Health Financing for Poor People Resource Mobilization and Risk Sharing Editors Alexander S. Preker Guy Carrin Other related titles by the editors Innovations in Health Service Delivery: The Corporatization of Public Hospitals. 2003. Washington, D.C.: World Bank. (Alexander S. Preker and April Harding, editors) Macroeconomic Environment and Health: With Case Studies for Countries in Greatest Need. 1993. Geneva: World Health Organization. (Guy Carrin, Michel Jancloes, and S. Ibi Ajayi, editors) Social ReInsurance: A New Approach to Sustainable Community Financing. 2002. Washington, D.C.: World Bank and International Labour Organization. (David M. Dror and Alexander S. Preker, editors) Strategies for Health Care Finance in Developing Countries: With a Focus on Community Financing in Sub-Saharan Africa. 1992. London: Macmillan Press Ltd. (Guy Carrin with Marc Vereecke, editors) Health Financing for Poor People Health Financing for Poor People Resource Mobilization and Risk Sharing Editors Alexander S. Preker and Guy Carrin THE WORLD BANK WORLD HEALTH ORGANIZATION INTERNATIONAL LABOUR OFFICE Washington, D.C. Geneva Geneva © 2004 The International Bank for Reconstruction and Development / The World Bank 1818 H Street, NW Washington, DC 20433 Telephone 202-473-1000 Internet www.worldbank.org E-mail feedback@worldbank.org All rights reserved. 1 2 3 4 07 06 05 04 The findings, interpretations, and conclusions expressed herein are those of the authors and do not necessarily reflect the views of the Board of Executive Directors of the World Bank or the govern- ments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The bound- aries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of the World Bank concerning the legal status of any terri- tory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this work is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. The World Bank encourages dissemination of its work and will normally grant permission promptly. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA, telephone 978-750-8400, fax 978-750-4470, www.copyright.com. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, World Bank, 1818 H Street, NW, Washington, DC 20433, USA, fax 202-522- 2422, e-mail pubrights@worldbank.org. ISBN: 0-8213-5525-2 Library of Congress Cataloging-in-Publication Data Health financing for poor people : resource mobilization and risk sharing / Alexander S. Preker. p. cm. Includes bibliographical references and index. ISBN 0-8213-5525-2 1. Poor--Medical care--Developing countries. 2. Public health--Developing countries--Finance. 3. Medical economics--Developing countries. 4. Minorities-- Medical care--Developing countries. 5. Human services--Developing countries-- Finance. I. Preker, Alexander S., 1951­ RA410.53.H437 2003 338.433621091724--dc21 2003057160 Contents Foreword xv Dean T. Jamison Preface xvii Acknowledgments xxiii PART 1 GLOBAL AND REGIONAL TRENDS 1 1. Rich-Poor Differences in Health Care Financing 3 Alexander S. Preker, Guy Carrin, David Dror, Melitta Jakab, William C. Hsiao, and Dyna Arhin-Tenkorang Overview and Context 4 Conceptual Underpinnings for Community-Based Action in Health Care Financing 9 Methodology for Assessing Impact, Strengths, and Weaknesses 14 Discussion of Main Findings from Background Reviews 24 Conclusions and Recommendations 41 Appendix 1A Statistical Data (Summary Tables) 44 2. Review of the Strengths and Weaknesses of Community Financing 53 Melitta Jakab and Chitra Krishnan Methods 54 What Is Community-Based Health Financing? 59 Performance of Community-Based Health Financing 68 Determinants of Successful Resource Mobilization, Social Inclusion, and Financial Protection 84 Concluding Remarks 92 Appendix 2A Performance Variables Reported in the Reviewed Studies 93 Appendix 2B Core Characteristics of Community Financing Schemes from the Review of the Literature 100 vi Contents 3. Experience of Community Health Financing in the Asian Region 119 William C. Hsiao What Is Community Financing? 122 A Summary of the Value Added by Types of Community-Financing Schemes 133 A Review of Selected Asian Community-Financing Schemes 135 4. Experience of Community Health Financing in the African Region 157 Dyna Arhin-Tenkorang Conceptual Framework 161 Evidence 171 Discussion 185 Conclusion 192 PART 2 COUNTRY CASE STUDIES USING HOUSEHOLD SURVEY ANALYSIS 199 5. Analysis of Community Financing Using Household Surveys 201 Melitta Jakab, Alexander S. Preker, Chitra Krishnan, Pia Schneider, François Diop, Johannes Paul Jütting, Anil Gumber, M. Kent Ranson, and Siripen Supakankunti Background 203 Methods 209 Results 216 Discussion 222 Concluding Remarks 225 Appendix 5A List of Reviewed Survey Instruments 226 6. Financial Protection and Access to Health Care in Rural Areas of Senegal 231 Johannes Paul Jütting Health Insurance in Rural Sub-Saharan Africa 232 Research Design and Methodology 235 Results 238 Conclusions 247 7. Community-Based Health Insurance in Rwanda 251 Pia Schneider and François Diop Background 253 Data Sources and Methodology 254 Results 258 Discussion and Conclusion 272 Contents vii 8. The SEWA Medical Insurance Fund in India 275 M. Kent Ranson Methods 277 Results 280 Discussion 285 9. The Potential Role of Community Financing in India 293 Anil Gumber Community Financing in India and the SEWA Program 295 Research Design and Methodology 301 Results 302 10. Impact of the Thailand Health Card 315 Siripen Supakankunti Methods 326 Discussion 335 Conclusions and Recommendations 352 PART 3 EXPENDITURE GAPS AND DEVELOPMENT TRAPS 359 11. Deficit Financing of Health Care for the Poor 361 Alexander S. Preker, John C. Langenbrunner, and Emi Suzuki Progress toward Achieving the MDGs 361 Key Drivers of Accelerated Progress toward Achieving the MDGs 366 Estimating the Cost of Achieving the MDGs 373 Financing the Expenditure Gap 384 Conclusions 389 12. Impact of Risk Sharing on the Attainment of Health System Goals 397 Guy Carrin, Riadh Zeramdini, Philip Musgrove, Jean-Pierre Poullier, Nicole Valentine, and Ke Xu Health System Goals and Functions in a Nutshell 398 The Organizational Form of Health Financing and Its Link to the Attainment of Health System Goals 399 Organization of Health Financing in the World 400 Modeling the Impact of the Organizational Form of Health Financing on Health System Attainment Indicators 401 Community Risk-Sharing Arrangements: Further Need to Measure Their Impact 410 Concluding Remarks 411 Appendix 12A Statistical Data 412 viii Contents About the Coeditors and Contributors 417 The Coeditors 417 Other Contributing Authors 418 Index 427 BOXES 1.1 Revenue Mobilization 26 1.2 Strengths of Community-Financing Schemes 28 1.3 Weaknesses of Community-Financing Schemes 30 2.1 Definitions of Community Health Financing 61 2.2 Contribution of CF Schemes to Operational Revenues 70 2.3 The Bamako Initiative 72 2.4 Turning Potatoes and Labor into Cash Revenues in Bolivia 86 2.5 Poor Management in the Nkoranza Scheme 89 4.1 Ghana's Policy Thrusts to Enable Evolution of Community Health Insurance 193 11.1 Millennium Development Goals (1990­2015) 362 FIGURES 1.1 Less Pooling of Revenues in Low-Income Countries 5 1.2 Flow of Funds through the System 5 1.3 Low-Income Countries Have Weak Capacity to Raise Revenues 6 1.4 Out-of-Pocket (OOP) Expenditure and Poverty without Risk Sharing 7 1.5 Pro-Rich Bias of Public Subsidies in Many Low-Income Countries 8 1.6 Determinants of Financial Protection, Health, and Social Inclusions 15 1.7 Hospitalization and Impoverishment 33 1.8 Stages of Financial Protection 42 2.1 Analytical Framework 58 3.1 Feasibility of Establishing Community Financing and the Amount the Average Person Is Willing to Pay as a Function of Expected Gains and Social Capital 124 3.2 Plausible Relationship between Locus of Control and Economic and Quality Gains 127 3.3 The Trade-Offs between Health Gains and Risk Protection by Type of Service Funded 133 4.1 Relationships between Stakeholders and the Scheme 167 4.2 Percentage of Community Enrolled, by Distance 179 4.3 Premiums, Participation, and Revenues: Predictions for Option C 187 4.4 Willingness to Pay for Adult Insurance 188 4.5 Stages of Financial Protection and Supporting Policies 191 Contents ix 6.1 Urban and Rural Health Insurance Schemes in Sub-Saharan Africa, Year of Inception and Size 233 7.1 Monthly Monetary Expenditure per Capita 262 11.1 Millennium Development Goals, Global Aggregate 363 11.2 Strong Correlation between Wealth and Health across Time 365 11.3 Income and Child Mortality 367 11.4 Cost-Effective and Affordable Public Health and Clinical Services 369 11.5 Three Nonfinancial Determinants of Good Outcomes 372 11.6 Lack of Management Capacity, Adverse Incentives, and Weak Institutions Break the Fulcrum 372 11.7 Expenditure Frontier and Six Countries 377 11.8 Production Frontiers for Total Expenditure on Health Care (Using Best Performance on Various Health Outcomes) 379 11.9 Production Frontiers for Public Expenditure on Health Care (Using Best Performance on Various Health Outcomes) 381 11.10 Income and Health Spending 385 11.11 Only 11 Percent of Global Spending for 90 Percent of the World's Population 386 11.12 Low-Income Countries Have Less Pooling of Revenues 387 11.13 Types of Private Financial Flows 388 TABLES 1.1 Conceptual Underpinnings of Community-Financing Schemes 10 1.2 Summary Statistics of the Literature Reviewed, by Publication Type 17 1.3 Core Characteristics of the Community-Based Financing Schemes 19 1.4 Characteristics of Five Survey Instruments 20 1.5 Types of Community-Based Financing 25 1.6 Number of Studies That Examined Core Health-Financing Subfunctions 25 1.7 Studies That Looked at Ways to Prevent Impoverishment 27 1.8 Studies That Looked at Ways to Combat Social Exclusion 27 1.9 Potential Value Added by Types of Community-Financing Schemes 32 1.10 Statistically Significant Determinants of Inclusion in Community Financing 36 1.11 Summary Findings: Statistically Significant Determinants of Utilization and Out-of-Pocket Expenditure Patterns 38 1.12 Estimation Results for the Basic Models 44 1.13 Estimation Results for the Enlarged Models 45 2.1 Summary Statistics of the Literature Reviewed, by Publication Type 55 x Contents 2.2 Summary of Literature Reviewed on Community-Based Health Financing Schemes, Based on Nature of Study and by Region 56 2.3 Selection Criteria to Assess the Performance of Community-Based Health Financing 59 2.4 Often Encountered Forms of Community Financing 65 2.5 Summary of Case Studies by Modalities 66 2.6 Cost Recovery from Prepaid Premiums 74 2.7 Summary of Findings: Who Is Covered by CF Arrangements? 75 2.8 Summary of Findings: Does CF Reduce the Burden of Seeking Health Care? 79 2.9 Determinants Associated with Effective Revenue Collection and Financial Protection 84 2.10 Performance Variables Reported in the Reviewed Studies 93 2.11 Core Characteristics of Community-Financing Schemes, from the Review of Literature 100 3.1 How Community Members Valued Service Availability, Quality, Risk Protection, and Costs 125 3.2 A Typology of Selected Characteristics of Community-Financing Schemes 129 3.3 Assessment of Potential Value Added by Selected Types of Community-Financing Schemes 134 3.4 Potential Value Added by Types of Community-Financing Schemes 136 3.5 Prevalence and Benefits of Community Health Financing in Five Provinces, 1991 139 3.6 Prevalence and Benefits of Community Health Financing in 30 Poor Counties, 1993 140 3.7 Management of Community Health Financing in 30 Poor Counties, 1993 140 3.8 Community Health Financing by Source in Selected Counties and Provinces, 1991 and 1993 143 3.9 Two Prototype Benefit Packages for China's Rural Poor 144 3.10 Current Financing of Health Spending by Source in China's Poverty Regions 145 3.11 Percentage of 2,236 Surveyed Community Leaders Citing Major Reasons for Lack of Rural Community Financing 145 4.1 Scheme Design Options 168 4.2 An Example of Goals Matched to Design Options 169 4.3 Features of "Potentially Large Population" Schemes for Informal Sector Households 172 4.4 "Potentially Large Population" Schemes' Financial Risk-Protection Performance 184 5.1 Socioeconomic Characteristics of Rwanda, Senegal, India, and Thailand, 1999 204 5.2 Health Outcomes and Expenditures in Rwanda, Senegal, India, and Thailand, 1999 209 Contents xi 5.3 Characteristics of Five Survey Instruments 210 5.4 Statistically Significant Determinants of Inclusion in Community Financing 217 5.5 Summary Findings: Statistically Significant Determinants of Utilization and Out-of-Pocket Expenditure Patterns 221 5.6 Reviewed Surveys for 21 Countries 227 6.1 Hospitalization Fees for Members and Nonmembers at St. Jean de Dieu Hospital 235 6.2 Selection Criteria for Mutual to Be Included in the Survey 235 6.3 Overview of Variables Used 239 6.4 Marginal Coefficients for Determinants of Participation in Mutual Health Insurance (Household Level) 240 6.5 Marginal Coefficients for Determinants of Participation in Mutual Health Insurance (Individual Level) 243 6.6 Probability of Hospitalization and Determinants of Expenditure in Case of Hospitalization 244 7.1 Community-Based Health Insurance in Rwanda, First Year Performance (July 1999­June 2000) 252 7.2 Summary Statistics: Independent Variables Used to Determine Probability of CBHI 256 7.3 Summary Statistics of Explanatory Variables for Probability of Professional Visit 257 7.4 Summary Statistics of Explanatory Variables for Estimated Out-of-Pocket Spending per Episode of Illness 259 7.5 Descriptive Sample Characteristics 260 7.6 Household Characteristics, by Income Quartile 261 7.7 Summary Statistics on Distribution of Monthly Monetary Expenditures per Capita 261 7.8 Logit Regression Results for Households' Probability to Demand Community-Based Health Insurance (Prepayment Schemes) 263 7.9 Probability of Using a Professional Provider by Insurance Status 266 7.10 Logit Regression Results for Probability of at Least One Professional Provider Visit for Members and Nonmembers 267 7.11 Average Health Expenditure per Sick Individual with or without a Visit to a Professional Provider, by Health Insurance Status and Expenditure Quartile, in Pilot Districts 268 7.12 Log-Linear Regression Results: Estimated Total Health-Related Expenditures per Episode of Illness for Sick Individuals with and without a Visit 270 7.13 Log-Linear Regression Results: Estimated Out-of-Pocket Health Expenditures per Episode of Illness for Sick Individuals Who Reported a Professional Visit in Past Two Weeks 271 8.1 Independent Variables Included in the Regression Analyses 279 8.2 Sample Characteristics 281 xii Contents 8.3 Hospital Utilization and Expenditure per Hospitalization by SEWA Coverage 282 8.4 Regression Results for Equation 1, the Odds of Being SEWA-INS Based on Sociodemographic Variables 283 8.5 Regression Results for Equation 2, the Probability of Being Hospitalized within the Last Year 284 8.6 Regression Results for Equation 3 286 8.7 Summary of Studies Looking at the Impact of CBHI Schemes on Utilization and Out-of-Pocket Expenditures 289 9.1 Type of Health Care Burden on Households Covered by Health Insurance Schemes 295 9.2 Salient Characteristics of Selected NGO-Managed Health Insurance Schemes 297 9.3 Prepayment and Insurance Mechanisms in Selected NGO-Managed Health Insurance Schemes 299 9.4 Coverage under SEWA Scheme 301 9.5 Overview of Variables Used 303 9.6 Determinants of Being Enrolled in SEWA Health Insurance Plan among SEWA Members 306 9.7 Determinants of Being Treated and Use of Public or Private Facility for Ambulatory Care 308 9.8 Determinants of Using Private Facility for Inpatient Care 310 9.9 Determinants of Out-of-Pocket Expenditure on Treatment by Type of Care 312 10.1 Health Service Utilization Pattern for Reported Ill Persons 316 10.2 Trend of Total per Capita Health Expenditure (Public and Private Spending) 317 10.3 Percent Source of Health Care Financing in 1984, 1986, and 1987 318 10.4 Budget Expenditures Classified by Program, Fiscal Years 1994­96 318 10.5 Modification of Health Card Program: Rationale, Objectives, and Activities 321 10.6 Unit Cost at Health Service Unit Used by Health Card Member 325 10.7 Demographic and Socioeconomic Characteristics of Health Card New Purchase and Health Card Nonpurchase 329 10.8 Demographic and Socioeconomic Characteristics of Health Card Dropout and Continued Card Purchase 330 10.9 Demographic and Socioeconomic Characteristics of Health Card Nonpurchase and Dropout Groups 331 10.10 Demographic and Socioeconomic Characteristics of Health Card Nonpurchase and Health Card Purchase (Dropout/Continued/New Purchase Groups) 332 10.11 Variables That Predict People Who Ever Purchased a Health Card Versus Nonpurchase Group 336 Contents xiii 10.12 Variables That Predict Health Card Nonpurchase and Dropout Groups Versus Continued and New Card Purchase Groups 337 10.13 Variables That Predict Health Card Nonpurchase Versus New Card Purchase Group 338 10.14 Variables That Predict New and Continued Health Card Purchase Versus Health Card Dropout Group 339 10.15 Demographic and Socioeconomic Characteristics of Health Care Seekers by Whether or Not They Possess Health Card 341 10.16 Frequency of Card Use among Card User Group 343 10.17 Health Care­Seeking Pattern in the Past Three Months among Card Users and Noncard Users 343 10.18 Multiple Regression Model 344 10.19 Logistic Regression Model: Variables that Predict Health Card Usage among Card Users 345 10.20 Attitudes toward Health Card Program of Cardholders and Noncardholders 346 10.21 Attitudes toward Health Card Program of Cardholders and Noncardholders (Mean Score) 347 10.22 Health Card Utilization Rate 348 11.1 Public Expenditure on Health Needed to Be More Like Frontiers 378 11.2 Total Health Expenditure Gap in US$ (in Millions) to Achieve Frontier Expenditure Levels (Countries with GDP < US$7,000 per Capita) 383 11.3 Public Health Expenditure Gap in US$ (in Millions) to Achieve Frontier Expenditure Levels (Countries with GDP < US$7,000 per Capita) 383 12.1 Descriptive Statistics (Full Samples) 402 12.2 Descriptive Statistics (Restricted Samples) 404 12.3 Classification of Countries by Degree of Risk Sharing in the Health-Financing System 412 12.4 Estimation Results for the Basic Model 413 12.5 Estimation Results for the Enlarged Models (Summary) 414 Foreword I n January 2000, Dr. Gro Harlem Bruntland, the former director general of the World Health Organization (WHO), established a Commission on Macroeco- nomics and Health (CMH) to provide evidence on the importance of health to economic development and poverty alleviation. This book is based on research undertaken for the Commission's Working Group 3. The mandate of Working Group 3 was to examine alternative approaches to domestic resources mobilization, risk protection against the cost of illness, and efficient use of resources by providers. Professor Alan Tait (former deputy director of Fiscal Affairs, International Monetary Fund, and currently honorary fellow at University of Kent at Canterbury and honorary fellow at Trin- ity College, Dublin) and Professor Kwesi Botchwey (director of Africa Research and Programs at the Harvard Center for International Development) chaired the group. Professor Jeffrey Sachs (then chairman of the Commission and director of the Harvard Center for International Development) presented the Commission's findings in a report submitted to the WHO on December 20, 2001--Macroeco- nomics and Health: Investing in Health for Economic Development. The Summary Report from the Commission recommended a six-pronged approach to domestic resource mobilization at low-income levels: "(a) increased mobilization of general tax revenues for health, on the order of 1 percent of GNP by 2007 and 2 percent of GNP by 2015; (b) increased donor support to finance the provision of public goods and to ensure access for the poor to essential health services; (c) conversion of current out-of-pocket expenditure into prepay- ment schemes, including community-financing programs supported by public funding, where feasible; (d) a deepening of the HIPC initiative, in country cover- age and in the extent of debt relief (with support from the bilateral donor com- munity); (e) effort to address existing inefficiencies in the way in which government resources are presently allocated and used in the health sector; and (f) reallocating public outlays more generally from unproductive expenditure and subsidies to social-sector programs focused on the poor." Most community-financing schemes have evolved in the context of severe economic constraints, political instability, and lack of good governance. Usually, government taxation capacity is weak, formal mechanisms of social protection for vulnerable populations absent, and government oversight of the informal health sector lacking. In this context of both public sector failure and market failure, community involvement in the financing of health care provides a criti- cal, albeit insufficient, first step in the long march toward improved access to health care by the poor and social protection against the cost of illness. xvi Foreword The Commission stressed that community-financing schemes are no panacea for the problems low-income countries face in resource mobilization. Instead, the Commission recommended that such community-based financing mecha- nisms be regarded as a complement to--not a substitute for--strong government involvement in health care financing and risk management related to the cost of illness. The key conclusions on community financing from Working Group 3 of the Commission on Macroeconomics and Health summarized in this book make a valuable contribution to our understanding of some of the strengths, weak- nesses, and policy options for securing better access for the poor to health care and financial protection against the impoverishing effects of illness, especially for rural and informal sector workers in low-income countries. Dean T. Jamison Professor School of Public Health Center for Pacific Rim Studies University of California Los Angeles (UCLA) Fellow Fogarty International Center National Institutes of Health Preface O ne of the most urgent and vexing challenges faced by many low- and middle-income countries is how to provide health care for the more than 1.3 billion poor people who live in rural areas or work in the informal sec- tor. As pointed out by Bill Hsiao from Harvard University in the chapter on the Asia region, this population is not a homogeneous group. Their occupations range from farmers, peddlers, day laborers, taxi drivers, and employees of the informal sector to shop owners and self-employed professionals. Yet this hetero- geneous group shares the same lack of access to health care that is often due to inadequate health care financing. This book focuses on how to mobilize finan- cial resources to pay for health care for such residents of rural communities in low-income countries. It also gives some attention to mobilizing health care financing for the urban poor. Most countries try to serve their rural populations by directly operating pub- lic clinics in rural areas, but it is often difficult to get qualified practitioners to staff them. Those who accept such postings frequently work sporadically and provide poor quality services. The facilities themselves often lack drugs and sup- plies. When individuals become ill, they are frequently forced to rely first on self-treatment with home remedies provided by traditional healers and pharma- cists. For serious illness episodes, the majority ultimately seek care from the few public and charity hospitals located in the rural areas. Patients often have to pay a formal copayment or informal charge when treated in hospitals, even in the public sector. As a result, many patients have to choose between bankrupting their families and purchasing needed treatment. Studies have found that higher proportions of women and children than men have to forgo medical treatments. In addition, studies consistently have found that even when the government provides free or nearly free services, poor house- holds pay a significant part of their income in informal charges. As much as 80 percent of total health care expenditure in low-income countries comes from direct out-of-pocket payment by patients. Studies in several countries found that large medical expenditure (such as for inpatient hospital services and costly out- patient drugs) is a major cause of poverty. These observations raise three serious sets of questions: First, do countries spend enough on health care? In many countries the answer is no, particularly in the case of health care expenditure for the poor. However, it is not always certain that governments can spend more. Most low- income countries have narrow tax bases and ineffective tax collection systems. The total amount of money mobilized through taxes is therefore limited. Com- peting demands for the scarce general government resources that are available xviii Preface often leaves little public funding for basic health care for the poor rural and urban households. Most developed countries use general revenues and social health insurance to pay for and provide health care for citizens working in rural areas and the informal sector. As will be seen in chapter one, the feasibility of these approaches may be weak in many low-income countries, as there are several factors that can hamper the move toward universal coverage. Private health insurance frequently is not affordable to the poor. User fees are inequitable and create a high barrier to access to health care by the poor. As for foreign aid, it is often small, even in low-income countries, compared with total spending on health care. Second, do countries have a capacity to transform the little money avail- able into effective services for the poor living in rural areas or working in the informal sector of urban centers? In many countries in which the government intends to fund and provide free, or nearly free, services for the rural residents and the poor, the target population is not utilizing the publicly provided health services. Why is this happening? Detailed studies in low-income coun- tries have consistently found that governments are inefficient in their funding of primary care at the village and township levels. Public funds usually sup- port the salaries of health workers regardless of whether they are delivering satisfactory services, while funds allocated to the purchase of drugs and sup- plies are inadequate. Consequently, this practice creates a public employment program rather than an effective health care delivery system. It thus turns out that the so-called free services may actually become expensive, as patients have to pay for drugs and medical consumables directly out-of-pocket. Fur- thermore, governments, in general, do not manage or monitor public services adequately at the local level. As a result, when the poor become ill they often choose to use their limited income to consult private practitioners and buy their own drugs. Third, is the money spent directly by households used in an efficient and cost-effective way? We know that the answer to this question, too, is often no. Out-of-pocket payment for private sector providers has some serious drawbacks. Because these resources are not channeled through collective purchasing arrange- ments, individual households seeking health care are frequently in a weak bar- gaining position against providers who can extract above-market prices due to their monopoly power. This is exacerbated at the village level, where the small population size means that the presence of multiple providers competing with each other to keep prices low is particularly unlikely. Throughout the world, community financing has been used to mobilize resources to fund and deliver health care for the poor in rural and urban com- munities, in settings where governments failed to fully meet this responsibility through the public sector. Some of these community-financing schemes have successfully addressed all three issues discussed above while others are primarily income-generating schemes for providers. Preface xix KEY FINDINGS Based on an extensive survey of the literature, the main strengths of community- financing schemes are the degree of outreach penetration achieved through community participation and the contribution to financial protection against the cost of illness for low-income rural and informal sector workers. The schemes' main weaknesses relate to both external and design factors. Often the level of revenues that can be mobilized from poor communities is low. As a result, without some form of subsidy the poorest of the poor are frequently excluded from participation in such schemes. The small size of the risk pool of many voluntary community schemes, the limited management capacity that exists in rural and low-income contexts, and the isolation of such schemes from formal health-financing mechanisms and provider networks are all major weak- nesses that must be addressed. The review of the literature provided a number of insights into the policy and institutional capacity-building measures that can be used to address many of these issues. The review of selected experiences in the Asia and Africa regions supported many of these conclusions. It emphasized the diversity of community-financing arrangements that exist there. Several of the schemes appear to improve finan- cial protection against the cost of illness, allow better access by poor households to essential health care, and confer greater efficiency in the collection, pooling, management, and use of scarce health care resources. The existence of risk-sharing arrangements, as well as trust and local commu- nity control over the schemes, appears to increase enrollment rates with such schemes. In particular, the literature emphasized that, although income is a key constraint to participation by the poorest of the poor, even they are often willing and able to participate if their contributions are subsidized by public or donor funds and if the benefits they receive provide access to quality services. House- holds were also more likely to enroll in these schemes when the households that would later use them were directly involved in their design and management. Other factors that increased the likelihood of enrollment included setting the contribution level based on an assessment of local ability and willingness to pay, and ensuring the availability of easy access to the health care providers who serve the members. Members like broad coverage that includes basic health services for frequently encountered health problems as well as hospitalization for rarer and more expensive conditions. In the context of extreme resource constraints, this creates a tension or tradeoff between prepayment for basic services and the need for insurance coverage for more expensive, life-threatening events that may only happen once or twice in a lifetime. This observation is consistent with the expe- rience in other areas of insurance, in which willingness to pay for rare cata- strophic events (life insurance) is often significantly reduced compared with coverage for events more likely to happen at a greater frequency (crop insur- ance). This highlights an area of market failure relating to voluntary community xx Preface involvement in health care financing that needs to be addressed by appropriate government policies, because it is precisely during hospital episodes that many of the poor become severely impoverished. The review of selected experiences from the Asia and Africa regions also pointed to a number of measures governments could take to strengthen com- munity financing. This included subsidizing the contributions for the poor, pro- viding technical assistance to improve a scheme's management capacity, and establishing links with formal health care networks. Satisfaction with the scheme was often related to the nature of the direct community involvement in design and management. A critical factor was matching willingness and ability to pay with the expectation of benefits to be received at some later point. The review also highlighted areas of government actions that appear to have nega- tive impacts on the function of community-financing schemes. Top-down inter- ference with the design and management of the schemes appeared to have a particularly negative impact on function and sustainability. The results of the microlevel household data analysis reinforced the conclu- sions from the survey of the literature and two regional reviews. Econometric analysis of household data from four countries indicated that prepayment and risk sharing through community involvement in health care financing--no mat- ter how small--increases access by poor populations to basic health services and protects them to a limited extent against the impoverishing effects of illness. Community involvement alone is not sufficient in preventing social exclusion since the poorest of the poor often do not participate fully in these schemes. However, the analysis provided evidence that this constraint in reaching the poorest could be overcome through well-targeted design features and implemen- tation arrangements. Finally, the results of the macrolevel cross-country analysis presented in this book give empirical support to the hypothesis that broad risk sharing in health- financing matters in terms of impact on both the level and the distribution of health, financial fairness, and responsiveness indicators. The results even sug- gested that risk sharing corrects for, and may outweigh, the negative effect of overall income inequality, suggesting that financial protection against the cost of illness may be a more effective poverty alleviation strategy in some settings than direct income support. CONCLUSIONS The underlying causes of many of today's health problems in lower-income coun- tries are often well known, and effective and affordable drugs, surgical procedures, and other interventions often exist. But because of a number of problems related to resource mobilization, risk sharing, and resource allocation and purchasing arrangements, as well as problems in the provision of goods and services to rural Preface xxi and low-income populations, potentially effective policies and programs fre- quently fail to reach the households and communities that need them the most. The research on community financing undertaken for Working Group 3 of the Commission on Macroeconomics and Health emphasized the importance of general tax revenues and payroll tax-based social health insurance contributions to the financing of health care at higher income levels. These methods can be equitable and efficient in mobilizing and utilizing resources. However, most community-financing schemes have evolved in settings with severe economic constraints, political instability, lack of good public sector governance, and cata- strophic out-of-pocket user charges that can lead to impoverishment. These con- ditions are very different from those enjoyed at higher income levels, in which public-financing instruments have been successful in financing health care. For years, many low- and middle-income countries--with assistance from the international development community--have tried to jump from no organized financing instruments to full reliance on financing through general taxation, social health insurance, or both. In the context of large rural populations, low for- mal labor market participation rates, and the limited scope of the above-men- tioned formal health financing methods, few have succeeded on this reform path. This book highlights the fact that community financing provides a more incre- mental, first step in the transition toward improved financial protection against the cost of illness and better access to priority health services for the 1.3 billion poor people in low- and middle-income countries. Community financing is not presented as a panacea for financing health care for rural and low-income workers in the informal sector. Rather, it is one of several options that can be considered by low-income countries in expanding coverage for the poor. The book highlights several concrete public policy measures that governments can introduce to strengthen and improve the effectiveness of community involvement in health care financing. These include (a) increased and well- targeted subsidies to pay for the contributions of low-income populations; (b) use of insurance to protect against health care costs and assessment of the feasibility of reinsurance to enlarge the effective size of small risk pools; (c) use of effective prevention and case management techniques to limit expenditure fluctuations; (d) technical support to strengthen the management capacity of local schemes; and (e) establishment and strengthening of links with the formal financing and provider networks. Acknowledgments T his book is based on work submitted to Working Group 3 (Chairmen Alan Tait and Kwesi Botchwey) of the Commission on Macroeconomics and Health (Chairman Jeffrey D. Sachs). The authors are grateful to the World Health Organization for having provided an opportunity to contribute to the work of the Commission on Macroeconomics and Health and to the World Bank for having published the background reports on community financing as HNP discussion papers. The following individuals contributed directly to the book: (a) preparation of the synthesis book by Alexander S. Preker, Guy Carrin, David Dror, Melitta Jakab, William C. Hsiao, and Dyna Arhin-Tenkorang; (b) survey of the literature by Melitta Jakab and Chitra Krishnan; (c) analysis of macrolevel data by Guy Carrin, Riadh Zeramdini, Philip Musgrove, Jean-Pierre Poullier, Nicole Valentine, and Ke Xu; (d) analysis of microlevel data by Melitta Jakab, Alexander S. Preker, Chitra Krishnan, Allison Gamble Kelly, Pia Schneider, François Diop, A. K. Nan- dakumar, Johannes Paul Jütting, Anil Gumber, M. Kent Ranson, and Siripen Supakankunti; (e) review of selected Asian and African experiences by William C. Hsiao and Dyna Arhin-Tenkorang; and (f) review of reinsurance of community schemes by David Dror and Alexander S. Preker. Valuable guidance on methodological issues was provided by Adam Wagstaff. We are also indebted to the following individuals for data access, guidance on research methodologies, reviews, and other indirect contributions to the book: Christian Jacquier, Christian Baeaza, Michael Cichon, Chris Murray, Kei Kawa- batak, Christopher Lovelace, Helen Saxenian, Davidson Gwatkin, David Peters, George Schieber, Charlie Griffin, Agnes Soucat, Abdo S. Yazbeck, Mariam Clae- son, Flavia Bustreo, Steve Cummings, and Shanta Devarajan. The authors of the book are also grateful for the access provided to parallel and ongoing research on community financing by the World Bank, the World Health Organization, and the International Labour Organization, with impor- tant input from Harvard University, the London School of Hygiene and Tropical Medicine, the University of Lyon, Abt Associates, Inc. (Partnerships for Health Reform USA), the National Council for Economic Research (India), the Center for Development Research (ZEF) (Germany), and the Chulalongkorn University Faculty of Economics (Thailand). PART 1 Global and Regional Trends 1. Rich-Poor Differences in Health Care Financing Alexander S. Preker, Guy Carrin, David Dror, Melitta Jakab, William C. Hsiao, and Dyna Arhin-Tenkorang 2. Review of the Strengths and Weaknesses of Community Financing Melitta Jakab and Chitra Krishnan 3. Experience of Community Health Financing in the Asian Region William C. Hsiao 4. Experience of Community Health Financing in the African Region Dyna Arhin-Tenkorang CHAPTER 1 Rich-Poor Differences in Health Care Financing Alexander S. Preker, Guy Carrin, David Dror, Melitta Jakab, William C. Hsiao, and Dyna Arhin-Tenkorang Abstract: Most community finance schemes have evolved in the context of severe eco- nomic constraints, political instability, and lack of good governance. Usually govern- ment taxation capacity is weak, formal mechanisms of social protection for vulnerable populations absent, and government oversight of the informal health sector lacking. In this context of extreme public sector failure, community involvement in financing health care provides a critical, though insufficient, first step in the long march toward improved health care access for the poor and social protection against the cost of illness. It should be regarded as a complement to--not a substitute for--strong government involvement in health care financing and risk management related to the cost of illness. Based on their extensive survey of the literature, the authors show that the main strengths of community-financing schemes are the extent of outreach penetration achieved through community participation, the contribution to financial protection against illness, and the increase in access to health care by low-income rural and informal sector workers. The schemes' main weaknesses are the low volume of revenues that can be mobilized from poor communities, the frequent exclusion of the very poorest from participation in such schemes without some form of subsidy, the small size of the risk pool, the limited management capacity existing in rural and low-income contexts, and the isolation from the more comprehensive benefits often available through more formal health-financing mechanisms and provider networks. The authors conclude by propos- ing concrete public policy measures that governments can introduce to strengthen and improve the effectiveness of community involvement in health care financing. These include: (a) increased and well-targeted subsidies to pay for the premiums of low-income populations; (b) use of insurance to protect against expenditure fluctuations and use of reinsurance to enlarge the effective size of small risk pools; (c) use of effective prevention and case management techniques to limit expenditure fluctuations; (d) technical support to strengthen the management capacity of local schemes; and (e) establishment and strengthening of links with the formal financing and provider networks. T his century has witnessed greater gains in health outcomes than any other time in history. These gains are partly the result of improvements in income that have been accompanied by improvements in health-enhanc- ing social policies (housing, clean water, sanitation systems, and nutrition) and greater gender equality in education. They are also the result of new knowledge about the causes, prevention, and treatment of disease and the introduction of policies, financing, and health services that make such interventions more equi- tably accessible. Improving ways to finance health care and protect populations against the cost of illness has been central to this success story (see Preker, Lan- genbrunner, and Jakab 2002; Preker and others 2002a, 2002b). 4 Health Financing for Poor People: Resource Mobilization and Risk Sharing OVERVIEW AND CONTEXT The share of the world's population protected against the catastrophic cost of ill- ness rose significantly during the twentieth century, with global spending on health increasing from 3 percent to 8 percent of global gross domestic product (US$2.8 trillion), or 4 percent of the GDP of developing countries (US$250 bil- lion). At the current global growth rate for GDP of 3.5 percent, spending on health-enhancing activities will increase annually by about $98 billion a year worldwide, or $8 billion a year in low- and middle-income countries. The Exclusion of Low-Income Rural Populations and Informal Workers Today the populations in most industrial countries (except Mexico, Turkey, and the United States) enjoy universal access to a comprehensive range of health ser- vices that are financed through a combination of general tax revenues, social insurance, private insurance, and charges (Preker 1998). A number of low-income countries (such as Costa Rica, Malaysia, Sri Lanka, and Zambia) have tried to follow a similar path, but the quest for financial pro- tection against the cost of illness in low- and middle-income countries has been a bumpy ride. Many of the world's 1.3 billion poor still do not have access to effective and affordable drugs, surgeries, and other interventions because of weaknesses in the financing and delivery of health care (ILO 2000a; WHO 2000; World Bank 1993, 1997). See figure 1.1. Although 84 percent of the world's poor shoulder 93 percent of the global burden of disease, only 11 percent of the $2.8 trillion spent on health care reaches the low- and middle-income countries. Vaccination strategies of mod- ern health care systems have reached millions of poor. However, when ill, low-income households in rural areas continue to use home remedies, tradi- tional healers, and local providers who are often outside the formal health system. The share of the population covered by risk-sharing arrangements is smaller at low-income levels (see figure 1.1). As a result, the rich and urban middle classes often have better access to the twenty-first century's health care advances. Origins of Rich-Poor Differences in Financial Protection The flow of funds through the health care system, and the public-private mix, is complex (see figure 1.2--modified from Schieber and Maeda 1997). It can be dif- ferentiated into three discrete functions: (a) collection of revenues (source of funds), (b) pooling of funds and spreading of risks across larger population groups, and (c) purchase of services from public and private providers of health services (allocation or use of funds) (see also WHO 2000). A combination of gen- eral taxation, social insurance, private health insurance, and limited out-of- pocket user charges has become the preferred health-financing instruments for FIGURE 1.1 Less Pooling of Revenues in Low-Income Countries Pooled health revenues as % of total 100 80 Share of the world's 1.3 billion Indonesia living on less than US$1/day 60 indicated by size of the bubbles China Mexico 40 Pakistan 20 India 0 ­10,000 0 10,000 20,000 30,000 40,000 50,000 GNP per capita FIGURE 1.2 Flow of Funds through the System Resource Service Revenue allocation or provision collection Pooling purchasing Taxes Government agency Public charges Social insurance or Mandates Public Public sickness funds providers Grants Private insurance organizations Loans Private Private providers insurance Employers Private Communities Individuals and Out-of-pocket households 5 6 Health Financing for Poor People: Resource Mobilization and Risk Sharing middle- and higher income countries, where income is readily identifiable and taxes or premiums can be collected at the source. Different issues arise in the cases of public and private engagements in health care financing and service delivery. The need for collective arrangements and strong government action in health care financing is often confused with public production of services. The poor and other excluded populations frequently seek care from private providers because public services in rural and low-income urban areas are often scarce or plagued by understaffing, supply shortages, and low-quality care. Poor households and community-financing schemes therefore often turn to private providers for the care they need. Private provider engage- ment can still be pro-poor if there are mechanisms to exempt the poor or subsi- dize user fees (Preker, Harding, and Girishankar 2001) and if purchasing arrangements include coverage for the poor (Preker and others 2001). Several factors make the policy options for financing health care at low-income levels different from financing those at higher income levels. Low-income coun- tries often have large rural and informal sector populations, limiting the taxation capacity of their governments (see figure 1.3--modified from World Bank 1997). When a country's taxation capacity is as low as 10 percent of GDP or less, it would take 30 percent of government revenues to meet a target of 3 percent of GDP health expenditure through formal collective health care financing channels. In most countries, public expenditure on health care is much lower than this, often not surpassing 10 percent of public expenditure, which means that less than 1 per- cent of GDP of public resources is available for the health sector. A related set of problems is faced during the pooling of financial resources at low-income levels. Pooling requires some transfer of resources from rich to poor, healthy to sick, and gainfully employed to inactive. In low-income countries, tax evasion by the rich and middle classes in the informal sector is widespread, allowing higher income groups to avoid contributing their share to the overall FIGURE 1.3 Low-Income Countries Have Weak Capacity to Raise Revenues Total government revenues as % GDP 100 · Governments in many countries often 80 raise less than 20% of GDP in public revenues; and 60 40 · The tax structure in many low-income countries is often regressive. 20 0 100 1,000 10,000 100,000 Per capita GDP (log scale) Rich-Poor Differences in Health Care Financing 7 revenue pool. Without such pooling of revenues and sharing of risks, low- income populations are exposed to serious financial hardship at times of illness (Diop, Yazbeck, and Bitran 1995). Figure 1.4 (Wagstaff, Watanabe, and van Doorslaer 2001) indicates households whose income drops below the poverty line (horizontal bar indicates poverty line) because of out-of-pocket expenditure on health care (vertical drop bars on the income distribution curve). Any pool- ing that does occur tends to be fragmented along income levels, preventing effective cross-subsidies between higher and lower income groups. In many poor countries, local community-financing schemes have emerged partially as an informal sector response to these shortcomings in revenue pooling at low- income levels. Faced with overwhelming demand and very limited resources, many low- income countries use nonspecific broad expenditure caps that push rationing and resource allocation decisions to lower levels of the provider system. This often leads to serious drug shortages, equipment breakdowns, capital stock depreciation, and the lowering of hygiene standards. Such an environment also means politically and ethically difficult rationing decisions about the targeting of public expenditure to the poor. As a result of such difficulties, the rich often benefit more from public subsidies and public expenditures than the poor (figure 1.5--Peters and others 2001; see also Gwatkin 2001). FIGURE 1.4 Out-of-Pocket (OOP) Expenditure and Poverty without Risk Sharing HH expenditure as multiple of PL 10 Vietnam 9 8 7 6 5 4 3 2 1 0 1 500 999 1498 1997 2496 2995 3494 3993 4492 4991 5490 5989 Poverty line = 1789870 dongs/day Pre-OOP HH income Post-OOP HH income 8 Health Financing for Poor People: Resource Mobilization and Risk Sharing FIGURE 1.5 Pro-Rich Bias of Public Subsidies in Many Low-Income Countries Concentration index 0.50 0.40 0.30 Pro-rich distribution 0.20 0.10 0.00 ­0.10 Pro-poor distribution ­0.20 ­0.30 h t lia India Peru Rica China Nadu Brazil Chile GuineaBihar Ghana Kenya AfricaKerala alaysia RajasthanPrades BulgariaPradesh Gujara Karnataka Indonesia stBengalVietnam Tamil Mongo HondurasGuyanaColombia M Costa ArgentinaUruguay South Uttar Madagascar We Andhra It has been less difficult for national policymakers to design effective health- financing schemes for individuals and households in formal employment whose income is readily identifiable and who can be taxed at the source. Unfortunately, the formal sector in most low-income countries is small in comparison with populations in rural areas and informal employment. In low-income countries, large segments of the population in informal employment remain without effec- tive collective arrangements to pay for health care or to protect them from the costs of illness (Guhan 1994; Midgley and Tracey 1996; Van Ginneken 1999; World Bank 1995). Role of Communities in Providing Financial Protection Community initiatives have recently begun to bridge the large gap in social pro- tection between people covered by formal schemes and those with no protection at all against the cost of illness who are exposed to the impoverishing effects of user charges. (Arhin-Tenkorang 1994, 1995, 2000; Atim 1998, 1999; Bennett, Creese, and Monasch 1998; Jakab and Krishnan 2001; Musau 1999; Ziemek and Jütting 2000). In the literature, the term community financing has evolved into a generic expression used to cover a large variety of health-financing arrangements (Abel- Rich-Poor Differences in Health Care Financing 9 Smith 1988; Dror and Jacquier 1999; Foster 1982; Hsiao 2001; Navarro 1984). On one hand, different authors use the term community financing in different ways. On the other hand, similar--more specific--terms are often used to describe sim- ilar financing arrangements. Microinsurance, community health funds, mutual health organizations, rural health insurance, revolving drug funds, and community involvement in user fee management have all been referred to as community-based financing. Yet each of these risk-sharing arrangements has different objectives, policies, and management, organizational and institutional characteristics, and different strengths and weaknesses. The Oxford English Dictionary defines community as (a) "joint or common, ownership, tenure or liability"; (b) "common character"; (c) "social fellow- ship"; (d) "life in association with others"; (e) "common or equal rights or rank"; and (f) "people organized into common political, municipal or social unity." Community-based health care financing reflects most of these concepts. One common feature of the definitions is the predominant role of collective action in raising, pooling, allocating or purchasing, and supervising the management of health- financing arrangements, even when there is interface with government pro- grams and services in terms of subsidies, supplemental insurance coverage, or access to public provider networks. Some community-financing schemes cover common geographic entities, while others are based on professional affiliations, religion, or some other joint activity. A second common feature relates to the beneficiaries of these schemes, who tend to be populations with no other finan- cial protection or access to collective financing arrangement to cover the cost of health care. A third common feature is the voluntary nature of these schemes and the tradition of self-help and social mobilization embraced by the poor in many low-income countries. CONCEPTUAL UNDERPINNINGS FOR COMMUNITY-BASED ACTION IN HEALTH CARE FINANCING If both markets and governments fail to provide financial protection mecha- nisms for the poor, what is it about community-based initiatives that makes the poor turn to such arrangements? The growth of community-based health- financing arrangements rests on developments in three related areas (see table 1.1 and Dror, Preker, and Jakab 2002): · Microfinance (microcredits, microsavings, microinsurance, financial interme- diation) · Social capital (community, network, institutional, and societal links) · Mainstream theories (welfare of society, public finance, social policy, and health policy). 10 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 1.1 Conceptual Underpinnings of Community-Financing Schemes Key conceptual underpinnings Microfinance 1. Microcredits Risk taking (take advantage of opportunity, avoid overcautious behavior) Current liquidity management (smooth out consumption, increase choice) Short-term shocks (drought, famine) 2. Microsavings Predictable life cycle events (education, marriage dowry, childbirth, death) Capital formation (purchase of equipment, down payment on land, growth) Future liquidity management (smooth consumption, increase choice) 3. Microinsurance Long-term income support (life and disability insurance, pensions) Short-term income support (sick pay, unemployment insurance--not well developed) Unpredictable health expenditure (health insurance) Replacement of loss (fire and theft insurance) 4. Financial intermediation Payment and money-transfer services (facilitate trade and investments) Social capital 1. Community links Between extended families, local organizations, clubs, associations, civic groups 2. Network links Between similar communities (horizontal) and different communities (vertical) 3. Institutional links To communities' political, legal, and cultural environments 4. Societal links Between governments and citizens through public-private partnerships and community participation Mainstream theories 1. Welfare of society Income and growth 2. Public finance Taxation and social insurance 3. Social policy Social services and safety nets 4. Health policy Public health priorities and health systems Links to Existing Microfinance Organizations The role of microfinance in poverty alleviation for low-income groups has become a prominent theme in recent years (ADB 2000; Brown and Churchill 2000; Otero and Rhyne 1994; Zeller and Sharma 2000). Poor and rich house- holds are equally exposed to a range of events that put them at financial risk and are beyond their immediate control. Such events range from predictable life cycle events, such as marriage, childbirth, education, and death, to less pre- dictable events, such as droughts, fire, floods, and catastrophic illness. Rich-Poor Differences in Health Care Financing 11 The difference between poor and nonpoor households is the availability of mechanisms to cope with the financial consequences of unpredictable events. Nonpoor households take advantage of a wide range of risk-protection mecha- nisms that are available even in the lowest income countries. This includes sav- ings, access to credit, insurance, and other financial intermediation mechanisms. Until recently, few risk-protection mechanisms were accessible to the poor. It was assumed that the poor--living on less than a dollar a day--were neither will- ing nor able to save or contribute to insurance against the risks they faced. In sum, the poor were thought to be "unbankable" and "uninsurable" (Zeller and Sharma 2000). This led to the growth of informal risk-protection mechanisms through families, friends, and community networks. However, the past decade has witnessed a steady expansion of successful initiatives to provide the poor with savings, credit, and insurance services. Growing experience with these mechanisms suggests that the poor can be creditworthy, can save, and can buy insurance. In particular, four microfinance instruments have been developed to improve the productive needs of low-income households. They are (a) credits that help improve the immediate human, physical, and social capital of the poor (for example, small short-term loans to help pay for training, a piece of farm equipment, and access to social networks); (b) savings to be used to build up the medium-term capital of the poor, such as education, the down payment on a piece of land, and dowry for the marriage of a daughter into a good family; (c) insurance to stave off unpredictable expenses, such as theft, loss, and ill- ness); and (d) financial intermediation (payment systems to facilitate trade and investments). Life, casualty, and crop insurance is often used to secure loans for low-income populations. Microfinance instruments help the poor avoid having to invest in less cost-efficient means of saving, credit, and insurance such as jewelry, live- stock, and staple food, or to resort to inefficient barter systems of payment (pay- ment in-kind). These instruments also contribute to the early transformation of barter transactions into more formal economic exchange and formalization of property rights. The extension of such techniques to the health sector is now being observed in many microfinance and development organizations in low-income countries, especially in the case of microinsurance (Brown and Churchill 2000; Dror and Jacquier 1999; ILO 2000b, 2001). Extending microinsurance techniques to health care presents a unique set of challenges under exploration. While life and crop insurance deals mainly with the financial cost of income loss, health insur- ance presents an additional set of issues related to financing tangible services for which the cost is neither fully predictable nor constant. This includes the range and severity of different illnesses, the range and scope of services provided, and the behavior of patients and providers (the latter influenced particularly by the payment mechanism due to moral hazard, adverse selection, and fraud, espe- cially in the form of supplier-induced demand). 12 Health Financing for Poor People: Resource Mobilization and Risk Sharing Links to Community-Level Social Capital Why have microfinance organizations been able to reach low-income individu- als and households while more formal national systems have failed to do so? Clues to the answer come from the social capital literature of the 1990s, which can be summed up as "it is not what you know, but whom you know" (Platteau 1994; Woolcock 1998; Woolcock and Narayan 2000). When hard times strike, it is often family and friends who constitute the ultimate safety net for low-income groups. Evidence suggests that social capital has four dimensions with potentially positive and potentially negative impacts on development. The four dimensions include: · Community links such as those between extended families, local organiza- tions, clubs, associations, and civic groups--people in small communities helping each other (Dordick 1997) · Network links between similar communities (horizontal) and between differ- ent communities (vertical), such as ethnic groups, religious groups, class structures, and genders (Granovetter 1973) · Institutional links such as those between communities' political, legal, and cultural environments (North 1990) · Societal links between governments and their citizens through complemen- tarity and embeddedness, such as public-private partnerships and the legal framework that protects the rights of association (for example, chambers of commerce and business groups) and community participation in public orga- nizations (for example, community members on city councils and hospital boards) (Evans 1992, 1995, 1996). Low-income households are likely to have greater trust in microhealth insur- ance programs that are linked to the community credit, savings, and insurance organizations to which they already belong and over which they feel they have some control. The people often regard national systems as impersonal and dis- tant and think they will never benefit from those programs. This view is rein- forced when the national programs ration care to focus on "global" public health priorities that--although they may have large externalities and benefits to society as a whole--often do not respond to the poor's immediate day-to-day health care needs. Such social capital has both benefits and costs. The downside of social capital occurs when communities and networks become isolated or parochial or work at cross-purposes to societal collective interests (for example, ghettos, gangs, car- tels). Intercommunity ties or bridges are needed to overcome the tendency of communities and networks to pursue narrow, sectarian interests that may run counter to broader societal goals. (Narayan 1999) Community-financing schemes are vulnerable to a number of the shortcomings associated with social capital: Rich-Poor Differences in Health Care Financing 13 · Community-financing schemes that share risk only among the poor will deprive its members of much needed cross-subsides from higher income groups. · Community-financing schemes that remain isolated and small deprive their members of the benefits of spreading risks across a broader population. · Community-financing schemes that are disconnected from the broader refer- ral system and health networks deprive their members of the more compre- hensive range of care available through the formal health care system. Links to Mainstream Public Economics Community-financing schemes--in addition to their links to microfinance and social capital--benefit from interconnectivity to the overall welfare of the soci- ety in which they exist, the system of public financing (no matter how weak it may be), and the broader social policy underpinning the prevailing national health system. Schemes that build such connections at an early stage are better able to evolve in terms of expanding the number of members covered, level of resources mobilized, size of the risk pool, and range of benefits they can cover as the local community they serve grows and evolves. Their members have more to gain through such connectivity than they would through isolation. Principal-agent problems also explain why community-based initiatives are expected to be more successful than purely market-based institutions at provid- ing financial protection products. These problems can be overcome in two ways: by designing incentives that align the interest of the agent (insurer) with that of the principal (member), and by designing monitoring systems that allow the principal (member) to effectively observe the actions of the agent (insurer). The proximity of community schemes (agents) to their members (principals) allows effective monitoring, which is much more difficult at the national level. Proponents of linkage between community involvement and public finance argue their case on philosophical and technical grounds. In most societies, care for the sick and disabled is considered an expression of humanitarian and philo- sophical aspirations. Proponents do not, however, have to resort to moral prin- ciples or arguments about the welfare state to justify collective intervention in health. The past century is rich in examples of the failure of the private sector and market forces alone to secure efficiency and equity in the health sector. There is ample justification for such an engagement on both theoretical and practical grounds. In the case of efficiency, there is ample evidence of the significant market fail- ure that exists in the health sector--information asymmetries, public goods, positive and negative externalities, distorting or monopolistic market power of many providers and producers, absence of functioning markets in some areas, and frequent occurrence of high transaction costs (Arrow 1963; Atkinson and Stiglitz 1980; Bator 1958; Evans 1984; Musgrave and Musgrave 1984). In the case 14 Health Financing for Poor People: Resource Mobilization and Risk Sharing of equity, there is equally good evidence that on a voluntary basis individuals and families often fail to protect themselves adequately against the risks of ill- ness and disability (Barer, Getzen, and Stoddart 1998; van Doorslaer, Wagstaff, and Rutten 1993). METHODOLOGY FOR ASSESSING IMPACT, STRENGTHS, AND WEAKNESSES To assess the impact, strengths, and weaknesses of community-based involve- ment in health care financing, we will use a modified version of the World Bank's Poverty Reduction Strategy Paper (PRSP) framework (Claeson and others 2001). According to this framework, community financing can be seen as having three independent objectives: (a) mobilizing financial resources to promote bet- ter health and to diagnose, prevent, and treat known illnesses; (b) protecting individuals and households against direct financial cost of illness when chan- neled through risk-sharing mechanisms; and (c) giving the poor a voice in their own destinies and making them active participants in breaking out of the social exclusion in which they are often trapped. We will not deal with the indirect impact of illness on loss of income due to interruption of employment, although this is clearly another important dimension of financial protection against the cost of illness. This framework is consistent with the three goals of health systems empha- sized by the World Health Report 2000 (WHO 2000): financial fairness (an indica- tor that measures inequality of the financial contribution for health across households), disability-adjusted life expectancy (DALE, an indicator that com- bines life expectancy and disability measures), and responsiveness (a consumer- satisfaction indicator that combines ethical and consumer quality dimensions). This framework is also consistent with the International Development Goals (IDGs) relating to achievement of better health and protection against impover- ishment by the year 2015. The determinants of financial protection, improved health, and social inclu- sion are complex (see figure 1.6). The PRSP framework emphasizes the following causal links: (a) close tracking of key outcome measures relating to improved financial protection, health, and social inclusion; (b) demand and utilization patterns; (c) supply in the health system and related sectors; and (d) policy actions by governments, civil society, the private sector, and donors. Outcome indicators. Much work is still needed to develop a meaningful set of indicators for improving health and protection against impoverishment and combating social exclusion. For this report, we have used both the financial fair- ness, DALE, and responsiveness indicators recommended by the World Health Organization (WHO) and several intermediate indicators (see next section for details). Demand and utilization in influencing financial protection. There is a complex interplay between household assets (human, physical, financial, and social), Rich-Poor Differences in Health Care Financing 15 FIGURE 1.6 Determinants of Financial Protection, Health, and Social Inclusions Demand/ utilization Supply Policies Individuals, Financing and other inputs Governments households, and Services Communities Outcomes communities Other sectors Donors Health services Financing Government Financial Willingness and Public health Revenues Stewardship protection ability to pay Ambulatory care Pooling Coordination In-patient care Purchasing Information Regulation Monitoring and evaluation Enforcement Health Household Input markets Communities outcomes behavior Human Self-regulation HNP risk resources Information factors Knowledge Monitoring and Needs Pharmaceuticals evaluation Expectations Equipment Enforcement Demand Consumables Capital Social Household Other sectors Donors inclusion assets GDP, growth, prices, inflation Loans Human Tax systems, insurance, Grants (education, microfinance Advice health status) Water, food, roads Physical (house and livestock) Financial (loans, savings, insurance) Social (communities, networks) household behavior (risk factors, needs, expectations, and demands for services), ability and willingness to pay, and the availability of insurance or subsidies (Sou- cat and others 1997). This part of the analysis emphasizes the importance of household and community behavior in improving health and reducing the financial risks. 16 Health Financing for Poor People: Resource Mobilization and Risk Sharing Supply in health system and related sectors. There is a hierarchy of interest from nonhealth sector factors in improving financial protection--such as GDP, prices, inflation, availability of insurance markets, effective tax systems, credit, and sav- ings programs--to more traditional parts of the health system (a) preventive and curative health services, (b) health financing, (c) input markets, and (d) access to effective and quality health services (preventive, ambulatory, and in-patient). In respect to the latter, organizational and institutional factors contribute to the incentive environment of health-financing and service delivery systems in addi- tion to the more commonly examined determinants such as management, input, throughput, and output factors (Harding and Preker 2001). Policy actions by governments, civil society, and the private sector. Finally, through their stewardship function, governments have a variety of policy instruments that can be used to strengthen the health system, the financing of services, and the regulatory environment within which the system functions (Saltman and Ferroussier-Davis 2000). This includes regulation, contracting, subsidies, direct public production, and ensuring that information is available. In countries with weak government capacity, civil society and donors can be encouraged to play a similar role. Four levels of analysis were used to assess the impact, strengths, and weak- nesses of community involvement in financial protection against the cost of ill- ness and improved health. They include (a) a survey of the literature on the impact, strengths, and weaknesses of different types of community involvement in health financing; (b) macrolevel cross-country analysis of the impact of dif- ferent health care financing mechanisms on national health systems' perfor- mance indicators--health, financial fairness, and responsiveness; (c) microlevel household data analysis of the specific impact of community-financing schemes on overall welfare of the poor--financial protection and access to health services for the poor; and (d) regional reviews of the Asia and Africa experience of com- munity involvement in health care financing, including different public policy options such as subsidies, reinsurance, linkages to formal public financing sys- tems, and management-capacity building. Methodology for Survey of Literature on Community Financing Despite the recent growth in research on community-based health care financ- ing, there is a paucity of systematic evidence regarding the performance of these schemes in terms of their impact on broad outcome goals such as improving health and protection against impoverishment and combating social exclusion. In particular, little is known about their effectiveness in mobilizing resources and improving access to effective and quality health care; their role in sharing risks across population groups; and their impact on addressing the problems associated with social exclusion. Despite progress made by the time the World Health Report 2000 was published, experts are still debating which indicators best capture progress toward achieving these goals. Rich-Poor Differences in Health Care Financing 17 TABLE 1.2 Summary Statistics of the Literature Reviewed, by Publication Type Internal documents of Peer-reviewed Published international organizations Conference journal articles reports or academic institutions proceedings Number of studies 20 15 4 4 The review looked at any past studies whose main focus had been to examine community involvement in health care financing. Based on this broad criterion, the review comprised 43 studies. The selected papers included articles published in peer-reviewed journals, reports published in formal publication series of inter- national organizations (such as WHO, International Labour Organisation, United Nations Children's Fund), internal unpublished documents of interna- tional organizations and academic institutions, and conference proceedings. Table 1.2 presents the breakdown of the reviewed studies, by publication type. Of these 43 studies, 5 were conceptual papers, 7 were large-scale comparative papers (analyzing five or more community-based health financing schemes) and the remaining 31 were case studies. The regional breakdown of the case studies was 15 in Africa, 11 in Asia, and 4 in Latin America. Language barriers and time constraints created a certain selection bias--Spanish literature was not included in our search while French literature was (Jakab and Krishnan 2001). Assessment of Performance Since past research of community-financing schemes varies considerably in the issues examined and methodologies used, a standard set of questions were asked relating to both the review of impact assessments and the review of determi- nants (key strengths and weaknesses of various types of schemes). The following three questions relating to the impact of community involvement on health, financial protection, and social inclusion were asked: Question 1: What and how robust was the evidence on the amount of resources that could be mobilized through community involvement to pay for health care and the sustainability of this source of financing? Question 2: What and how robust was the evidence on the effectiveness of com- munity involvement in protecting individuals against the impoverishing effects of illness? Question 3: What and how robust was the evidence on the role that community involvement played in combating social exclusion by allowing low-income groups to have a more direct role in the financing of their health care needs and protecting them against the financial burden of illness? A number of studies offered conclusions on resource mobilization, financial pro- tection, and social exclusion based on the experience of authors or review of 18 Health Financing for Poor People: Resource Mobilization and Risk Sharing other studies but did not provide actual evidence in support of their conclusions. Our review excluded studies of performance assessments from the analysis. It also excluded studies that did not use controls from the performance evaluation. This approach yielded 11 studies for the performance assessment of the review. Assessment of Institutional Determinants of Performance The direct and indirect determinants of improved health, financial protection against the cost of illness, and social inclusion are complex. As described earlier by the PRSP framework, policy actions by governments, civil society, and the pri- vate sector are mediated through supply and demand factors related to both the health sector and other sectors that affect the outcome measures being exam- ined. This would include indicators of the service delivery system (product mar- kets), input generation (factor markets), the stewardship or government oversight function (policymaking, coordination, regulation, monitoring, evalua- tion), and market pressures. The current body of literature on community financing is not comprehensive so the report looked only at factors directly related to health care financing. Table 1.3 provides a list of the core technical design, management, organiza- tional, and institutional characteristics related to health care financing in general. Based on this framework, the study reviewed 43 assessments of community- financing schemes for their impacts, strengths, and weaknesses. Methodology for Regional Reviews of Selected Asia and Africa Experiences The main objective of the reviews of selected Asia and Africa experiences was to provide additional insights about several key issues from the perspective of the two regions of the world that carry the heaviest burden of mortality and mor- bidity, have the weakest risk-sharing arrangements to protect their populations against the impoverishing effects of illness, and have the greatest number of poor living in absolute poverty and social exclusion (Arhin-Tenkorang 2001; Hsiao 2001). In addition to contributing to an understanding about the current roles of community involvement in health care financing, the regional reviews also focused on future policy options. Key questions asked include the following: · Using the same framework described under the survey of the literature, what are the main characteristics of existing community involvement in financing health care in the Africa and Asia regions in terms of impacts, strengths, and weaknesses of existing schemes (describe successful and unsuccessful features)? · To what extent do community-financing schemes serve the objective of secur- ing adequate, equitable, and sustainable financing for the low-income and rural populations served (impact on the poor)? · What are the main challenges and obstacles to improving community arrangements to provide adequate, equitable, and sustainable financing? Rich-Poor Differences in Health Care Financing 19 TABLE 1.3 Core Characteristics of the Community-Based Financing Schemes Key policy questions Technical design 1. Revenue-collection mechanisms characteristics Level of prepayment compared with direct out-of-pocket spending Extent to which contributions are compulsory compared with voluntary Degree of progressivity of contributions Subsidies for the poor and buffer against external shocks 2. Arrangements for pooling revenues and sharing risks Size Number Redistribution from rich to poor, healthy to sick, and gainfully employed to inactive 3. Purchasing and resource allocation Demand (for whom to buy) Supply (what to buy and in which form, and what to exclude) Prices and incentive regime (at what price and how to pay) Management 1. Staff characteristics Leadership Capacity (management skills) 2. Culture Management style (top down or consensual) Structure (flat or hierarchical) 3. Access to information (financial, resources, health information, behavior) Organizational 1. Organizational forms (extent of economies of scale and scope, and contractual characteristics relationships) 2. Incentive regime (extent of decision rights, market exposure, financial responsibility, accountability, and coverage of social functions) 3. Linkages (extent of horizontal and vertical integration or fragmentation) Institutional 1. Stewardship (who controls strategic and operational decisions and characteristics regulations) 2. Governance (what are the ownership arrangements) 3. Insurance markets (rules on revenue collection, pooling, and transfer of funds) 4. Factor and product markets (from whom to buy, at what price, and how much) Outcome Indicators Health Financial Social Outcomes Protection Inclusion · Are there other viable alternatives to community financing in the country set- tings where they exist today? · In the context of these study findings, what role could the international donor community play to improve financing for rural and other low-income population groups? 20 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 1.4 Characteristics of 5 Survey Instruments Year of Sample size Organization associated Name of scheme data collection (households) with the survey Rwanda 54 prepayment schemes 2000 2,518 Partnerships for Health Reform in 3 districts of Byumba, (PHR) in collaboration with Kabgayi, and Kabutare National Population Office Senegal 3 Mutual Health 2000 346 Institute of Health and Insurance Schemes Development, Dakar in (Thiés Region) collaboration with ILO India (1) Self-Employed Women's 1998­99 1,200 National Council of Applied Association (SEWA) Economic Research (NCAER) India (2) Self-Employed Women's 1997 1,200 London School of Hygiene and Association (SEWA) Tropical Medicine Thailand Voluntary Health Card 1994­95 1,005 National Statistics Office Scheme (HCP) Methodology for Microlevel Household Survey Analysis The aim of the microlevel household survey analysis was to shed light on two questions (Jakab and others 2001): What characteristics affect the decision of households to join community-based prepayment schemes? Do community health-financing schemes provide financial protection for their members against the cost of illness? Eleven household budget surveys, four Living Standard Measurement Surveys (LSMS), and nine Demographic and Health Surveys (DHS) were screened for community-financing data. Most of these surveys did not allow an identification of households with access to community-based health financing. Of the 11 smaller scale nonstandardized surveys that matched the requirements for the core list of variables, 5 were available for further analysis and were included in this report. Table 1.4 summarizes the key characteristics of these surveys. The remaining 6 were either not accessible for further analysis (4), data collection was incomplete (1), or authors were not available to collaborate (1). The five household surveys identified and accessible for analysis for the purposes of this report represent nonstandardized, relatively small-scale data- collection efforts with a sample size of 346 to 1,200 households. The surveys were not nationally representative; they were a random sample of the local pop- ulation. With the exception of Thailand's, all surveys are very recent. Determinants of Inclusion To assess the determinants of social inclusion in community-financing schemes, we assume that the choice of whether to enroll is influenced by two main deter- minants: individual and household characteristics, and community characteris- Rich-Poor Differences in Health Care Financing 21 tics. Individual and household characteristics influence the cost and the benefit calculation of the rational individual decision maker. This choice is moderated, however, through certain social characteristics of the community. The individual rational choice model of weighting costs and benefits of joining a prepayment scheme is altered by the social values and ethics of the local culture. For example, two individuals with similar individual and household characteristics (such as income, household size, assets, education level, health status) may decide differently about joining a prepayment scheme depending, for example, on encouragement from community leaders, availabil- ity of information, and ease of maneuvering unknown processes. To estimate the weight of these determinants, a binary logit model was applied to four of the data sets, and a binary probit was applied to the Senegal data set. The model can be formally written as follows: (1.1) Prob (membership > 0) = X11 + X22 + The independent variable takes on a value of 1 if the individual belongs to a com- munity-financing scheme and 0 if he or she does not. X1 represents a set of inde- pendent variables for characteristics of the individual and the household, such as income, gender, age, and marker on chronic illness or disability. X2 represents a set of independent variables that approximate the social values in the communi- ties: religion and marker on various communities where appropriate. Other vari- ables specific to the surveys as well as interaction terms were included where appropriate. 1 and 2 are vectors of coefficient estimates and is the error term. The two variables of primary interest are income (measure of social inclusion) and a marker for community factors (dummy variable). Control variables also included gender, age, disability or chronic illness, religion, and distance to the health center under the scheme. Some of these variables are important to control for the different probability of health care use (for example, age, health status, and distance from provider). These variables also allow us to test the presence and importance of adverse selection to which all voluntary prepayment schemes are subject. Other variables included control for the different individual and house- hold attitudes toward investment in health at a time when illness is not necessar- ily present (for example, gender and religion). Literature has shown that the distance to the hospitals and local health centers and existence of outreach pro- grams influence the decision to purchase membership to the scheme. Determinants of Financial Protection To empirically assess the impact of scheme membership on financial protection, a two-part model was used.1 The first part of the model analyzes the determinants of using health care services. The second part of the model analyzes the determi- nants of health care expenditures for those who reported any health care use. There are several reasons for taking this approach. First, using health expendi- ture alone as a predictor of financial protection does not allow capture of the lack of financial protection for people who choose not to seek health care 22 Health Financing for Poor People: Resource Mobilization and Risk Sharing because they cannot afford it. As the first part of the model assesses the determi- nants of utilization, this approach allows us to see whether membership in com- munity financing reduces barriers to accessing health care services. Second, the distribution of health expenditures is typically not a normal distribution. Many nonspenders do not use health care in the recall period. The distribution also has a long tail due to the small number of very high spenders. To address the first cause of nonnormality, the study restricted the analysis of health expenditures to those who report any health care use. As the first part of the model assesses determinants of use, we will still be able to look into whether scheme member- ship removes barriers to care. To address the second part of nonnormality, a log- linear model specification is used. Part one of the model is a binary logit model for the India, Rwanda, and Thai- land data sets and a probit model in the Senegal model. The model estimates the probability of an individual visiting a health care provider. Formally, part one of the model can be written as follows: (1.2) Prob (visit > 0) = X + . Part two is a log-linear model that estimates the incurred level of out-of-pocket expenditures, conditioned on positive use of health care services. Formally, part two of the model can be written as follows: Log (out-of-pocket expenditurevisit > 0) = X + µ where X represents a set of individual and household characteristics hypothe- sized to affect individual patterns of utilization and expenditures. and are vectors of coefficient estimates of the respective models; and µ are error terms. The two variables of primary interest are scheme membership status and income. Other control variables were also included in the estimation model to control for the differences in need for health care (for example, age and gender), differences in preferences toward seeking health care (for example, gender and religion), and differences in the cost (direct and indirect) of seeking health care (for example, distance). Methodology for Macrolevel Cross-Country Analysis For the dependent variables of the macrolevel country analysis, the study used the standard indicators proposed by WHO for health systems performance (WHO 2000). These are the disability-adjusted life expectancy (DALE), the index of level of responsiveness (IR), the index of fairness of financial contribution (IFFC), the index of distribution of responsiveness (IRD), and the index of equal- ity of child survival (IECS). Only the observed data for these indicators were included in the analysis. For the independent variables of the macrolevel analysis, countries were divided into three groups based on the extent of their risk-sharing arrangements. Rich-Poor Differences in Health Care Financing 23 We assign countries to the first category, advanced risk sharing, when they have either a social health insurance scheme or a health-financing scheme based on general taxation, and when these two schemes are associated with the principle of universal coverage. Countries with no explicit reference to overall coverage of the population, who usually have mixed health-financing systems, with some part of the population partially covered via general taxation and specific popula- tion groups covered by health insurance schemes, are associated with the second category, medium risk sharing. Finally, countries with general taxation systems that incompletely cover the population are associated with the third category, low risk sharing. This classification system allows us to define the two main orga- nizational dummy variables: DARS = 1 when a country belongs to the set of advanced risk-sharing systems and 0 otherwise; DMRS = 1 when a country belongs to the set of medium risk-sharing systems and 0 otherwise. The methodology for this analysis is described by Carrin and others (2001). The objective of the analysis is to examine the degree to which risk sharing has a beneficial impact on the five indicators of health systems performance. The analysis used the following specification for the impact of risk sharing on the level of health: (1.3) Ln (80 ­ DALE) = a1 + b1 Ln HEC + c1 Ln EDU + d1 DARS. HEC refers to the health expenditure per capita (in U.S. dollars). EDU refers to the educational attainment in society and is measured by the primary enroll- ment. The dependent variable is the logarithm of the difference between the observed DALE and a maximum. Several alternative models were also tested. The hypothesis is that advanced risk sharing (among indirect determinants such as education) is associated with a better definition of the benefit package of health services to which citizens are entitled, which translates into an increased overall level of health. The analysis used two alternative functional forms to assess the impact of risk sharing on responsiveness: (1.4a) Ln (IR/[1 ­ IR]) = a21 + b21 HEC + c21 EDU + d21 DARS and (1.4b) Ln (1 ­ IR) = a22 + b22 LnHEC + c22 Ln EDU + d22 DARS. The hypothesis to be tested is that advanced risk-sharing systems are associated with a larger degree of stewardship. The latter, in turn, is likely to positively influence the mechanisms and incentives that entail a greater responsiveness. The analysis used three measures for distributional impact. This included an IECS, an IFFC, and an IRD. Several models were tested. A model was developed that examined the impact of the dummy variable (DARS) on the distributional variables for health, fair- ness, and responsiveness. We have adopted the same functional forms as in equations (1.4a and 1.4b): 24 Health Financing for Poor People: Resource Mobilization and Risk Sharing (1.5a) Ln (Ij/[1 ­ Ij]) = a31 + b31 DARS and (1.5b) Ln (1 ­ Ij) = a32 + b32 DARS, where Ij (j = 1,...,3) refers to the three above-mentioned indexes, respectively. The effect of DARS on the indicator of fair financing is expected to be positive when using the logit form of the equation. The hypothesis to be tested is that in countries with advanced risk sharing, more so than in other systems, people make financial contributions according to their capacity to pay. This would be associated with a higher IFFC. In addition, systems with universal coverage generally pay more attention to the objective of equal treatment for equal need. It is therefore assumed that such systems also respond to people's expectations as to the nonmedical aspects of health care in a more equal way. Hence the effect of DARS on the distrib- ution of responsiveness is anticipated to be positive as well. Finally, it is assumed that universal coverage systems are more likely to provide people with a similar benefit package than other systems, irrespective of their socioeconomic background, with a resulting positive impact on the distributional aspects of child health. DISCUSSION OF MAIN FINDINGS FROM BACKGROUND REVIEWS Based on a review of the 43 papers discussing community-based health financing, the first and foremost conclusion is that there is a paucity of systematic empirical work regarding the performance of these financing mechanisms or the determi- nants of good outcomes in achieving good health (Jakab and others 2001). Discussion of Survey of Existing Literature on Community Health Financing Although several authors have tired to create a typology for community-based schemes (Atim 1998; Bennett, Creese, and Monasch 1998; Criel, van der Stuyft, and van Lerberghe 1999; Hsiao 2001), the possibilities for variations is almost limitless, given the great diversity in objectives, design, context, and implementation arrangements. Nevertheless, the review revealed four commonly encountered and well-identifiable types of schemes. In the first type, resource mobilization relies mainly on out-of-pocket payments at the point of contact with providers, but the community is actively involved in designing these fees and managing the collec- tion, pooling, and allocation of the funds mobilized in this way (community cost- sharing). In the second type, the community collects payments in advance of treatment (prepayment) and then manages these resources in paying for providers (community prepayment or mutual health organization). In the third type, providers serving a particular community collect the prepayments themselves (community provider-based health insurance). In the fourth type, the community acts as "agent" to reach rural and excluded populations on behalf of the formal government or social health insurance system (government or social insurance) via contracts or agreements. Rich-Poor Differences in Health Care Financing 25 TABLE 1.5 Types of Community-Based Financing Four community-based finance modalities Type of Government Type 4 Type 3 Type 2 Type 1 Direct user scheme schemes: Linked Community Community- Community- fees (spot social- community provider- based managed market) insurance health fund, based health prepayment user fees and tax- revolving insurance schemes based fund, or prepayment Table 1.5 summarizes these four types of community-based financing schemes based on their core design features and management, organizational, and institutional characteristics. Assessment of Impact Following the framework presented in table 1.3, the survey of the literature looked at three indicators of performance of community-based financing schemes (Jakab and others 2001): (a) their effectiveness in mobilizing resources and improving access to effective and quality health care; (b) their role in shar- ing risks across population groups; and (c) their impact on addressing the prob- lems associated with social exclusion (see table 1.6). This is followed by a discussion on the key conclusions from the performance review of the literature. Resource Mobilization. There is good evidence from the literature that community- financing arrangements make a positive contribution to the financing of health care at low-income levels, thereby improving access to drugs, primary care, and even more advanced hospital care (Dave 1991). Such community involvement allowed rural and low-income populations to mobilize more resources to pay for health care than would have been available without this involvement (Diop, Yazbeck, and Bitran 1995; McPake, Hanson, and Mills 1993; Soucat and others 1997). But there are great variations in the volume of resources that can be mobi- lized this way, constrained largely by the low income of the contributing popu- lation (Atim 1998; Bennett, Creese, and Monasch 1998; Hsiao 2001; Jütting 2000--see box 1.1). This is particularly true when most members of the community TABLE 1.6 Number of Studies that Examined Core Health-Financing Subfunctions Financing function Revenue collection Pooling of revenues Resource allocation or purchasing Type 1 5 2 3 Type 2 6 4 9 Type 3 2 2 3 Type 4 3 3 2 Multiple 10 2 3 26 Health Financing for Poor People: Resource Mobilization and Risk Sharing BOX 1.1 REVENUE MOBILIZATION Based on data from Bennett, Creese, and Monasch (1998), this graph shows the cost recovery from prepayment of six Modality II schemes. The range is from 12 percent to 51 percent of recurrent expenditure. This shows that, for these schemes, the resources collected contribute significantly to the full recurrent costs but do not fully cover them, thereby necessitating other sources of fund- ing, such as out-of-pocket spending, government subsidies, and donor grants. Percent of recurrent costs from prepayment 60 50 40 30 20 10 0 GK SSSS Caja Abota Grameen Lalitpur (Bangladesh) (India) (Bolivia) (Guinea- (Bangladesh) (Nepal) Bissau) schemes are already below the poverty line. None of the studies reviewed reported the share of aggregate national resources that were mobilized through community- financing arrangements. There is an urgent need to strengthen the evidence base of community-financing arrangements through more rigorous registration, monitor- ing, and evaluation of the resource mobilization capacity of these schemes. Financial Protection. Where household survey data have been analyzed, a consis- tent observation was that community-based health financing has been effective in reaching more low-income populations who would otherwise have no finan- cial protection against the cost of illness (Litvack and Bodart 1992). Improved financial protection is achieved through reducing the members' out-of-pocket spending while increasing their utilization of health care services (Atim 1998; Criel, van der Stuyft, and van Lerberghe 1999; Desmet, Chowdhury, and Islam 1999, Gumber and Kulkarni 2000; Jütting 2000; Supakakunti 1997). At the same time, some of the research suggested that the poorest and socially excluded groups are often not included in community-based health-financing initiatives (Arhin-Tenkorang 1994; Criel, van der Stuyft, and van Lerberghe 1999; Jütting 2000). Those studies that compared the level of financial protection of scheme members with that of nonmembers found that belonging to some form of pre- payment scheme reduced the financial burden of seeking health care (Arhin- Tenkorang 1994; Diop, Yazbeck, and Bitran 1995; DeRoeck and others 1996; Rich-Poor Differences in Health Care Financing 27 TABLE 1.7 Studies that Looked at Ways to Prevent Impoverishment Studies that confirmed key Utilization of members Level of out-of-pocket expenditures hypothesis being tested relative to nonmembers of members relative to nonmembers Increase Decrease Increase Decrease Type 1 Community user fees 3 1 0 1 Type 2 Community prepayment 4 2 0 6 Type 3 Provider prepayment 3 0 0 0 Type 4 Linked to formal system 3 0 0 2 TABLE 1.8 Studies that Looked at Ways to Combat Social Exclusion Distance Scheme Poorest Inability to pay, Rich gradient to Studies that confirmed key reaches not main reason for do not scheme hypothesis being tested the poor covered not being covered participate provider Type 1 Community user fees 3 1 1 0 0 Type 2 Community prepayment 5 1 2 1 1 Type 3 Provider prepayment 2 2 1 1 1 Type 4 Linked to formal system 3 1 2 1 1 Gumber and Kulkarni 2000; see table 1.7). Two studies indicated that community financing does not eliminate the need for broader coverage in the case of cata- strophic health care expenditures (Pradhan and Prescott 2000; Xing-Yuan 2000). Combating Social Exclusion. Community-based health-financing schemes appear to extend coverage to many rural and low-income populations who would otherwise be excluded from collective arrangements to pay for health care and protect them against the cost of illness. However, the poorest are often excluded even from community-financing arrangements, and higher income groups often do not belong, thereby segmenting the revenue pool by income group (see table 1.8). Identification of Determinants The survey of the literature also looked at factors that would contribute to strengths and weaknesses of the schemes (Jakab and others 2001) in the following four areas: (a) technical design characteristics, (b) management characteristics, (c) organizational characteristics, and (d) institutional characteristics. The key advan- tages and disadvantages of community-based schemes lie in their ability to fill the policy, management, organizational, and institutional void left by extreme gov- ernment failure to secure more organized financing arrangements for the poor. In this context, a number of strengths (see box 1.2) and weaknesses (see box 1.3) of community-financing schemes have been identified by various authors. 28 Health Financing for Poor People: Resource Mobilization and Risk Sharing BOX 1.2 STRENGTHS OF COMMUNITY-FINANCING SCHEMES Technical Design Characteristics Revenue Collection Mechanisms · Shift away from point-of-service payment to increasing prepayment and risk sharing · Flat-rate premium, which facilitates revenue collection, reduces the scope for manipulation, and contributes to low transaction costs · Contribution payment that accommodates the income-generating pat- terns of households employed in agriculture and the informal sector (irregular, often noncash) · Modest degree of household-level affiliation · Pro-poor orientation even at low-income levels through exemptions of premiums and subsidies, despite flat-rate contribution rate · Some buffering against external shocks though accumulation of reserves and links to formal financing schemes Arrangements for Pooling Revenues and Sharing Risks · Some transfers from rich to poor, healthy to sick, and gainfully employed to inactive through some pooling of revenues and sharing of risk within community groups Purchasing and Resource Allocation · Most community schemes make a collective decision about who is cov- ered through scheme, based on affiliation and direct family kinship (for whom to buy). · Many community schemes define the benefit package to be covered in advance (what to buy, in what form, and what to exclude). · Some community schemes engage in collective negotiations about price and payment mechanisms. Management · Most community schemes are established and managed by community leaders. Community involvement in management allows social controls over the behavior of members and providers that mitigates moral hazard, adverse selection, and induced demand. · Many schemes seek external assistance in strengthening management capacity. · The management culture tends to be consensual (high degree of democra- tic participation). · Most schemes have good access to local utilization and behavior patterns. Rich-Poor Differences in Health Care Financing 29 BOX 1.2 continued Organizational Structure · Most community schemes are distributed organizational configurations that reach deep into the rural and informal sectors. · Incentive regimes include: (a) extensive decision rights, (b) strong internal accountability arrangements to membership or parent community organi- zation, (c) ability to accumulate limited reserves if successful but unsuc- cessful schemes often ask governments for bailouts, (d) mainly factor-market exposure since few overlapping schemes compete with each other in the product market, and (e) some limited coverage of indigent populations through community or government subsidies. · Vertical integration may lead to increased efficiency and quality services. Schemes that have a durable partnership arrangement or contractual arrangement with providers able to negotiate preferential rates for their members. This in turn increases the attractiveness of the scheme to the population and contributes to sustainable membership levels. · Better organized schemes use horizontal referral networks and vertical links to formal sector. Institutional Environment · Stewardship function is almost always controlled by local community, not central government or national health insurance system, which is apt to make the schemes responsive to local contexts. · Ownership and governance arrangements (management boards or com- mittees) are almost always directly linked to parent community schemes; freestanding health insurance schemes are rare. · There is little competition in the product market. · Competition is limited in factor markets and through consumer choice. Discussion of Main Findings from Asia Regional Review The review of selected Asia experiences emphasized the heterogeneity of community-financing schemes and the fact that their performance is highly depen- dent on the nature of their technical design and management, organizational, and institutional characteristics. For the purpose of this review, Hsiao (2001) classified community involvement in health care financing into five types: (a) direct subsidy to individuals (Thai Health Card and Tanzania Community Health Fund), (b) coop- erative health care (Jiangsu Province and Tibet), (c) community-based third-party insurance (Rand Experiment in Sichuan Province and Dana Sehat), (d) provider sponsored insurance (Dkaha Community Hospital, Gonoshasthya and Bwamanda), and (e) producer-consumer cooperative (Grameen). 30 Health Financing for Poor People: Resource Mobilization and Risk Sharing BOX 1.3 WEAKNESSES OF COMMUNITY-FINANCING SCHEMES Technical Design Characteristics Revenue Collection Mechanisms · Without subsidies, resource mobilization is limited when everyone in the pool is poor. · Many of the poorest do not join since they cannot afford premiums. · Pro-poor orientation is undermined by regressive flat-rate contributions and by a lack of subsidies or premium exemption, which create a finan- cial barrier for the poor. · Community-based voluntary prepayment schemes are also prone to adverse selection. · Few schemes have reinsurance or other mechanisms to buffer against large external shocks. Revenue-Pooling and Risk-Sharing Arrangements · The scope for transfers within very small pools is limited (often fewer than 1,000 members per scheme). Purchasing and Resource Allocation · Without subsidies, the poorest are often left out (for whom to buy). · The benefit package is often very restricted (what to buy, in what form, and what to exclude). · Providers can often exert monopoly power during price and payment negotiations. Management · Community leaders are as vulnerable to adverse incentives and corruption as national bureaucrats. · Even with external assistance, absorptive capacity in management training is limited. · Extensive community consultation is time consuming and can lead to conflicting advice. · Most schemes do not use modern information management systems. Organizational Structure · Even widely distributed organizational configurations may have difficulty reaching deep into the rural and informal sectors. · There are often conflicting incentives, especially among extensive decision rights, soft budget constraints at time of deficits (bailouts by governments and external sources of funding such as nongovernmental organizations), limited competitive pressures in the product markets, and lack of financ- ing to cover the poorest population groups. · The less-organized schemes are often cut off from formal sector networks. Rich-Poor Differences in Health Care Financing 31 BOX 1.3 continued Institutional Environment · Government stewardship and oversight function are often very weak, lead- ing to a poor regulatory environment and lack of remedies in the case of fraud and abuse. · Ownership and governance arrangements are often driven by nonhealth and financial protection objectives. · Choice in strategic purchasing is limited by small number of providers in rural areas. · True consumer choice is often limited by lack of a full insurance and product market, leading to (a) adverse selection (signing on only the better-off, working age, and healthy), (b) moral hazard (members mak- ing unnecessary claims because they have insurance coverage), (c) free- rider effect (households waiting until they think they will be sick before joining), and (d) information asymmetry (for example, concealing pre- existing conditions). Sources: Bennett, Creese, and Monasch (1998); Carrin, Desmet, and Basaza (2001). Based on this typology, the Asia review ranked the community-financing schemes examined according to their potential impacts on several intermediate outcome indicators (coverage, equity in financing, efficiency and cost contain- ment in service delivery, access, quality, and degree of risk sharing). The results are summarized in table 1.9. Based on this framework, the review made the following observations: · Rural households and urban poor households are willing to prepay a portion of their health services. The resources that can be raised in this manner depend on both economic and social factors. · Since the membership of many community-financing schemes consists of poor households, their ability to raise significant resources to pay for health care is limited by the community's overall income, exposure to out-of-pocket payment when not enrolled, availability and size of subsidies, and satisfaction with the services provided. The poor and near poor are more motivated to prepay if their contributions are supplemented by government or donor sub- sidies. For the poorest households, this subsidy has to be a large share of the total payment. · The social factors that influence membership rates include a sense of kinship, mutual community concern, and trust and confidence in the management of the scheme. unless subsidyt pooling Greater gov' risk Low Low Modest w/ High High Medium Improve quality Low Low High Low Low High for insured for insured Improve access Low Modest High High those High those High and cost None None High Low Low High Increase efficiency reduce Schemes in Low Low Low Low Low Low Equity financing to and funds families families higher higher member Community-Financing covered of Population raise be Low Modest High Cover income Cover income Cover ypesT and NGO by fund controls of purpose Added community Who use alueV Government Individual households Local special Community Hospitals Cooperatives Potential fees health insurance consumer (2001). 1.9 user -sponsored or Hsiao of ABLET ypeT care third-party insurance cooperative community-financing scheme Prepay Cooperative Community-based Provider Provider Source: 32 Rich-Poor Differences in Health Care Financing 33 FIGURE 1.7 Hospitalization and Impoverishment Bihar Uttar Pradesh Punjab Rajasthan Gujarat Madhya Pradesh West Bengal North East All India Maharashtra Orissa Haryana Andhra Pradesh Karnataka Tamil Nadu Kerala 0 5 10 15 20 25 30 35 40 Percent falling into poverty Source: Peters and others (2001). · A major additional value of well-performing community-financing schemes is expanded access to quality services, improved efficiency of management and service delivery, and cost containment. · Governmental and nongovernmental organizations (NGOs) often catalyzed the startup of the community-financing schemes in question and contributed to its management and sustainability. · Finally, members appear to prefer coverage for both primary care and more expensive hospital care. Since many schemes do not mobilize sufficient resources to pay for both, a number of communities opt for primary care cov- erage, which they will use regularly for their basic health care needs, rather than insurance coverage for rarer and more expensive events that may only happen once or twice in a lifetime and whose concept is often poorly under- stood. This creates a tension or trade-off between individual needs and demands for basic care and household and community needs for financial protection (see figure 1.7). 34 Health Financing for Poor People: Resource Mobilization and Risk Sharing Discussion of Main Findings from Africa Regional Review The review of selected Africa experiences (Arhin-Tenkorang 2001) emphasized that a common feature of many of the reforms introduced during the past two decades have consisted of copayments to influence utilization patterns and direct out-of-pocket user charges to mobilize much-needed additional resources (Vogel 1990). Most of the population currently does not benefit from formal insurance coverage, and government expenditures often do not meet the basic health needs of the poor, let alone the entire population (Abel-Smith and Rawal 1994). These user charges add significantly to the financial hardship of poor households, often fully exposed to the financial risks associated with illness. This has been especially true during recent years, due to the rising incidence and prevalence of HIV/AIDS, tuberculosis, and other communicable diseases. A central premise of the Africa review is that individuals in the informal sector of poor countries cannot access appropriate health care--particularly curative care-- at the time of need partially because of lack of adequate insurance coverage (Arhin- Tenkorang 2001). Although preventive measures may have long-term payoffs in improving the overall welfare and productivity of the population, the income shock associated with seeking access to curative and palliative care has become such a great financial burden for the poor that some form of insurance coverage has to be considered an essential part of any serious poverty alleviation strategy. The first section of the chapter conceptualizes how the interaction between sev- eral design features and institutional factors influences scheme performance in terms of risk protection and resource mobilization. In the absence of risk protection, several African studies demonstrated that poor households often deferred visits to formal health facilities until their illnesses became quite severe or used ineffective self- medication sometimes injurious to their health, leading to more severe health and financial consequences than would have been the case had care been sought earlier. Key design features included the methodology, nature, and quality of the data used to determine contribution levels, benefit packages, and subsidy levels. The argument is that appropriate specifications require (a) data on the target popula- tion's willingness-to-pay (WTP) and ability-to-pay, often not collected or available, (b) data on projected costs of the benefits to be consumed, and (c) operational modalities that facilitate interaction between individuals in an informal environ- ment and in a range of formal organizations. The review concludes that in an informal environment, decisions cannot rely on such written information because the needed data are usually not available in this form. To be effective and afford- able, activities undertaken by community-financing schemes must be based on simple and directly observable behavior patterns with low transaction costs. Key institutional features included the degree of congruence between the scheme's operating rules and the participating population's normal behavior patterns. They also included the degree of participating health care providers' past experience with third-party payments and contractual arrangements. The review found that these institutional factors had a significant influence on the nature and extent of community participation in any given scheme, as well as Rich-Poor Differences in Health Care Financing 35 the quality of its management and monitoring of performance. The review did not examine other institutional factors, such as government regulations and laws governing insurance and health care provision. The second part of the chapter proposes the design features of several potential "high population schemes" for Africa's informal sector and assesses their perfor- mances with respect to risk protection and resource mobilization. Potential "high population" schemes examined include the Abota Village Insurance Scheme (Guinea-Bissau); Bwamanda Hospital Insurance Scheme (Democratic Republic of Congo); Carte d'Assurance Maladie, or CAM, program (Burundi); Dangme West Health Insurance Scheme (Ghana); Nkoranza Community Financing Health Insur- ance Scheme (Ghana); and Community Health Fund, or CHF, (Tanzania). These schemes had large target populations and provided a comprehensive range of ben- efits and geographically accessible care to its members. Key factors influencing enrollment appeared to include a matching of the premium to the willingness and ability to pay, availability of government subsidies for the poor who cannot afford the basic premium; and ready access to basic care for common health problems and emergency care--both geographic proximity and availability of range of basic services appeared to significantly affect enrollment. The final part of the chapter presents a set of policy measures that national and international health policymakers may consider implementing to increase the level of risk protection provided for informal sector populations. The financial risk protection and resource mobilization that can be achieved by any given scheme appears to be influenced by the compatibility between the way it is designed and operated and the behavior of the individuals and households from the informal sector that enroll in the schemes. The enrollment rate of a given population with such schemes appears to reflect the target population's WTP, in turn, closely related to its ATP. In most cases, some central government support in the form of fiscal transfers, budget allocations, or both is necessary, given the small volume of resources available at low-income levels in poor communities. Schemes that are operated as solidarity-based partnerships with service providers appear to create additional incentives to increase efficiency and accountability. The authors con- clude that national government policies, a legal framework, and financial support for these organizations are likely to be a good investment of scarce government resources. The authors emphasize that, in the absence of established practices in the design of community-financing schemes, donor funding, procedures, and reg- ulations supporting community financing through communities, local govern- ments and local NGOs need further pilot testing to identify the elements that would be needed to expand the schemes or to go to scale with them. Discussion of Main Findings from Microlevel Household Survey Analysis Determinants of Social Inclusion in Community Financing In terms of the determinants of social inclusion through community financing, the results from the microlevel household survey analysis are varied. Table 1.10 36 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 1.10 Statistically Significant Determinants of Inclusion in Community Financing Rwanda Senegal India (1) India (2) Thailand Model Logit Probit Logit Logit Logit Dependent variable Dependent Proportion of Proportion of Proportion of Proportion of Proportion of variable population population population population population that enrolled in 1 enrolled in 1 enrolled in enrolled in purchased new of 66 schemes of 4 schemes SEWA- SEWA- health card, insurance insurance continued, dropped out, never purchased Independent variables: individual & household characteristics Income/assets No Yes No No Yes Age No No Yes Yes No Education Yes No No No Yes Gender No No -- -- No Health status No -- Yes Yes Yes Household size Yes No Yes No -- Marital status Yes No No No Religion -- Yes -- No -- Distance of Yes -- -- -- -- household from scheme provider Independent variables: community characteristics Community Yes Yes -- -- -- marker for unobservable characteristics Solidarity N/A Yes N/A N/A N/A Note: Yes--variable is significant at least at the 10 percent level. No--variable is not significant. (--)--not included in the particular model. presents the determinants that were found statistically significant in the five household surveys (Gumber 2001; Jütting 2001; Ranson 2001; Schneider and Diop 2001; Supakankunti 2001). · Income and other socioeconomic determinants. In Senegal and Thailand, house- hold income was a significant determinant of being member of a prepayment scheme; in Rwanda and India income was not significant. · Other individual and household characteristics. Health status was included in the analysis of the Rwanda, Thailand, and both India surveys. In all four surveys, the analysis confirmed the presence of the adverse selection that characterizes Rich-Poor Differences in Health Care Financing 37 voluntary prepayment schemes. Patients with recent illness episodes or with chronic illnesses are more likely to purchase a prepayment plan. Distance of the household from the provider of the scheme was included in the Rwanda analysis. Households less than 30 minutes from the health facility of the scheme were four times more likely to belong to the prepayment scheme than households living farther away. · Community characteristics. Dummy variables for community characteristics were significant predictors of the probability of enrolling in the prepayment scheme (Senegal and Rwanda). Determinants of Financial Protection in Community Financing The results are varied in terms of the determinants of financial protection through community financing. Table 1.11 presents the determinants found statistically sig- nificant in four of the household surveys. The household survey conducted in Thailand did not permit analysis of the determinants of out-of-pocket payments and was therefore excluded. The key findings from this part of the study include: · Insurance effect. In three of the five household surveys, membership in a com- munity-financing scheme was a significant determinant of the probability of using health care and in reducing out-of-pocket payments. This confirms our original hypothesis that even small-scale prepayment and risk pooling reduce financial barriers to health care (Rwanda, Senegal, and India). · Socioeconomic determinants. The analysis indicated that even with insurance, low income remains a significant constraint to health care utilization and ability to pay out-of-pocket payments (Rwanda, Senegal, and India). · Other determinants. Distance from scheme provider was a significant determi- nant of the likelihood of using health care (Rwanda and Senegal). Discussion of Main Findings from Macrolevel Cross-Country Analysis A first observation was that most routine national statistical sources do not include data on the share of overall financing channeled through either community-based or private health insurance schemes (Carrin and others 2001). The analysis there- fore had to focus on the extent of collective risk sharing provided at low-income levels through different combinations of general tax revenues and social insurance. The equations have been estimated with the ordinary least squares method, using data for the explanatory variables HEC, EDU, and PHE percents that per- tain to the year 1997. The Gini index pertains to specific years, depending upon the country, in 1986­99. In this synthesis chapter, we present only the "best" regressions2 in summary tables 1.12 and 1.13 in the appendix. Except for the functional form of the regression for DALE, we present only the results of the logit specification. The estimation results for the basic model presented in sum- mary table 1.12 are discussed next. use are conditional care 0) OOPs for > group they annual for for as (2) (use Log-linear on altoT hospital s s out-of-pocket payment of Ye Only richest quintile Only oldest Ye -- seY s Ye oN seY -- but, India of at studies w/ group the one for for of Utilization Logit Proportion sample least hospitalization No Only richest quintile Only oldest oN -- esY No oN s Ye -- some in included Patterns Ps conditional 0) indirect OO > were annual and health of use small (use nditure (1) Log-linear on otalT direct cost care No seY -- -- -- seY size Only hh -- -- -- variables India Expe of care control reporting for health severe Other Utilization Logit Proportion sample any use seY No No No No Only very No oN -- -- Out-of-Pocket model. and characteristics effect spending particular conditional the OOPs 0) > household in Utilization Insurance of (use variable hospitalization and Log-linear on Out-of-pocket for esY seY s Ye No No seY -- -- -- No included Model Senegal variables: Dependent Individual of at w/ (--)--not Determinants one for Utilization Logit Proportion sample least hospitalization Independent seY variables: Only richest terzile No No seY s -- -- -- -- Ye significant. Percent) 10 not to is at Independent payment illness Least of conditional episode 0) (at OOPs poorest > full illness-related for No--variable (use episode the s Log-linear on otalT out-of-pocket per for esY Only quartile No No Ye No No No level. table. Significant Rwanda to this percent in visit sample health 10 of one them the at Utilization least Statistically provider at least Logit Proportion w/ professional care seY seY seY No No seY No ---- ---- seY include at not did Findings: we, significant is paper Summary the provider in variable size household 1.11 membership of status/severity status scheme es--variableY illness discussed ABLET of from Dependent Scheme Income/assets Age Education Gender Health Household Marital Religion Distance Note: not 38 Rich-Poor Differences in Health Care Financing 39 First, concerning the level of health (DALE), the effects of DARS, HEC, and EDU are as expected and are statistically significant at the 1 percent significance level. Second, from the equation for the level of responsiveness (IR), we see that HEC and EDU do not have a statistically significant impact. One major reason is likely to be that the index of responsiveness contains elements of both respect for per- sons and client orientation and that both are influenced differently by HEC and EDU. For instance, HEC may be important in explaining client orientation, but it may not be when explaining respect for persons. Therefore, when analyzing the determinants of the overall index of responsiveness, the effect of HEC may disappear. Notice, however, that both the coefficients of DARS and DMRS have the expected sign and are statistically significant. Third, the explanatory power of the regression for the index of fair financing (IFFC) is minimal; DARS does not have a statistically significant impact on the IFFC. We submit that the major reason for this unsatisfactory result is the rela- tively small sample size. Moreover, the sample did not include sufficient data on countries with advanced and low risk sharing. For instance, the (full sample) data on advanced risk sharing are those of Bulgaria, Jamaica, Kyrgyz Republic, Romania, and Russia and do inadequately reflect the experience of high-income countries with either social health insurance or general taxation financing. Fourth, in the equation for the distribution of responsiveness (IRD), the coeffi- cient of DARS is statistically significant. The impact of DSHI is statistically insignificant. Fifth, the results for the index of equality of child survival (IECS) show that both DARS and DMRS have statistically significant impacts. We next present the estimation results for the enlarged model with the Gini index as an explanatory variable in the equations for the distributional mea- sures. The results are presented in table 1.13 (appendix 1A). In the fair financing equation (IFFC), which has very low explanatory power, the coefficient of the Gini index has the anticipated sign but is not statistically significant. The coeffi- cient of DARS is also not statistically significant. Related to the distribution of responsiveness (IRD), the result shows significant impacts of both DARS and DMRS, as well as of the Gini index. All coefficients have the expected sign. One can conclude that these risk-sharing arrangements are efficient in counterbalancing the overall effect of income inequality. A threshold for the Gini indexes can be computed, indicating the value above which risk sharing is no longer able to counteract the effect of overall income inequality. In the case of a country with an advanced risk-sharing scheme, the threshold value is 57.9. In the case of medium risk-sharing schemes, the thresh- old is 26.3. From these estimates we can infer that advanced risk-sharing schemes are more effective in counteracting the effects of overall income inequality in society. For example, let us assume that a country has a Gini coeffi- cient of 35. If this country has an advanced risk-sharing scheme, its effect will outweigh the impact of income equality: the combined effect will be +0.8588. However, if the country has a medium risk-sharing arrangement, the combined effect will be ­0.3252. 40 Health Financing for Poor People: Resource Mobilization and Risk Sharing In the regression result related to the inequality of child survival (IECS), the sign of the Gini coefficients is against our expectations. Surprisingly, the Gini coefficient is also statistically significant at the 10 percent level. The coefficient of DARS has the anticipated sign, however, and is statistically significant at the 1 percent level. Inclusion of the interaction variables with PHE percent in the equations did not result in a general improvement of the estimation results. For instance, in a number of cases, the coefficients of DARS have the correct sign but are statisti- cally insignificant. In other instances, the coefficient of DARS has a negative sign. Further estimations were done with transformed interaction variables. In the case of the interaction between DARS and PHE percents, the variable con- structed was DARS*(PHE percent ­ 0.5). The coefficient associated with this vari- able reveals the impact of the difference between the PHE percent and a threshold of 50 percent. The results for IR, IFFC, IRD, and IECS are not satisfac- tory: the coefficient of the new interaction variable has the wrong sign, is not statistically significant, or both. Only in the case of DALE did we obtain a satis- factory result: both the coefficients of DARS and the interaction variable have the expected sign and are statistically significant. This result is presented in table 1.13 (appendix 1A). In other words, for advanced risk-sharing systems with a PHE percent above 50, the level of the PHE percent reinforces the "average" effect of DARS. For instance, in the case of Oman with a PHE percent of 63.31, the combined impact of DARS and DARS*(PHE percent ­ 0.5) becomes ­0.2694. For countries with a PHE percent below 50 (Chile, Republic of Korea, Brunei Darussalam, and United Arab Emirates), the initial effect of DARS is weakened. For instance, for Chile with a PHE percent of 40.1, the combined effect of DARS and DARS*(PHE percent ­ 0.5) on the dependent variable becomes ­0.1637. Key conclusions can be drawn from the various estimates. A first conclusion is that the extent of advanced risk sharing, as measured by the dummy variable DARS, is significant in the equations for four of the five goal measurements. No effect could be found in the case of the index of fair financing, but we submit this is due to the small sample size. In addition, in at least two of these measure- ments (level of responsiveness and distribution of health), the variable DMRS also has been shown to have a statistically significant impact. Second, when enlarging the set of explanatory variables in the models for the distributional measures with the Gini index, DARS remains statistically signifi- cant in the equations for IRD and IECS. In addition, DMRS has a statistically sig- nificant impact in the equations for IRD. An additional interpretation emerges from the results, namely that risk sharing corrects for, or may even outweigh, the negative effect of overall income inequality on the fair financing index and the index of distribution of responsiveness. Third, using interaction terms with PHE percent leads to plausible results for DALE only: the level of PHE percent reinforces the average positive effect of advanced risk sharing. An analysis with preliminary updated data was also undertaken; since publi- cation of the World Health Report 2000, WHO has developed updated estimates Rich-Poor Differences in Health Care Financing 41 for the level (HEC) and share of public health expenditure in total health expen- diture (PHE percent). When using updated data for HEC in the equations for DALE and IR, similar results to those presented here are obtained (in terms of explanatory power, sign, and statistical significance of coefficients). The use of the updated PHE percent does not significantly change the estimates for the equations with the interaction terms. Estimates of the index of fair financing (IFFC) were also obtained for an additional 30 countries. Reestimation of the equations, using an enlarged sample of 50, now leads to two interesting results: the advanced risk-sharing dummy variable DARS exerts a statistically significant effect on the fair financing index; and the Gini index has a statistically signifi- cant impact on IFFC but is counterbalanced by a health-financing system char- acterized by advanced risk sharing. These preliminary results prove to be more in line with those obtained for the other distributional measures. CONCLUSIONS AND RECOMMENDATIONS Most community-financing schemes have evolved in the context of severe eco- nomic constraints, political instability, and lack of good governance. Usually, government taxation capacity is weak, formal mechanisms of social protection for vulnerable populations absent, and government oversight of the informal health sector lacking. In this context of extreme public sector failure, community involvement in the financing of health care provides a critical, though insuf- ficient, first step in the long march toward improved access to health care for the poor and social protection against the cost of illness. It should be regarded as a complement to--not as a substitute for--strong government involvement in health care financing and risk management related to the cost of illness. Based on an extensive survey of the literature, the main strengths of commu- nity-financing schemes are the degree of outreach penetration achieved through community participation, their contribution to financial protection against ill- ness, and increase in access to health care by low-income rural and informal sec- tor workers. Their main weaknesses are the low volume of revenues that can be mobilized from poor communities, the frequent exclusion of the poorest from participation in such schemes without some form of subsidy, the small size of the risk pool, the limited management capacity that exists in rural and low-income contexts, and isolation from the more comprehensive benefits often available through more formal health-financing mechanisms and provider networks. The results of the macrolevel cross-country analysis presented in this report give empirical support to the hypothesis that risk sharing in health financing mat- ters in terms of its impact on both the level and distribution of health, financial fairness, and responsiveness indicators. The results even suggest that risk sharing corrects for, and may outweigh, the negative effect of overall income inequality, suggesting that financial protection against the cost of illness may be a more effec- tive poverty alleviation strategy in some settings than direct income support. 42 Health Financing for Poor People: Resource Mobilization and Risk Sharing FIGURE 1.8 Stages of Financial Protection Renewed policy commitment Broad insurance coverage Optimized subsidy of low Advocacy and income by high income consumer protection Insurance pool funding and reinsurance consolidation Interpool subsidies and consolidation policies Established insurance Framework for pool pools management and interactions Set-up funding and reinsurance Evolution of community- based Capacity building and Commitment to dissociation of pools technical support contribution from utilization Evidence-based advocacy National Dominance of out-of- Donor policy thrusts pocket payments policy thrusts Source: Adapted from Arhin-Tenkorang (2001). The results of the microlevel household data analysis indicate that prepay- ment and risk sharing through community involvement in health care financ- ing--no matter how small--increases access by poor populations to basic health services and protects them to a limited extent against the impoverishing effects of illness. Community involvement alone is not sufficient in preventing social exclusion since the very poorest often do not participate fully in these schemes. However, the study provides evidence that this constraint in reaching the poor- est could be overcome through well-targeted design features and implementa- tion arrangements. The Asia regional review supports many of these conclusions. In particular, the review emphasizes that, although income is a key constraint to participation by the very poorest, even they are often willing to participate if their contribu- tions are supplemented by a government subsidy and if the benefits they receive provide access to quality services that address their most frequent health prob- lems. In the context of extreme resource constraints, this creates a tension or Rich-Poor Differences in Health Care Financing 43 trade-off between prepayment for basic services and the need for insurance cov- erage for rarer, more expensive, and life-threatening events that may only hap- pen once or twice in a lifetime. This highlights an area of market failure relating to voluntary community involvement in health care financing that needs to be addressed by appropriate government policies since it is precisely during hospi- talization that many of the poor become even more impoverished. More rigorous research is still needed on understanding the institutional strengths and weaknesses of community involvement in health care financing, and in monitoring and evaluating their impacts on financial protection, increas- ing access to needed health care, and combating social exclusion of the poor. Yet the research for this report points to five key policies available to governments for improving the effectiveness and sustainability of community financing: (1) in- creased and well-targeted subsidies to pay the premiums of low-income popula- tions, (2) insurance to protect against expenditure fluctuations and reinsurance to enlarge the effective size of small risk pools, (3) effective prevention and case- management techniques to limit expenditure fluctuations, (4) technical support to strengthen the management capacity of local schemes, and (5) establishment and strengthening of links with the formal financing and provider networks. See page 46 for acknowledgments, notes, and references. APPENDIX 1A. STATISTICAL DATA (SUMMARY TABLES) TABLE 1.12 Estimation Resultsa for the Basic Models Explanatory variables DALEb IRc IFFCb IRDd IECSe Ln (80­ DALE) (Logit) (Logit) (Logit) (Logit) Constant 4.9423 ­0.4896 2.2874 1.6327 0.2798 (0.3328) (0.2160) (0.2786) (0.4507) (0.2038) (14.8493) (­2.2663) (8.2099) (3.6228) (1.3329) HEC ­0.1919 0.0000 (0.0197) (0.0003) (­9.7498) (0.1150) EDU ­0.2141 0.0032 (0.0834) (0.0026) (­2.5684) (1.2540) DARS ­0.2963 0.7244 ­0.1146 4.2257 6.6269 (0.0654) (0.2244) (0.6072) (0.8228) (0.3868) (­4.5321) (3.2275) (­0.1888) (5.1355) (17.1343) DSHI ­0.2521 ­1.4049 (0.1987) (0.9107) (­1.2688) (­1.5427) DMRS 0.2673 0.7217 1.0737 (0.1148) (0.5355) (0.4202) (2.3294) (1.3478) 2.5550 DMRS1 ­0.1079 (0.4607) (­0.2343) DMRS2 ­0.6458 (0.3995) (­1.6165) R­squared 0.7874 0.5678 0.0021 0.5749 0.8778 Adjusted R­squared 0.7821 0.4597 ­0.0566 0.5276 0.8671 S.E. of regression 0.2639 0.2134 1.0791 1.1924 0.7350 Ak. Info criterion 0.2049 ­0.0525 3.0894 3.3097 2.3149 Sample size 124 26 19 31 51 a. The first and second coefficients in the parentheses refer to the standard error and t-statistic, respectively. b. Restricted samples. c. Bulgaria is excluded from the sample. d. Chile and Poland are excluded from the full sample. e. Uzbekistan is excluded from the restricted sample. 44 Appendix--Statistical Data 45 TABLE 1.13 Estimation Resultsa for the Enlarged Models Explanatory variables IFFCb IRD IECSb DALEb (Logit) (Logit) (Logit) Ln(80­ DALE)) Constant 2.8260 3.0610 ­0.7471 4.9446 (1.3698) (0.7956) (0.9164) (0.3306) (2.0630) (3.8539) (­0.8153) (14.9580) Gini ­0.0119 ­0.0375 0.0355 (0.0296) (0.0180) (0.0206) (­0.4020) (­2.0853) (­0.8153) DARS ­0.2568 2.1713 5.3537 ­0.2088 (0.7162) (0.5222) (0.5531) (0.0843) (­0.3586) (4.1577) (9.6789) (­2.4774) DARS*[PHE percent­ 0.5] ­0.4556 (0.2798) (­1.6284) DMRS 0.9873 (0.4637) (2.1291) HEC ­0.1897 (0.0196) ­9.6837 EDU ­0.2166 (0.0828) (­2.6155) R­squared 0.0121 0.5191 0.7053 0.7920 Adjusted R­squared ­0.1114 0.4590 0.6906 0.7850 S.E. of regression 1.1067 0.9320 1.1912 0.2621 Ak. Info criterion 3.1846 2.8286 3.2550 0.1990 Sample size 19 28 43 124 a. The first and second coefficients in the parentheses refer to the standard error and t-statistic, respectively. b. Restricted samples. 46 Health Financing for Poor People: Resource Mobilization and Risk Sharing Acknowledgments: Valuable guidance on methodological issues was provided by Adam Wagstaff. We are also indebted to the following individuals for data access, guidance on research methodologies, reviews, and other indirect contributions: Christian Jacquier, Christian Baeaza, Michael Cichon, Chris Murray, Kei Kawabatak, Christopher Lovelace, Helen Saxenian, Jack Langenbrunner, David Peters, George Schieber, Charlie Griffin, Agnes Soucat, Abdo S. Yazbeck, Mariam Claeson, Flavia Bustreo, Steve Cummings, and Shanta Devarajan. The authors are also grateful for the access provided to parallel and ongoing research on community financing by the World Bank, the World Health Organization (WHO), and the International Labour Organisation (ILO), with important inputs from Harvard University, the London School of Hygiene and Tropical Medicine, the University of Lyon, Partnerships for Health Reform--Abt Associates Inc. (USA), the National Council for Economic Research (India), the Center for Development Research (ZEF) (Germany), and Chulalongkorn University Faculty of Economics (Thailand). The authors are grateful to WHO for having provided an opportunity to contribute to the work of the Commission on Macroeconomics and Health and to the World Bank for having published the material in this chapter as an HNP Discussion Paper. NOTES 1. This model is similar to the two-part demand model developed as part of the Rand Health Insurance experiment to estimate demand for health care services (Duan and others 1982; Manning and others 1987). For a recent application of the model that ana- lyzes the access impact of school health insurance in Egypt, see Yip and Berman 2001. 2. "Best" according to the adjusted R-squared and/or the Akaike criterion, as well as the theoretical consistency of the model. In addition, we present only the results using restricted samples where data points have been deleted from the "full" samples because of uncertainty in the risk-sharing classification, or restricted samples with additional deletion of influential data. REFERENCES Abel-Smith, B. 1988. "The Rise and Decline of the Early HMOs: Some International Expe- riences." Milbank Quarterly 66(4): 694­719. Abel-Smith, B., and P. Rawal. 1994. "Employer's Willingness to Pay: The Case for Compul- sory Health Insurance in Tanzania." Health Policy and Planning 9(4): 409­18. ADB (Asian Development Bank). 2000. Finance for the Poor: Microfinance Development Strat- egy. Manilla. Arhin-Tenkorang, D. C. 1994. "The Health Card Insurance Scheme in Burundi: A Social Asset or a Non-Viable Venture?" Social Science and Medicine 39(6): 861­70. ------. 1995. Rural Health Insurance: A Viable Alternative to User Fees. London School of Hygiene and Tropical Medicine, Discussion Paper. London. ------. 2000. "Mobilizing Resources for Health: The Case for User Fees Re-Visited." Report submitted to Working Group 3 of the Commission on Macroeconomics and Health, Jeffrey D. Sachs (Chairman). World Health Organization (WHO), Geneva. Rich-Poor Differences in Health Care Financing 47 ------. 2001. "Health Insurance for the Informal Sector in Africa: Design Features, Risk Protection, and Resource Mobilization." Report submitted to Working Group 3 of the Commission on Macroeconomics and Health, Jeffrey D. Sachs (Chairman). WHO, Geneva. Arrow, K. W. 1963. "Uncertainty and the Welfare Economics of Medical Care." American Economic Review 53(5): 940­73. Atim, C. 1998. Contribution of Mutual Health Organizations to Financing, Delivery, and Access to Health Care: Synthesis of Research in Nine West and Central African Countries. Technical Report 18. Partnerships for Health Reform Project, Abt Associates Inc., Bethesda, Md. ------. 1999. "Social Movements and Health Insurance: A Critical Evaluation of Volun- tary, Non-Profit Insurance Schemes with Case Studies from Ghana and Cameroon." Social Science and Medicine 48(7): 881­96. Atkinson A. B., and J. E. Stiglitz. 1980. Lectures on Public Economics. Maidenhead, Berkshire: McGraw-Hill. Barer, L. M., T. E. Getzen, and G. L. Stoddart, eds. 1998. Health, Health Care, and Health Eco- nomics: Perspectives on Distribution. Chichester, West Sussex: John Wiley and Sons. Bator, F. 1958. "The Anatomy of Market Failure." Quarterly Journal of Economics 72(3): 351­79. Bennett S., A. Creese, and R. Monasch. 1998. Health Insurance Schemes for People Outside Formal Sector Employment. ARA Paper No. 16. WHO, Geneva. Brown, W., and C. Churchill. 2000. Insurance Provision in Low-Income Communities: Part II--Initial Lessons from Microinsurance Experiments for the Poor. Microenterprise Best Prac- tices Project, Development Alternatives Inc., Bethesda, Md. Carrin, G., M. Desmet, and R. Basaza. 2001. "Social Health Insurance Development in Low-Income Developing Countries: New Roles for Government and Non-Profit Health Insurance Organisations." In X. Scheil-Adlung and D. D. Hoskins, eds., Build- ing Social Security: The Challenge for Privatization. Geneva: International Social Security Association. Carrin, G., R. Zeramdini, P. Musgrove, J-P. Poullier, N. Valentine, and K. Xu. 2001. The Impact of the Degree of Risk-Sharing in Health Financing on Health System Attainment. Report submitted to Working Group 3 of the Commission on Macroeconomics and Health, Jeffrey D. Sachs (Chairman). WHO, Geneva. Claeson, M., C. C. Griffin, T. A. Johnston, M. McLachlan, A. Wagstaff, and A. S. Yazbeck. 2001. Health, Nutrition, and Population Sourcebook for the Poverty Reduction Strategy Paper. HNP Discussion Paper. Washington, D.C.: World Bank. Criel, B., P. Van der Stuyft, and W. Van Lerberghe. 1999. "The Bwamanda Hospital Insur- ance Scheme: Effective for Whom? A Study of Its Impact on Hospitalisation and Utili- sation Patterns." Social Science and Medicine 48(7): 879­911. Dave, P. 1991. "Community and Self-Financing in Voluntary Health Programmes in India." Health Policy and Planning 6(1): 20­31. DeRoeck, D., J. Knowles, T. Wittenberg, L. Raney, and P. Cordova. 1996. Rural Health Ser- vices at Seguridad Social Campesino Facilities: Analyses of Facility and Household Surveys. Health Financing and Sustainability Project, Technical Report No. 13. Partnerships for Health Reform Project, Abt Associates, Inc., Bethesda, Md. 48 Health Financing for Poor People: Resource Mobilization and Risk Sharing Desmet, A., A. Q. Chowdhury, and K. Islam, M.D. 1999. "The Potential for Social Mobi- lization in Bangladesh: The Organization and Functioning of Two Health Insurance Schemes." Social Science and Medicine 48: 925­38. Diop, F., A. Yazbeck, and R. Bitran. 1995. "The Impact of Alternative Cost Recovery Schemes on Access and Equity in Niger." Health Policy and Planning 10(3): 223­40. Dordick G. 1997. Something Left to Lose: Personal Relations and Survival among New York's Homeless. Philadelphia: Temple University Press. Dror, D., and C. Jacquier. 1999. "Microinsurance: Extending Health Insurance to the Excluded." International Social Security Review 52(1): 71­97. Dror, D., A. S. Preker, and M. Jakab. 2002. "Role of Communities in Combating Social Exclusion." In D. Dror and A. S. Preker, eds., Social Reinsurance: A New Approach to Sus- tainable Community Health Financing. Washington, D.C.: World Bank/International Labour Organisation (ILO). Duan, N., W. G. Manning Jr., C. M. Morris, and J. P. Newhouse. 1982. A Comparison of Alternative Models for the Demand for Medical Care. Report R-2754-HHS. The Rand Cor- poration, Santa Monica, Calif. Evans, P. 1992. "The State as Problem and Solution: Predation, Embedded Autonomy, and Structural Change." In Stephan Haggard and Robert Kaufman, eds., The Politics of Eco- nomic Adjustment: International Constraints, Distributive Conflicts, and the State. Prince- ton, N.J.: Princeton University Press. ------. 1995. Embedded Autonomy: States and Industrial Transformation. Princeton, N.J.: Princeton University Press. -------. 1996. "Government Action, Social Capital, and Development: Reviewing the Evi- dence on Synergy." World Development 24(6): 1119­32. Evans, R. G. 1984. Strained Mercy. Toronto: Butterworth. Foster, G. 1982. "Community Development and Primary Health Care: Their Conceptual Similarities." Medical Anthropology 6(3): 183­95. Granovetter, M. 1973. "The Strength of Weak Ties." American Journal of Sociology 78: 1360­80. Guhan, S. 1994. "Social Security Options for Developing Countries." International Labour Review 33(1): 35­53. Gumber A. 2001. Hedging the Health of the Poor: The Case for Community Financing in India. World Bank, HNP Discussion Paper. Washington, D.C. Gumber, A., and V. Kulkarni. 2000. "Health Insurance for Informal Sector: Case Study of Gujarat." Economic and Political Weekly, 30 September, 3607­13. Gwatkin, D. 2001. "Poverty and Inequalities in Health within Developing Countries: Fill- ing the Information Gap." In D. Leon and G. Walt, eds., Poverty, Inequality, and Health: An International Perspective. Oxford: Oxford University Press. Harding, A., and A. S. Preker. 2001. A Framework for Understanding Organizational Reforms in the Hospital Sector. World Bank, HNP Discussion Paper. Washington, D.C. Hsiao, W. C. 2001. Unmet Health Needs of Two Billion: Is Community Financing a Solution? Report submitted to Working Group 3 of the Commission on Macroeconomics and Health, Jeffrey D. Sachs (Chairman). WHO, Geneva. Rich-Poor Differences in Health Care Financing 49 ILO (International Labour Organisation). 2000a. Mutuelles de santé en Afrique: Charactéris- tiques et mise en place. Manuel de formateurs. Programme Stratégies et Technique contre l'Exclusion sociale et la Pauvreté, Département de Sécurité Sociale (STEP). Geneva. ------. 2000b. World Labour Report: Income Security and Social Protection in a Changing World. Geneva. ------. 2001. "Mutuelles de santé et associations de microentrepreneurs--Guide." Pro- gramme Stratégies et Technique contre l'Exclusion Sociale et la Pauvreté, Département de Sécurité Sociale (STEP), Small Enterprise Development Programme (SEED). Geneva. Jakab, M., and C. Krishnan. 2001. Community Involvement in Health Care Financing: A Sur- vey of the Literature on the Impact, Strengths, and Weaknesses. Report submitted to Work- ing Group 3 of the Commission on Macroeconomics and Health, Jeffrey D. Sachs (Chairman). WHO, Geneva. Jakab, M., A. S. Preker, C. Krishnan, P. Schneider, F. Diop, J. Jütting, A. Gumber, K. Ranson, and S. Supakankunti. 2001. Social Inclusion and Financial Protection through Community Financing: Initial Results from Five Household Surveys. Report submitted to Working Group 3 of the Commission on Macroeconomics and Health, Jeffrey D. Sachs (Chair- man). WHO, Geneva. Jütting, J. 2000. Do Mutual Health Insurance Schemes Improve the Access to Health Care? Pre- liminary Results from a Household Survey in Rural Senegal. Paper presented at the Interna- tional Conference on Health Systems Financing in Low-Income African and Asian Countries, November 30-December 1, Centre d`Etudes et de Recherches sur le Développement International (CERDI), Clermont-Ferrand, France. ------. 2001. The Impact of Health Insurance on the Access to Health Care and Financial Pro- tection in Rural Areas of Senegal. World Bank, HNP Discussion Paper. Washington, D.C. Litvack, J. I., and C. I. Bodart. 1992. "User Fees Plus Quality Equals Improved Access to Health Care: Results of a Field Experiment in Cameroon." Social Science and Medicine 37(3): 369­83. Manning, W. G., J. P. Newhouse, N. Duan, E. B. Keeler, A. Leibowitz, and M. S. Marquis. 1987. "Health Insurance and the Demand for Medical Care: Evidence from a Random- ized Experiment." American Economic Review 77(3): 251­77. McPake, B., K. Hanson, and A. Mills. 1993. "Community Financing of Health Care in Africa: An Evaluation of the Bamako Initiative." Social Science and Medicine 3(11): 1383­95. Midgley, J., and M. B. Tracey. 1996. Challenges to Social Security: An International Explo- ration. Westport, Conn.: Auburn House. Musau, S. 1999. Community-Based Health Insurance: Experiences and Lessons Learned from East and Southern Africa. Technical Report 34. Partnerships for Health Reform Project, Abt Associates Inc., Bethesda, Md. Musgrave, R. A., and P. B. Musgrave. 1984. Public Finance in Theory and Practice. New York: McGraw-Hill. Narayan, D. 1999. "Bonds and Bridges: Social Capital and Poverty." Policy Research Work- ing Paper 2167. World Bank, Poverty Reduction and Economic Management Network, Washington, D.C. Navarro, V. A. 1984. "Critique of the Ideological and Political Position of the Brandt Report and the Alma Ata Declaration." International Journal of Health Services 2(14): 159­72. 50 Health Financing for Poor People: Resource Mobilization and Risk Sharing North, D. C. 1990. Institutions, Institutional Change, and Economic Performance. New York: Cambridge University Press. Otero, M., and E. Rhyne, eds. 1994. The New World of Microenterprise Finance: Building Healthy Financial Institutions for the Poor. West Hartford, Conn.: Kumarian Press. Peters, D., A., G. Yazbeck, N. V. Ramana, and R. Sharma. 2001. India--Raising the Sights: Better Health Systems for India's Poor. Washington, D.C.: World Bank. Platteau, J. P. 1994. "Behind the Market Stage Where Real Societies Exist" (parts 1 and 2). Journal of Development Studies 30: 533­77 and 753­817. Pradhan, M., and N. Prescott. 2000. Social Risk Management for Medical Care in Indonesia. Paper presented at the International Conference on Health Systems Financing in Low- Income African and Asian Countries, November 30­December 1, CERDI, Clermont- Ferrand, France. Preker, A. S. 1998. "The Introduction of Universal Access to Health Care in the OECD: Lessons for Developing Countries." In S. Nitayarumphong and A. Mills, eds., Achieving Universal Coverage of Health Care. Bangkok: Ministry of Public Health. Preker A. S., H. Harding, and N. Girishankar. 2001. "Private Participation in Supporting the Social Contract in Health: New Insights from Institutional Economics." In A. Ron and X. Scheil-Adlung, eds., Recent Health Policy Innovations in Social Security, Interna- tional Social Security Series, vol. 5. London: Transaction Publishers. Preker, A. S., J. Langenbrunner, and M. Jakab. 2002. "Development Challenges in Health Care Financing," In D. Dror and A. S. Preker, eds., Social Reinsurance: A New Approach to Sustainable Community Health Financing. Washington, D.C.: World Bank/ILO. Preker A. S., J Langenbrunner, and E Suzuki. 2001. The Global Expenditure Gap: Securing Financial Protection and Access to Health Care for the Poor. Report submitted to Working Group 3 of the Commission on Macroeconomics and Health, Jeffrey D. Sachs (Chair- man). WHO, Geneva. Preker, A. S., J Langenbrunner, M. Jakab, and C. Baeza. 2001. Resource Allocation and Pur- chasing (RAP) Arrangements That Benefit the Poor and Excluded Groups. World Bank, HNP Discussion Paper. Washington, D.C. Preker A. S., G. Carrrin, D. Dror, M. Jakab, W. Hiao, and D. Arhin-Tenkorang. 2002a. "Effectiveness of Community Health Financing in Meeting the Cost of Illness." Bulletin of the World Health Organization 8(2): 143­50. ------, eds. 2002b. Chapter 4, "Alternative Ways of Financing Health in the Informal Sec- tor." In Jeffrey D. Sachs, Alan Tait, and Kwesi Botchwey (Chairmen), Report of Working Group 3 of the Commission on Macroeconomics and Health. Geneva: WHO. Ranson, K. 2001. The Impact of SEWA's Medical Insurance Fund on Hospital Utilization and Expenditure in India. World Bank, HNP Discussion Paper. Washington, D.C. Saltman R. B., and O. Ferroussier-Davis. 2000. "The Concept of Stewardship in Health Pol- icy." Bulletin of the World Health Organization 78(6): 732­39. Schieber G., and A. Maeda. 1997. "A Curmudgeon's Guide to Health Care Financing in Developing Countries." In G. Schieber, ed., Innovations in Health Care Financing. Washing- ton D.C.: World Bank. (Proceedings of a World Bank Conference, March 10­11, 1997.) Rich-Poor Differences in Health Care Financing 51 Schneider P., and F. Diop. 2001. Synopsis of Results on the Impact of Community-Based Health Insurance (CBHI) on Financial Accessibility to Health Care in Rwanda. World Bank, HNP Discussion Paper. Washington, D.C. Soucat A., T. Gandaho, D. Levy-Bruhl, X. de Bethune, E. Alihonou, C. Ortiz, P. Gbedonou, P. Adovohekpe, O. Camara, J. M. Ndiaye, B. Dieng, and R. Knippenberg. 1997. "Health Seeking Behavior and Household Expenditures in Benin and Guinea: The Equity Impli- cations of the Bamako Initiative." International Journal of Health Planning and Manage- ment 12(S1): 137­63. Supakankunti, S. 1997. Future Prospects of Voluntary Health Insurance in Thailand. Takemi Research Paper 13. Takemi Program in International Health, Harvard School of Public Health. Harvard University, Cambridge, Mass. ------. 2001. Determinants for Demand for Health Card in Thailand. World Bank, HNP Dis- cussion Paper. Washington, D.C. van Doorslaer E., A. Wagstaff, and F. Rutten, eds. 1993. Equity in the Finance and Delivery of Health Care: An International Perspective. Oxford: Oxford Medical Publications. Van Ginneken, W. 1999. "Social Security for the Informal Sector: New Challenges for the Developing Countries." International Social Security Review 52(1): 49­69. Vogel, R. J. 1990. Health Insurance in Sub-Saharan Africa. Working Paper Series 476. World Bank, African Technical Department, Washington, D.C. Wagstaff, A. N., N. Watanabe, and E. van Doorslaer. 2001. Impoverishment, Insurance, and Health Care Payments. World Bank, HNP Discussion Paper. Washington, D.C. Woolcock, M., 1998. "Social Capital and Economic Development: Toward a Theoretical Synthesis and Policy Framework." Theory and Society 27: 151­208. Woolcock, M., and D. Narayan. 2000. "Social Capital: Implications for Development The- ory, Research, and Policy." World Bank Research Observer 15(2): 225­49. World Bank. 1993. World Development Report 1993: Investing in Health. New York: Oxford University Press. ------. 1995. World Development Report: Workers in an Integrating World. Washington D.C. ------. 1997. Sector Strategy for HNP. Washington, D.C. WHO (World Health Organization). 2000. World Health Report 2000: Health Systems-- Measuring Performance. Geneva. Yip, W., and P. Berman. 2001. "Targeted Health Insurance in a Low-Income Country and Its Impact on Access and Equity in Access: Egypt's School Health Insurance." Health Economics 10(2): 207­20. Zeller, M., and M. Sharma. 2000. "Many Borrow, More Save, and All Insure: Implications for Food and Microfinance Policy." Food Policy 25(2): 143­67. Ziemek, S., and J. Jütting. 2000. Mutual Insurance Schemes and Social Protection. STEP Research Group on Civil Society and Social Economy. Geneva: ILO. CHAPTER 2 Review of the Strengths and Weaknesses of Community Financing Melitta Jakab and Chitra Krishnan Abstract: The chapter reviews 45 published and unpublished reports on community financing completed between 1990 and 2001. The main objective of the study was to explore performance measures reported in the literature regarding community financing. The study concluded that the reviewed literature is rich in describing scheme design and implementation. At the same time, evidence on the performance of community-financing schemes is limited. The study focused on reporting measures on three indicators in particular: · Resource mobilization capacity. Community-financing mechanisms mobilize significant resources for health care. However, there is a large variation in the resource mobiliza- tion capacity of various schemes. This review did not find systematic estimates about how much community-financing contributes to health revenues at the local or national level. · Social inclusion. Community financing is effective in reaching a large number of low- income populations who would otherwise have no financial protection against the cost of illness. There is large variation in the size of various schemes. At the same time, there are no estimates about the total population covered through community financ- ing. There are indications that the poorest and socially excluded groups are not auto- matically reached by community-financing initiatives. · Financial protection. Community-based health-financing schemes are systematically reported to reduce the out-of-pocket spending of their members while increasing their utilization of health care services. C ommunity-based health-financing (CF) mechanisms play an increasingly important role in the health system of many low- and middle-income coun- tries. The expectation is that CF mechanisms reach population groups that government and market-based health-financing arrangements do not. Popula- tions with low income, who obtain their subsistence in the informal sector (urban and rural), and socially excluded groups (excluded because of cultural factors, physical or mental disability, or other chronic illness) are often not able to take advantage of government or market-based health-financing arrangements. Thus CF has been attracting widespread attention for its potential to provide vulnerable population groups with increased financial protection and access to health care. With increasing interest in CF in academic, development, and policy circles, the relevant literature has also been growing. In the past decade, there has been an exponential growth in the number of conceptual works, country case studies, comparative papers, and empirical papers describing and analyzing various aspects of community-based health financing. 54 Health Financing for Poor People: Resource Mobilization and Risk Sharing Despite the diverse nature of the literature, the papers reviewed for this study address many similar questions. What is community financing? How should this phenomenon that has been discussed under many different terms be described? What explains the growing enthusiasm for community financing? Are these ini- tiatives successful in raising resources for health care, reducing the financial bur- den of those seeking care, and increasing access to health care? If so, what allows these schemes to succeed where the more entrenched institutions of govern- ments and markets have failed? This chapter provides a comprehensive review of CF literature dating from between 1990 and 2001. The chapter seeks to answer three specific questions: What is community-based health financing and what are its main modalities? How do community-financing schemes perform as health-financing instru- ments in terms of mobilizing resources--including the poor's--and providing financial protection by removing financial barriers to access? What are the key structural determinants that explain the performance of community financing? In addition to addressing these specific questions, we hope to determine the focus of studies completed to date; integrate common findings; identify knowl- edge gaps; and present key contested issues requiring further research. The chapter is structured as follows. In the first section, it discusses the methodology of the review. Next, it discusses the definition and possible modal- ities of CF. In the third section, it reviews the performance of CF with regard to resource mobilization capacity, social inclusion, and financial protection. Finally, it discusses the performance determinants of CF with regard to key tech- nical design features and management, organizational, and institutional charac- teristics. Then it presents the conclusions. METHODS Forty-five papers were reviewed for this study. Broad selection criteria were applied: studies were included in the review if their main objective was to dis- cuss health-financing arrangements in which the community was actively involved in some form. The selected papers included articles published in peer- reviewed journals, reports published in formal publication series of international organizations (such as the World Health Organization and the International Labour Organization), internal unpublished documents of international organi- zations and academic institutions, and conference proceedings. Table 2.1 pre- sents the breakdown of the reviewed studies according to publication type. Of these 45 papers, 6 were conceptual in nature, 8 were large-scale compara- tive papers (analyzing 5 or more community financing schemes), and the remaining 31 were case studies. The regional breakdown of the case studies is fairly even between Africa and Asia (15 and 11, respectively), but there are only 4 on Latin America (see table 2.2). This breakdown reflects selection bias: litera- ture available only in Spanish was not included in the review. Review of the Strengths and Weaknesses of Community Financing 55 TABLE 2.1 Summary Statistics of the Literature Reviewed, by Publication Type Peer-reviewed Published Internal documents of international Conference journal articles reports organizations or academic institutions proceedings Number of studies 20 16 5 4 The analytical approach applied to the 45 papers (see figure 2.1) followed the framework proposed by Preker and others (2001). Health care financing through community-based involvement can be seen as having three independent objec- tives: (1) it provides the financial resources to promote better health and to diag- nose, prevent, and treat known illness; (2) it provides an opportunity to protect individuals and households against direct financial cost of illness when channeled through risk-sharing mechanisms; and (3) it gives the poor a voice and makes them active participants in breaking out of the social exclusion in which they are often trapped. These three objectives can be influenced through the design of CF schemes in terms of technical characteristics of revenue collection, pooling, and purchasing; and management, organizational, and institutional characteristics. To approximate the three objectives of CF as proposed above, the review defined the following research questions that can be answered from available studies: What is the potential of CF schemes to mobilize resources in a sustainable manner? · What is the contribution of community financing to the resources available for local health systems? · What is the share of community financing in total health revenues (of a dis- trict, state, country)? · How does community financing compare to other resource mobilization instruments in terms of per capita amount of resources mobilized? Is CF inclusive of the poor? · Do community-financing schemes reach the poor? What is the socioeco- nomic composition of schemes? How effective are CF schemes in preventing impoverishment due to the cost of illness? · Do CF scheme members have better access to health care than nonmembers? · Does CF eliminate the financial barriers to health care? Of the 45 studies, 31 were included in the review of performance. Table 2.3 presents the list of variables we used as selection criteria for the performance section and the number of studies that reported the selected performance variable. Of the 45 studies, 26 provided some information to assess resource-generation capacity of CF, 13 pro- vided information on social inclusion, and 20 provided useful information to assess financial protection. Appendix 2A presents the detailed list of the 45 reviewed stud- ies and the kind of performance variables they report. 56 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 2.2 Summary of Literature Reviewed on Community-Based Health Financing Schemes, Based on Nature of Study and by Region Conceptual studies 1. Dror, D., and C. Jacquier. 1999. "Micro-Insurance: Extending Health Insurance to the Excluded." 2. Brown, W., and C. Churchill. 2000. Insurance Provision in Low-Income Communities: Part II--Initial Lessons from Micro-insurance Experiments for the Poor. 3. Ziemek, S., and J. Jütting. 2000. Mutual Insurance Schemes and Social Protection. 4. Criel, B. 2000. Local Health Insurance Systems in Developing Countries: A Policy Research Paper. 5. Ekman, B. 2001. Community-Based Health Insurance Schemes in Developing Countries: Theory and Empirical Experiences. 6. Hsiao, W. C. 2001. Unmet Needs of Two Billion: Is Community Financing a Solution? Large-scale comparative studies (> 5 schemes) 7. Dave, P. 1991. "Community and Self-Financing in Voluntary Health Programmes in India." 8. McPake, B., K. Hanson, and A. Mills. 1993. "Community Financing of Health Care in Africa: An Evaluation of the Bamako Initiative." 9. Gilson, L. 1997. "The Lessons of User Fee Experience in Africa." 10. Atim, C. 1998. Contribution of Mutual Health Organizations to Financing, Delivery, and Access to Health Care: Synthesis of Research in Nine Western and Central-African Countries. 11. Bennett, S., A. Creese, and R. Monasch. 1998. Health Insurance Schemes for People Outside Formal Sector Employment. 12. Musau, S. 1999. Community-Based Health Insurance: Experiences and Lessons Learned from East and Southern Africa. 13. CLAISS. 1999. "Synthesis of Micro-Insurance and Other Forms of Extending Social Protection in Health in Latin America and the Caribbean." 14. Narula, I. S., and others. 2000. Community Health Financing Mechanisms and Sustainability: A Comparative Analysis of 10 Asian Countries. Case studies--Africa 15. Arhin, D. C. 1994. "The Health Card Insurance Scheme in Burundi: A Social Asset or a Non-Viable Venture?" 16. Diop, F., R. Bitran, and M. Makinen. 1994. Evaluation of the Impact of Pilot Tests for Cost Recovery on Primary Health Care in Niger. 17. Arhin, D.C. 1995. Rural Health Insurance: A Viable Alternative to User Fees. 18. Diop F., A. Yazbeck, and R. Bitran 1995. "The Impact of Alternative Cost Recovery Schemes on Access and Equity in Niger." 19. Ogunbekun, I., and others. 1996. "Costs and Financing of Improvements in the Quality of Maternal Health Services through the Bamako Initiative in Nigeria." 20. Roenen, C., and B. Criel 1997. The Kanage Community Financed Scheme: What Can Be Learned from the Failure? 21. Soucat, A., T. Gandaho, and others. 1997. "Health Seeking Behavior and Household Expenditures in Benin and Guinea: The Equity Implications of the Bamako Initiative." 22. Soucat, A., D. Levy-Bruhl, and others. 1997. "Local Cost Sharing in Bamako Initiative Systems in Benin and Guinea: Assuring the Financial Viability of Primary Health Care." (continued) Review of the Strengths and Weaknesses of Community Financing 57 TABLE 2.2 Continued Case studies--Africa (continued) 23. Atim, C. 1999. "Social Movements and Health Insurance: A Critical Evaluation of Voluntary, Nonprofit Insurance Schemes with Case Studies from Ghana and Cameroon." 24. Criel, B., P. van der Stuyft, and W. van Lerberghe. 1999. "The Bwamanda Hospital Insurance Scheme: Effective for Whom? A Study of Its Impact on Hospitalization Utilization Patterns." 25. Atim, C., and M. Sock. 2000. An External Evaluation of the Nkoranza Community Financing Health Insurance Scheme, Ghana. 26. Jütting, J. 2000. Do Mutual Health Insurance Schemes Improve the Access to Health Care? Preliminary Results from a Household Survey in Rural Senegal. 27. Schneider, P., F. Diop, and S. Bucyana. 2000. Development and Implementation of Prepayment Schemes in Rwanda. 28. Gilson, L., and others. 2000. "The Equity Impacts of Community Financing Activities in Three African Countries." 29. Okumara, J., and T. Umena. 2001. "Impact of Bamako Type Revolving Drug Fund on Drug Use in Vietnam." Case studies--Asia 30. Hsiao, W. C. 1995. "The Chinese Health Care System: Lessons for Other Nations." 31. Ron, A., and A. Kupferman. 1996. A Community Health Insurance Scheme in the Philippines: Extension of a Community-Based Integrated Project. 32. Liu, Y., and others. 1996. "Is Community Financing Necessary and Feasible for Rural China?" 33. Supakankunti, S. 1997. Future Prospects of Voluntary Health Insurance in Thailand. 34. Supakankunti, S. 1998. Comparative Analysis of Various Community Cost Sharing Implemented in Myanmar. 35. Carrin, G., and others. 1999. `'The Reform of the Rural Cooperative Medical System in the People's Republic of China: Interim Experience in 14 Pilot Counties." 36. Desmet, A., A. Q. Chowdhury, and K. Islam. 1999. "The Potential for Social Mobilization in Bangladesh: The Organization and Functioning of Two Health Insurance Schemes." 37. Chen, N., A. Ma, and Y. Guan. 2000. `'Study and Experience of a Risk-Based Cooperative Medical System in China: Experience in Weifang of Shandong Province." 38. Gumber, A., and V. Kulkarni. 2000. "Health Insurance for Informal Sector: Case Study of Gujarat." 39. Xing-yuan, G., and F. Xue-shan. 2000. "Study on Health Financing in Rural China." 40. Preker, A. S. 2001. "Philippines Mission Report." Case studies--Latin America and the Caribbean 41. Toonen, J. 1995. Community Financing for Health Care: A Case Study from Bolivia. 42. DeRoeck, D., and others. 1996. Rural Health Services at Seguridad Social Campesino Facilities: Analyses of Facility and Household Surveys. 43. Fiedler, J. L., and R. Godoy. 1999. An Assessment of the Community Drug Funds of Honduras. 44. Fiedler, J. L., and J. B. Wight. 2000. "Financing Health Care at the Local Level: The Community Drug Funds of Honduras." Case studies--mixed regions 45. Ron, A. 1999. "NGOs in Community Health Insurance Schemes: Examples from Guatemala and Philippines." FIGURE 2.1 Analytical Framework Ultimate performance indicators Health Preventing impoverishment Social inclusion Intermediate performance indicators · Level of mobilized resources · Level of health care utilization · Equity · Sustainability of resource · Financial access and barriers · Efficiency mobilization · Successful risk management · Quality Community financing schemes Core technical design Revenue-collection mechanisms characteristics · Level of prepayment compared with direct out-of-pocket spending · Extent to which contributions are compulsory compared with voluntary · Degree of progressivity of contributions · Subsidies to cover the poor Arrangements for pooling of revenues and sharing of risks · Size of the pool · Number of pools · Redistribution from rich to poor; from health to sick; and from gainfully employed to inactive Resource allocation or purchasing arrangements · Demand (for whom to buy) · Supply (what to buy, in which form, and what to exclude) · Prices and incentive regime (at what price and how to pay) Management Staff characteristics · Leadership · Capacity (management skills) Culture · Management style (top down or consensual) · Structure (flat or hierarchical) Access to information · Financial, resources, health information, behavior Organizational Organizational forms characteristics · Economies of scale and scope · Contractual relationships Organizational incentive regime · Decision rights, market exposure, financial responsibility, accountability, social functions Organizational integration/fragmentation · Horizontal, vertical Institutional Stewardship (government oversight, coordination, regulation, monitoring, etc.) characteristics Governance (public-private mix in ownership) Rules on revenue transfers and risk pooling Market structure (factor market and product market) Source: Preker and others (2001). 58 Review of the Strengths and Weaknesses of Community Financing 59 TABLE 2.3 Selection Criteria to Assess the Performance of Community-Based Health Financing Number of studies reporting selected Performance variable of interest performance variablea Resource mobilization capacity Contribution of CF to the resources of local health systems (providers) 26 Share of CF in total health revenues (of district, state, country) 1 Per capita amount of resources mobilized through CF 2b W/ control relative to other health-financing instruments 2 Social inclusion Socioeconomic composition of reviewed schemes 13 Financial protection Utilization rate of scheme members w/ control 16 OOP payment of CF members w/ control 9 a. Studies that offered conclusions about various performance criteria but did not present supporting evidence in the study were not included in this count. b. Dave (1991) provides total expenditures and covered populations for 12 schemes and thus the per capita amount could be calculated based on these figures. However, the author does not present them in this format. A number of studies offered conclusions on resource mobilization, social inclu- sion, and financial protection based on the experience of authors or review of other studies and schemes but did not provide actual evidence in support of their conclusions. We excluded these studies from the analysis. We also dropped stud- ies that reported performance figures for the scheme or schemes they analyzed but did not present controls that would enable us to make unbiased conclusions. The direct and indirect determinants of financial protection, health, and social inclusion are complex. To assess these determinants, this chapter reviewed four characteristics of community-financing arrangements: · Technical design characteristics · Management characteristics · Organizational characteristics · Institutional characteristics. Nearly all the studies reviewed provided some insight into these characteristics. The literature is particularly rich in describing the function, design, and imple- mentation of CF arrangements. WHAT IS COMMUNITY-BASED HEALTH FINANCING? The term community-based health financing has evolved into an umbrella term that covers a wide spectrum of health-financing instruments (Hsiao 2001; Dror and Jacquier 1999). Microinsurance, community health funds, mutual 60 Health Financing for Poor People: Resource Mobilization and Risk Sharing health organizations, rural health insurance, revolving drug funds, and com- munity involvement in user fee management have all been loosely referred to as community-based health financing. The rationale of referring to these diverse financing instruments under the same heading is that they exhibit a number of similarities that effectively distin- guish them from other resource mobilization instruments such as general taxa- tion, social insurance, and out-of-pocket payments. At the same time, there are important distinctions among them in terms of their core characteristics, orga- nizational structure, management, and institutional environment. These differ- ences make these arrangements dissimilar enough that comparisons are impossible without some kind of typology. In this section, we present defini- tions and categorizations from the 45 reviewed papers and establish a typology that we will use throughout the analytical part of the chapter. A number of studies offered an explicit definition for the type of community financing they investigated (Atim 1998; Ziemek and Jütting 2000; Hsiao 2001; Dror and Jacquier 1999; Musau 1999; McPake, Hanson, and Mills 1993). Box 2.1 presents these definitions. Regardless of the terminology used, the definitions converge on several points. In particular, the role of the community, the nature of the beneficiary group, and the social values underlying the design of the schemes stand out as key descriptors of the investigated health-financing arrangements. Each of these common characteristics will be reviewed in turn. The first important common feature of the definitions is reference to the pre- dominant role of community in mobilizing, pooling, allocating, and managing or supervising (or both) health care resources (Atim 1998; Ziemek and Jütting 2000; Hsiao 2001; Musau 1999; McPake, Hanson, and Mills 1993). The Oxford English Dictionary defines community as the quality of (a) "joint or common ownership, tenure or liability"; (b) "common character"; (c) "social fellowship"; (d) "life in association with others"; (e) "common or equal rights or rank"; and (f) "people organized into common political, municipal or social unity." Thus community-- according to this definition--is a broader concept than that commonly used to refer to a geographic entity defined for political and administrative purposes. Various forms of CF reflect most of the concepts in the above definition. Mem- bers of many CF schemes are bound together not only by geographic proximity but also by shared professional and cultural identity. A narrow geographic definition would exclude many CF schemes whose members are not geographically linked but instead belong to the same craft, profession, or religion or share some other kind of affiliation that facilitates their cooperation for financial protection. This is particu- larly reflected in the tradition of mutual health organizations in francophone Africa or microfinance organizations that provide health insurance to their borrowers. The predominance of community action does not mean that CF mecha- nisms do not rely on government, donor, or other external support. On the contrary, reviewers of successful community initiatives often point to the role of government and donor support--both financial and nonfinancial--as a key determinant of sustainability (Carrin and others 1999; Criel, van der Stuyft, Review of the Strengths and Weaknesses of Community Financing 61 BOX 2.1 DEFINITIONS OF COMMUNITY HEALTH FINANCING Mutual Health Organizations (MHO): "A voluntary, nonprofit insurance scheme, formed on the basis of an ethic of mutual aid, solidarity, and the col- lective pooling of health risks, in which the members participate effectively in its management and functioning. . . . [T]hey are nonprofit, autonomous orga- nizations based on solidarity between, and democratic accountability to, their members whose objective is to improve their members' access to good quality health care through their own financial contributions and by means of any of a range of financing mechanisms that mainly involve insurance, but that may also include simple prepayments, savings and soft loans, third-party subscrip- tion payments, and so on" (Atim 1998, p. 2). "Mutual insurance schemes can be broadly defined as systems based on voluntary engagement and the principles of solidarity and reciprocity. Mem- bers usually have to meet certain obligations, e.g., payment of premiums, and are bound together by a common objective and a strong local affiliation. Many times, these schemes evolve out of traditional systems or form as a response to the low coverage provided by formal systems" (Ziemek and Jütting 2000, p. 2). "Community financing can be broadly defined as any scheme that has three features: community control, voluntary, and prepayment for health care by the community members. This definition would exclude financing schemes such as regional compulsory social insurance plans and community-managed user fee programs" (Hsiao 2001, p. 4). Microinsurance is "voluntary group self-help schemes for social health insur- ance. . . . The underpinning of microinsurance is that excluded populations have not been covered under the existing health insurance schemes because of two concurrent forces. The first is that . . . insurers have done little to include these population segments. The second factor has been that excluded people have for- gone claiming access because of their disempowerment within society. Microin- surance proposes to change both factors" (Dror and Jacquier 1999, p. 78). "The term community-based health insurance is used in this study to refer to any nonprofit health financing scheme. It covers any not-for-profit insur- ance scheme that is aimed primarily at the informal sector and formed on the basis of an ethic of mutual aid and the collective pooling of health risks, and in which the members participate in its management" (Musau 1999, p. 5). "[T]he term community financing [means] a system comprising consumer payment (either as a user fee, some form of prepayment mechanism, or other charge) for health services at community level, the proceeds from which are retained within the health sector and managed at local level. . . . In addition it is sometimes argued that community financing is a form of community partic- ipation which ensures that communities are not just passive recipients of ser- vices" (McPake, Hanson, and Mills 1993, p. 1384). 62 Health Financing for Poor People: Resource Mobilization and Risk Sharing and van Lerberghe 1999; Supakankunti 1997; Atim 1998). However, the commu- nity has a predominant role in designing the rules of the game, managing and supervising the schemes in raising resources, pooling them, and allocating them. The second common feature of the definitions is the description of the benefi- ciary group. Typically, it is expected that community financing will attract those with no access to financial protection and no access to other health care financ- ing arrangements. In other words, those who are not employed in the formal sector and thus are not eligible to be part of social insurance schemes; those who cannot take advantage of general tax-financed health services because of geo- graphic access barriers, unavailability of needed care and drugs, or both; and those who cannot pay for market-based private health care. These population groups include the poor with no means of subsistence, those engaged in economic activity in the informal sector and in agriculture; and those who are socially excluded because of ethnicity, religion, mental and physical disability, or other illness (Musau 1999; Dror and Jacquier 1999; Atim 1998; Atim and Sock 2000; Gumber and Kulkarni 2000). Finally, the third common feature of the definitions is reference to the social values and principles underlying the design of community-based financing. This includes the principles of voluntary participation, built-in solidarity mecha- nisms, and reciprocity. In many societies, these principles originate from the poor's traditional self-help mechanisms, which embrace not only health (or pri- marily not health) but also many other risks with potentially devastating finan- cial implications (Atim 1999; Musau 1999; DeRoeck and others 1996). Based on the above, we adopted a broad definition of community financing that reflects all three of these common characteristics. For the purpose of this review, we included studies of health-financing arrangements characterized by the following: · The community (geographic, religious, professional, ethnic) is actively en- gaged in mobilizing, pooling, and allocating resources for health care. · The beneficiaries of the scheme have predominantly low income, earning a subsistence from the informal sector (rural and urban), or are socially excluded. · The schemes are based on voluntary engagement of the community (although not necessarily of the individual community members). · The structure of resource mobilization and benefits reflect principles of solidarity. · The primary purpose of the schemes is not commercial (that is, not-for-profit). The advantage of this broad definition is that it is inclusive of many different health-financing arrangements with these common characteristics. Further, it effectively distinguishes community-based health financing from other resource mobilization instruments, including out-of-pocket payments, voluntary private insurance, social insurance, and general taxation. At the same time, the disadvantage of this definition is that it does not address the problem of "apples and oranges." In other words, this definition does not facil- Review of the Strengths and Weaknesses of Community Financing 63 itate comparability across the schemes. Health-financing arrangements that meet the above definition can still differ significantly from each other in terms of objec- tives, structure, management, organization, and institutional characteristics. For example, community-level revolving drug funds in Honduras would qualify as CF, and so would the hospital-based prepayment­risk-sharing scheme of Bwamanda (DRC) and the individual savings account for pregnant women in Indonesia. Yet these various arrangements have different capacities to mobilize resources, to pro- vide financial protection, and to include the poor. We aimed to address this problem by grouping community-financing schemes. The possibilities for creating a typology are endless, and this is reflected in the reviewed papers. Four of the 45 papers reviewed propose a typology, and each pro- poses different ways of grouping CF. The common characteristics in the proposed typologies are that they combine the technical health-financing characteristics of the schemes with descriptors of the organizational structure that governs the oper- ation of the schemes. · Bennett, Creese, and Monasch (1998) separate the schemes based on the nature of the health risks they cover and ownership. The authors distinguish between high-cost, low-frequency events (type 1) and low-cost, high-frequency events (type 2). Additionally, schemes are also presented by ownership arrangements, distinguishing them by health facility, community, cooperative or mutual, nongovernmental organization (NGO), government, or joint ownership. · Atim (1998) reviews the experience of mutual health organizations in West and Central Africa and separates schemes based on their ownership (tradi- tional clan or social network, social movement or association, provider and community comanaged, community) and their geographic and socioprofes- sional criteria (rural, urban, and profession, enterprise, or trade union-based). · Criel, van der Stuyft, and van Lerberghe (1999) distinguish between two poles of voluntary health insurance systems: the mutualistic, or participatory, model, and the provider-driven, or technocratic, model. Their starting point is the risk cate- gorization offered by Bennett, Creese, and Monasch, and they arrive at these two typologies by adding three additional characteristics: size of target population, degree of overlap between the scheme and the existing providers, and intermedi- ary institutions between the source of funding and the destination of the funds. · Hsiao (2001) distinguishes among five types of CF initiatives: direct demand- side subsidies channeled to individuals (for example, Thai health card), coopera- tive health care, community-based third-party insurance, provider-sponsored insurance, and producer or consumer cooperative. The categorization takes into account not only whether community involvement is present but also the strength of community involvement. From our review of three dozen case studies, four commonly encountered and well identifiable modalities emerged that proved useful for our analytical purpose. The first modality groups schemes in which the resource mobilization instrument 64 Health Financing for Poor People: Resource Mobilization and Risk Sharing is out-of-pocket payments but the community is actively involved in fee design, collection, allocation, and management. We refer to this modality as "community cost-sharing" to distinguish it from cost-sharing arrangements in which the com- munity is not involved in any aspect of health financing. The second modality is the community-based prepayment scheme or mutual health organization. The third one is provider-based community health insurance. We label the final cate- gory community-based prepayment scheme linked to government or social insur- ance system. Table 2.4 summarizes these four types of schemes based on their core design features and management, organizational, and institutional characteristics. Community cost-sharing. In these types of arrangements, the community partic- ipates in mobilizing resources for health care through user fees. The health- financing instrument in this case is out-of-pocket payments, but the community is involved in setting user fee levels, allocating the collected resources, developing and managing exemption criteria, and general management and oversight. The community may also be involved in management of at least the first level of health care, the health centers, through participatory structures. The most impor- tant characteristics distinguishing this type of financing arrangement from the other three modalities is the lack of prepayment and risk sharing. The Bamako Initiative is a good illustration of this kind of health-financing mechanism. Community prepayment or mutual health organizations. These schemes are char- acterized by voluntary membership, prepayment of usually a one-time annual fee, and risk sharing. Some of these schemes cover catastrophic benefits (includ- ing hospital care and drug expenditures); others do not. The community is strongly involved in designing and managing the scheme. Schemes are typically not-for-profit. Examples include the Grameen Health Plan in Bangladesh and the Boboye District Scheme in Niger. Provider-based health insurance. These schemes often revolve around single provider units such as a town or city or regional hospital. They are characterized by voluntary membership, prepayment of usually a one-time annual fee, risk sharing, and coverage of catastrophic risks. They are frequently started by the providers themselves or through donor support. The involvement of the com- munity is often more supervisory than strategic. Examples include the Bwa- manda Hospital Insurance Scheme in the Democratic Republic of the Congo and the Nkoranza Community Health-financing Scheme in Ghana. Government or social insurance-supported community-driven scheme. These community-based health-financing schemes are attached to formal social insur- ance arrangements or government-run programs. The community actively par- ticipates in running the scheme, but the government (Thailand) or the social insurance system (Ecuador) contributes a significant amount of the financing. These schemes are not always voluntary (Burundi), and some have referred to this category as district or regional health insurance. Often such financing ini- tiatives are initiated by the government and not the community. One example of this type of scheme is Ecuador's Seguro Social Campesino. Table 2.5 presents the list of reviewed papers grouped by these modalities. The comparative studies review several types of schemes of varying modalities. TABLE 2.4 Often Encountered Forms of Community Financing Community-driven prepayment scheme Community Community Provider-based attached to social involvement prepayment community insurance or in user fee scheme or mutual health government-run collection health organization insurance system Example Bamako Grameen in Bwamanda in Seguro Social Initiative in Bangladesh Democratic Campesino, Benin and Mutual health Republic of Ecuador Guinea organizations in Congo Thai Health Card Thiès, Senegal Nkoranza in Scheme CMS in China Ghana Indonesia ASKES Technical Point-of-service Prepayment Prepayment Prepayment design payment Risk sharing Risk sharing Risk sharing characteristics No risk sharing Typically primary Hospital care Primary and Preventive care care; also some hospital care subsidized by drug and curative care sometimes hospital care Management Community Strongcommunity Community Community characteristics involvement in involvement in involvement is involvement in setting fees and management and more decisionmaking exemptions strategy informational schedules and Community not and supervisory allocation of necessarily than managerial collected defined in resources geographic sense but also by professional associations Organizational No formal Separated Integrated Durable characteristics organizational financing and financing and organizational form but provision provision structures informal links Varying degree Often poor Linkages with with health of linkages linkages with social security centers between primary care if and government schemes and not included entities providers ranging from third-party payment to durable institutionalized relationships Institutional Government Often started Often donor Very strong characteristics commitment to and supported initiated and government Bamako by donor and donor supported involvement Donor support government (financial, initiatives supervisory) 65 66 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 2.5 Summary of Case Studies by Modalities Modality 1: Community involvement in user fee collection 1. McPake, B., K. Hanson, and A. Mills. 1993. "Community Financing of Health Care in Africa: An Evaluation of the Bamako Initiative." 2. Ogunbekun, I., and others. 1996. "Costs and Financing of Improvements in the Quality of Maternal Health Services through the Bamako Initiative in Nigeria." 3. Soucat, A., T. Gandaho, and others. 1997. `'Health Seeking Behavior and Household Expenditures in Benin and Guinea: The Equity Implications of the Bamako Initiative." 4. Soucat A., D. Levy-Bruhl, and others. 1997. `'Local Cost Sharing in Bamako Initiative Systems in Benin and Guinea: Assuring the Financial Viability of Primary Health Care." 5. Gilson, L. 1997. "The Lessons of User Fee Experience in Africa." 6. Supakankunti, S. 1998. Comparative Analysis of Various Community Cost Sharing Implemented in Myanmar. 7. Fiedler, J.L., and R. Godoy. 1999. An Assessment of the Community Drug Funds of Honduras. 8. Fiedler, J. L., and J. B. Wight. 2000. "Financing Health Care at the Local Level: The Community Drug Funds of Honduras." 9. Gilson, L., and others. 2000. "The Equity Impacts of Community Financing Activities in Three African Countries." 10. Okumara, J., and T. Umena. 2001. "Impact of Bamako Type Revolving Drug Fund on Drug Use in Vietnam." Modality 2: Community prepayment or mutual health organizations 11. Arhin, D. C. 1995. Rural Health Insurance: A Viable Alternative to User Fees. 12. Toonen, J. 1995. Community Financing for Health Care: A Case Study from Bolivia. 13. Hsiao, W. C. 1995. "The Chinese Health Care System: Lessons for Other Nations." 14. Ron, A., and A. Kupferman. 1996. A Community Health Insurance Scheme in the Philippines: Extension of a Community-Based Integrated Project. 15. Liu, Y., and others. 1996. "Is Community Financing Necessary and Feasible for Rural China?" 16. Desmet, A., A. Q. Chowdhury, and K. Islam. 1999. "The Potential for Social Mobilization in Bangladesh: The Organization and Functioning of Two Health Insurance Schemes." 17. Ron, A. 1999. "NGOs in Community Health Insurance Schemes: Examples from Guatemala and Philippines." 18. Gumber, A., and V. Kulkarni. 2000. "Health Insurance for Informal Sector: Case Study of Gujarat." 19. Carrin, G., and others. 1999. `'The Reform of the Rural Cooperative Medical System in the People's Republic of China: Interim Experience in 14 Pilot Counties." 20. Chen, N., A. Ma, and Y. Guan. 2000. `'Study and Experience of a Risk-Based Cooperative Medical System in China: Experience in Weifang of Shandong Province." 21. Xing-yuan, G., and F. Xue-shan. 2000. "Study on Health Financing in Rural China." 22. Jütting, J. 2000. Do Mutual Health Insurance Schemes Improve the Access to Health Care? Preliminary Results from a Household Survey in Rural Senegal. 23. Schneider, P., F. Diop, and S. Bucyana. 2000. Development and Implementation of Prepayment Schemes in Rwanda. 24. Preker, A. S. 2001. "Philippines Mission Report." (continued) Review of the Strengths and Weaknesses of Community Financing 67 TABLE 2.5 Continued Modality 3: Provider-based community health insurance 25. Roenen, C., and B. Criel. 1997. The Kanage Community Financed Scheme: What Can Be Learned from the Failure? 26. Criel, B., P. van der Stuyft, and W. van Lerberghe. 1999. "The Bwamanda Hospital Insurance Scheme: Effective for Whom? A Study of Its Impact on Hospitalization Utilization Patterns." 27. Atim, C., and M. Sock. 2000. An External Evaluation of the Nkoranza Community Financing Health Insurance Scheme, Ghana. Modality 4: Community-driven prepayment scheme attached to social insurance or government-run system 28. Arhin, D. C. 1994. "The Health Card Insurance Scheme in Burundi: A Social Asset or a Non-Viable Venture?" 29. DeRoeck, D., and others. 1996. Rural Health Services at Seguridad Social Campesino Facilities: Analyses of Facility and Household Surveys. 30. Supakankunti, S. 1997. Future Prospects of Voluntary Health Insurance in Thailand. Studies that address multiple modalities 31. Dave, P. 1991. "Community and Self-Financing in Voluntary Health Programmes in India." 32. Diop, F., R. Bitran, and M. Makinen. 1994. Evaluation of the Impact of Pilot Tests for Cost Recovery on Primary Health Care in Niger. 33. Diop F., A. Yazbeck, and R. Bitran 1995. "The Impact of Alternative Cost Recovery Schemes on Access and Equity in Niger." 34. Atim, C. 1998. Contribution of Mutual Health Organizations to Financing, Delivery, and Access to Health Care: Synthesis of Research in Nine Western and Central-African Countries. 35. Bennett, S., A. Creese, and R. Monasch. 1998. Health Insurance Schemes for People Outside Formal Sector Employment. 36. Musau, S. 1999. Community-Based Health Insurance: Experiences and Lessons Learned from East and Southern Africa. 37. Atim, C. 1999. "Social Movements and Health Insurance: A Critical Evaluation of Voluntary, Nonprofit Insurance Schemes with Case Studies from Ghana and Cameroon." 38. CLAISS. 1999. "Synthesis of Micro-Insurance and Other Forms of Extending Social Protection in Health in Latin America and the Caribbean." 39. Narula, I. S., and others. 2000. Community Health Financing Mechanisms and Sustainability: A Comparative Analysis of 10 Asian Countries. 40. Hsiao, W. C. 2001. Unmet Needs of Two Billion: Is Community Financing a Solution? Conceptual papers that did not address any specific schemes classified under the modalities 41. Dror, D., and C. Jacquier. 1999. "Micro-Insurance: Extending Health Insurance to the Excluded." 42. Brown, W., and C. Churchill. 2000. Insurance Provision in Low-Income Communities: Part II--Initial Lessons from Micro-insurance Experiments for the Poor. 43. Ziemek, S., and J. Jütting. 2000. Mutual Insurance Schemes and Social Protection. 44. Criel, B. 2000. Local Health Insurance Systems in Developing Countries: A Policy Research Paper. 45. Ekman, B. 2001. Community-Based Health Insurance Schemes in Developing Countries: Theory and Empirical Experiences. 68 Health Financing for Poor People: Resource Mobilization and Risk Sharing PERFORMANCE OF COMMUNITY-BASED HEALTH FINANCING This section synthesizes the conclusions and evidence presented in the 45 reviewed studies regarding the performance of community-financing arrange- ments. Although there are several interesting performance aspects, this review focuses specifically on three questions. Question 1: What is the potential of community-based health-financing schemes as sustainable health care financing mechanisms? Which modality of com- munity financing performs better in terms of resource mobilization? Question 2: How inclusive of the poor are CF schemes? Which modality is more inclusive? Do the rich participate in pooling arrangements? Question 3: How effective are CF schemes in providing financial protection for their members? Which modality of community financing performs better to provide financial protection? Summary findings for resource mobilization capacity · Community-financing arrangements contribute significantly to the resources available for local health care systems, whether for primary care, drugs, or hospital care. · It appears that the involvement of the community--in various forms--results in access to more household resources than would otherwise be available for health care. · At the same time, there is large variation in the share of CF in the total resources of local health systems. · There continues to be a need for more rigorous evaluation of the resource gen- eration capacity of community-based schemes. Summary findings for social inclusion · Community-based health financing is effective in reaching a large number of low-income populations who would otherwise have no financial protection against the cost of illness. · The poorest and socially excluded groups are not automatically included in CF initiatives. · High-income groups are frequently underrepresented relative to the entire population, undermining redistribution of resources from the rich to the poor. Summary findings for financial protection · Generally, community-based health-financing schemes (modalities 2­4) are reported to reduce the out-of-pocket spending of their members while increasing their utilization of health care services. Review of the Strengths and Weaknesses of Community Financing 69 Resource Mobilization Capacity The most striking conclusion from the review of the literature is that there is lit- tle systematic evidence that would allow assessment of the overall resource generation capacity of various CF initiatives. It is also difficult, at this point, to assess how the various modalities fare when compared with each other as well as when com- pared with other health-financing instruments. None of the reviewed studies reported the amount of resources raised through community-financing arrange- ments as a share of the country's total health revenues. In a few cases, there are estimates about the per capita expenditures of the schemes. However, in the absence of concurrent estimates about the proportion of the population covered, extrapolation to a national level was not possible. This lack of evidence is not surprising. In most cases, community-financing arrangements are not registered. For example, 60 percent of 50 reviewed CF schemes in West and Central Africa were not registered with authorities (Atim 1998). Thus centrally maintained data do not exist. Surveys of a nationally rep- resentative nature do not ask questions based on which extrapolation would be statistically appropriate. In the absence of systematic assessments, the following findings aim to pro- vide a synthesis of authors' conclusions and an approximation of sustainability of resource mobilization through community financing. Community-financing arrangements contribute significantly to the resources avail- able for local health care systems, be it primary care, drugs, or hospital care. Twenty- six studies report the contribution of community-financing schemes to the operational revenues of local providers. A few examples are shown in box 2.2. The most striking finding is the large variation in the capacity of CF schemes to contribute to the operational expenditures of local providers. Some schemes achieve full financing of the recurrent costs of their local health center, even some drug and referral expenditures. Others, particularly hospital-based schemes (modality 3), have a modest contribution to the resources of the facility and external contributions are required. This large variation in the resource-generation ability holds not only across countries but also within countries. For example, Dave (1991) compares the experi- ence of 12 community-financing schemes in India. The Sewagram scheme, for example, was found to generate enough revenues through membership fees to cover 96 percent of all community-based health care costs, including salaries, drug costs, and mobile costs. On the other hand, the RAHA scheme covered only between 10 percent and 20 percent of the total community costs. The author attrib- utes the difference to the subscription policies: at Sewagram at least 75 percent of the households had to enroll in the scheme before services were reimbursed. This increased the risk-pooling and resource mobilization ability of the scheme. On the other hand, at RAHA, subscription occurred on an individual basis. It appears that the involvement of the community--in various forms--results in access to more household resources than would be otherwise available for health care. Most of the evidence in this regard originates from the analysis of the Bamako Initiative. BOX 2.2 Contribution of CF Schemes to Operational Revenues Community cost sharing/total local operating costs 100 80 60 40 20 0 t sau Mali Coas Benin Zaire Guinea Senegal Ivory Cameroon Guinea-Bis This graph is based on data from a study conducted by Soucat, Gandaho, Levy- Bruhl, and others on cost sharing in Benin and Guinea. The level of cost recov- ery from user fees in the health centers of those countries varies from 24 percent to 99 percent of the total local operating costs. This excludes the salaries generally paid by the government. Source: Soucat, Gandaho, Levy-Bruhl, and others (1997). Percent of recurrent costs from prepayment 60 50 40 30 20 10 0 GK SSSS Caja Abota Grameen Lalitpur (Bangladesh) (India) (Bolivia) (Guinea- (Bangladesh) (Nepal) Bissau) Based on data from Bennett, Creese, and Monasch, this graph shows the cost recov- ery from prepayment of six modality 2 schemes. The range is from 12 percent to 51 percent of recurrent expenditure. This shows that for these schemes the resources collected do not cover the full recurrent costs, thereby necessitating other sources of funding: out-of-pocket payments, donor, and government subsidy. Source: Bennett, Creese, and Monasch (1998). (continued) 70 Review of the Strengths and Weaknesses of Community Financing 71 BOX 2.2 Continued Percent of health facility operation costs financed by scheme 8 IV 7 III 6 5 4 3 III 2 1 0 Chogoria (Kenya) Kisiizi (Uganda) CHF (Tanzania) Based on data from Musau, the graph shows the cost-recovery level of three schemes: Chogoria, Kisiizi (modality 3 schemes), and Community Health Fund of Tanzania (modality 4). The contribution of the schemes to the financing of the health facility ranges from 2.1 percent to 7.2 percent. Compared to Chogo- ria, the higher resource mobilization capacity of Kisiizi is attributed to lower premiums, which attracts more members. The CHF scheme is highly subsi- dized by the government from funds provided under a World Bank project. Source: Musau (1999). The Bamako Initiative is categorized in our typology as modality 1: the community is involved in setting the level of user fees, designing and managing exemption schemes, and allocating the collected funds. (See box 2.3 for a more detailed description of the Bamako Initiative.) The financing instrument still lacks risk-pool- ing and prepayment elements a priori. However, the involvement of the commu- nity appears to alleviate the collection difficulties providers have experienced with user-fee mechanisms and the regressivity associated with out-of-pocket payments. In particular, community involvement can lead to better allocation of collected resources to services and drugs that community members value and want (McPake, Hanson, and Mills 1993; Diop, Yazbeck, and Bitran 1995; Soucat 1997). A few highlights: · Soucat, Gandaho, and others (1997) analyzed the impact of the Bamako Initia- tive in Benin and Guinea. They showed that direct household expenditure (through user fees) contributed to 25 percent of the health centers' local oper- ating costs in Benin and 40 percent in Guinea. The revenue was used to cover drug costs, outreach, local maintenance, and replacing supplies, and preventive care was subsidized more than curative care, thereby promoting utilization. In another study Soucat, Levy-Bruhl, and others (1997) found that in Benin, about 72 Health Financing for Poor People: Resource Mobilization and Risk Sharing BOX 2.3 THE BAMAKO INITIATIVE The Bamako Initiative (BI) was launched in 1987 by a group of African minis- ters of health in Bamako, Mali, in a meeting sponsored by WHO and UNICEF. The BI was a response to the rapid deterioration of access experienced in sev- eral health systems during the 1980s. Deterioration in access was partially attributed to the imposition of user fees on public facilities. In contrast, the BI model emphasizes that revenue should be raised and controlled at the local level through community-based activities that are national in scope. Commu- nity participation is seen as a mechanism to build accountability to the users of health care in that the revenues are used in ways that address the persistent quality weaknesses of primary care (Gilson 1997). By late 1994, the BI had been implemented in 33 countries, of which 28 were in sub-Saharan Africa. The other 5 were in Cambodia, Myanmar, Peru, Vietnam, and Yemen. half of the local operating costs was covered by the government, 23 percent by donors, and 28 percent by a surplus generated by community financing. In Guinea, 44 percent of the operating costs were covered by the government, 26 percent by donors, and 30 percent by community financing, with a lower aver- age surplus than in Benin. · Pilot studies conducted in Niger compared three resource mobilization methods: newly introduced user fees, prepayment scheme plus user fees, and the control dis- trict where health services remained "free" (Diop, Yazbeck, and Bitran 1995). Rev- enues from fees were managed by local providers and by local health committees organized by the population. Revenues were pooled at the district level and used mainly to purchase drugs. The study found that both intervention districts showed substantial increases in revenue collection compared with the control district. It also revealed that the revenue generation per capita under the prepayment method was two times higher than under the user-fee method. The authors add that the sustain- ability of these financing mechanisms critically depends on cost-containment. There is little evidence from the analysis of other modalities to determine how well modalities 2, 3, and 4 fare relative to each other in mobilizing resources for health care. While some of the literature is enthusiastic about the contribution of CF to health care resources, others are less optimistic about the sustainable resource- generation capacity of these arrangements. In the studies in which the schemes' contribution was low in the total financing of providers, authors tended to be pessimistic about sustainable resource generation through CF arrangements. (Ben- nett, Creese, and Monasch 1998; Atim 1998; Musau 1999; Arhin 1994; Roenen and Criel 1997; CLAISS 1999). The key factor that undermines the revenue-raising potential of community financing is their predominantly poor contributing popu- lation. Whether in rural or in urban areas, community-based health-financing schemes reach the poorest half of the population. If most members of community- Review of the Strengths and Weaknesses of Community Financing 73 financing schemes are poor, redistribution within the community takes place within a much limited overall resource pool (Hsiao 2001; Jütting 2000; Atim 1998; Bennett, Creese, and Monasch 1998). · Bennett, Creese, and Monasch (1998) recognize that prepaid premiums are important resource-generating instruments, but the authors conclude that "there is little evidence that voluntary prepayment schemes for those outside the formal sector can be `self-financing' for anything other than the short term." They show that for most schemes the resources collected from the combination of prepayment and user fees does not cover the recurrent costs of the scheme, and thus external funding is required. See table 2.6 for their findings. · In six Central and West African countries, Atim (1998) concludes that mutual health organizations (MHOs) "have had little impact on the finances of health facilities." For instance, in the Thiès region of Senegal, 30 percent of the admis- sions into St. Jean de Die are MHO members, and yet MHO resources account for less than 2.5 percent of hospital revenues. Atim notes, however, that these findings are not surprising, given the recent growth in the MHO phenomenon in the region. He concludes that the potential of MHOs to mobilize resources is much greater than current figures on contribution revenues would suggest. · Similar experiences are reported from five schemes in East Africa. Two hospital- based prepayment schemes in Uganda and Kenya contributed 8 percent and 2 per- cent, respectively, to the operational expenditures of their hospitals. Similar results were observed in Tanzania with regard to dispensaries: 5.4 percent of the partici- pating dispensaries' income came from the prepayment scheme (Musau 1999). · A study of the National Health Card Insurance Scheme (CAM) in Burundi revealed that the revenue generated from "premiums was insufficient to fund even the recurrent costs of outpatient drugs consumed by participating house- holds" (Arhin 1994). · In the review of Latin American community-financing schemes, 9 out of 10 schemes were found to need large external contributions to ensure future sus- tainability (CLAISS 1999). · Roenen and Criel (1997) reported that the sum of the premiums generated in the Kanage Community Financed Scheme covered only a fraction of what members spent on care. The scheme was largely financed by the revenues of the Murunda hospital to which the scheme was affiliated. Social Inclusion In this section, we explore whether CF schemes are effective in reaching the poor and socially excluded groups. To address this question, we looked at the socioeconomic composition of CF scheme membership. In particular, we were interested in whether CF arrangements reach the poor and whether higher income groups participate in pooling and income redistribution arrangements. TABLE 2.6 Cost Recovery from Prepaid Premiums Cost recovery from prepayment No. Scheme Country Last date available 42 SWRC India 10 percent of recurrent expenditure 25 RAHA India 10­20 percent off community costs & 100 percent referral costsa 24 SSSS India 15 percent of recurrent expenditureb 18 Caja-Tiwanaku Bolivia 18 percent of recurrent expenditurec 31 Abota Guinea Bissau 23 percent of recurrent expenditured 64 Bajada Philippines 30 percent of recurrent expenditure 58 CAM Burundi 34 percent of outpatient drug costse 17 GK Bangladesh 12 percent of recurrent expendituref 14 Grameen Bangladesh 24.7 percent of recurrent expenditureg 41 BRAC Bangladesh 50 percent of recurrent expenditure 62 Health Card Thailand 50 percenth 67 Bwamanda former Zaire 65­70 percent recurrent, excluding personal allowances 79 SWHI Thailand 50­60 percenti 59 Lalitpur Nepal 51 percent of recurrent expenditurej 21 Kisiizi Uganda 72 percent of recurrent expenditurek 46 KSSS India 88 percent of recurrent expenditure 60 Boboye Niger 89 percent of drug and management costsl 26 Sewagram India 96 percent of community health program costs 32 Medicare II Philippines 100 percent of recurrent expenditurem 33 PHACOM Madagascar 100 percent of drug costsn 61 UMASIDA Tanzania 100 percent of costs 2 ORT Philippines 100 percent excluding professional salaries 66 Nsalasani Congo-Brazzaville 100 percent 29 Bao Hiem Y Te Vietnam 130 percento Source: Bennett, Creese, and Monasch (1998, p. 40) a. Nonmember fee collections cover roughly 60 percent of community cost. b. Copayments cover 31 percent of costs and balance is financed from fund-raising activities. c. Without the costs associated with expatriate assistance the caja contribution would have been 48 percent of budget. d. In a study of 18 village schemes the cost recovery ranges from 3 percent to 123 percent based on the assumption that all communities consume a given amount of drugs estimated by government. e. There is no link between prepayment revenues collected and financing of services as revenues revert to government. A study in Muyinga Province showed that the revenue from premiums was sufficient to fund approximately 34 percent of drug costs. f. The remaining was covered by user fees (24 percent), subsidies from GK commercial ventures (14 percent), and international solidarity (50 percent). g. The remaining was covered by user fees (41.3 percent members and nonmembers) and a long-term loan from Grameen Bank (34 percent). h. The scheme is currently half-financed by government budget and half by cardholders; this is a relatively recent reform, and previous estimates show recovery of approximately 35 percent of recurrent costs. i. Balance from cross-subsidy from richer households. j. Cost recovery from prepayment ranges from 30 percent to 56.6 percent. k. Average cost recovery for the hospital as a whole is 48 percent. l. 149 percent of drug costs only. m. Fund utilization is relatively low, ranging from 38 to 78 percent of total collections. Only in 1992 after a large drop of membership disbursement exceeded collection in Unisan, Quezon, pilot scheme. n. Drugs are bought with membership fees but often only last three months of the year. o. The 130 percent includes a cross-subsidy from formal sector workers to informal sector workers. 74 Review of the Strengths and Weaknesses of Community Financing 75 Of the 13 studies that report evidence regarding the socioeconomic composi- tion of CF members, the findings appear to be consistent. Community-based health-financing schemes extend coverage to a large number of people who would oth- erwise not have financial protection. However, there seems to be some doubt whether the poorest are included in the benefits of community-based health financing. Where data are available, the most frequently cited reason for not being included in a community-financing scheme is lack of affordability. Distance to scheme hospi- tal is also reported as affecting the decision to enroll in the scheme in several cases. These findings do not show systematic variation with the modality of the reviewed scheme. Table 2.7 summarizes these findings. TABLE 2.7 Summary of Findings (1): Who Is Covered by CF Arrangements? Poorest of Ability to pay Distance Scheme the poor is the main Rich gradient reaches are not reason for not do not to scheme the poor covered being covered participate provider Modality 1 Diop, Yazbeck, and Bitran (1995), Niger McPake, Hanson, and Mills (1993), Burundi Gilson and others (2000), Benin, Kenya, Zambia Modality 2 Desmet, Chowdhury, and Islam (1999), Bangladesh Jütting (2000), Senegal Diop, Yazbeck, and Bitran (1995), Niger Liu and others (1996), China Carrin and others (1999), China Modality 3 Criel, van der Stuyft, and van Lerberghe (1999), Dem. Rep. of Congo Atim and Sock (2000), Ghana Modality 4 Arhin (1994), Burundi DeRoeck and others (1996), Ecuador Supakankunti (1997), Thailand Total number of studies confirming selected finding 13 5 6 3 3 76 Health Financing for Poor People: Resource Mobilization and Risk Sharing Findings for modality 2 schemes: Community health fund or mutual health organization: · The Gonosasthya Kendra (GK) in Bangladesh is effective in reaching the poor. Of the households classified as destitute in the area, 80 percent are covered by the scheme--46 percent of the poor, 20 percent of the middle class, and 10 per- cent of the rich, amounting to an overall subscription rate of 27.5 percent. The reason among the destitute and the poor for not subscribing, even after 15 years of operation of the scheme, is the level of the premium and copayments associated with it. The distribution of the scheme's membership by income group is as follows: 33.5 percent are classified as destitute and poor, 57.5 percent as middle income, and 9 percent as rich. In terms of equity, this suggests that pooling and income redistribution does take place but predominantly between the middle income and the poor (Desmet, Chowdhury, and Islam 1999). · Subscription rates to the scheme demonstrate a distance gradient to the GK hospital: subscription rates between the two lowest socioeconomic groups were 90 percent for the villages near the hospital and 35 percent for the dis- tant villages. Lack of transportation to the GK hospital was the second most- cited reason among the destitute and the poor for not subscribing to the scheme (Desmet, Chowdhury, and Islam 1999). · The Grameen Bank (GB) health scheme is operated by the microfinance orga- nization in Bangladesh. The GB scheme covered 57.8 percent of the poor in the areas while only 1.8 percent of the nonpoor families signed up for the scheme. This suggests that the scheme effectively enlisted the membership of the local poor. At the same time, solidarity and income redistribution is undermined as the rich do not take part in the pooling arrangement (Desmet, Chowdhury, and Islam 1999). · In Rwanda, the pilot prepayment scheme increased the utilization rate of members as compared with nonmembers despite a copayment charge. · Consultation rates of nonmembers = 0.2 per capita in all five districts. · Consultation rates of members = 1.3 per member in Byumba, 1.87 in Kab- gayi, and 1.76 in Kabutare. · Comparing utilization rates pre- and post-intervention, members' consul- tation rates were three to six times higher than reported before implemen- tation of the prepayment scheme. · Similar findings are reported for Senegal's Thiès district. Analyzing the member- ship characteristics of four mutual health organizations, Jütting (2000) reports that the average income of members is three times that of nonmembers. He con- cludes that the poorer people do not participate in MHOs because they do not have the financial resources to pay the regular premium. At the same time, Jüt- ting suggests that this finding does not mean that MHOs increase inequality for the population. Based on the national poverty line, most of the scheme mem- Review of the Strengths and Weaknesses of Community Financing 77 bers qualify as poor. Thus on average it can be concluded that these mutual health organizations have helped poor rural populations cope with health risks, even though they have not been able to include the very poorest. Findings for modality 3 schemes: Provider-based community health insurance: · In the Bwamanda hospital prepayment scheme in the Democratic Republic of the Congo, the very low- (less than US$20/month) and high-income (greater than US$200/month) groups were less represented among scheme members than in the nonmember population. Of the member population, 14.9 percent was from very low-income HHs versus 18.7 percent among nonmembers; 5.9 percent of the member population was from high-income HHs versus 10.5 percent among nonmembers (Criel, van der Stuyft, and van Lerberghe 1999). · In the Nkoranza scheme in Ghana, being "hard core poor" (defined as those falling below one-third of average income) is one of the reasons for not join- ing prepayment schemes. Eight percent of the Nkoranza town and 17 percent of the whole district are identified as hard core poor. The reason most often stated for not being a part of the scheme was "financial" and the fact that the registration period coincided with a low financial situation. One focus group also cited distance from the hospital as a reason for being uninsured (Atim and Sock 2000). Findings for modality 4 schemes: Community financing supported by govern- ment, social insurance, or both: · Under the Thai health card scheme in Khon Kaen province, health cardhold- ers have significantly lower income than those without a health card. This suggests a pro-poor targeting of the health card program. Separating card- holders into new cardholders, renewed cardholders, and dropouts, Supakankunti (1997) reports that dropouts have the lowest income, suggest- ing that the health card scheme may pose affordability problems to the lowest income population groups in addition to adverse selection due to lower levels of reported chronic illness in this group. · Comparing revolving drugs funds and the prepaid health card scheme in Burundi, 25 percent of the households were reported to be part of the prepay- ment scheme. Socioeconomic status and membership in the prepayment scheme were positively correlated. The poor were more likely to pay through user charges than purchase a prepayment plan. The low subscription rate in the prepayment scheme was associated with difficulty in coming up with one- time large payments in cash-constrained situations and the poor quality of services at government facilities. These factors limited ability, as well as will- ingness to pay (McPake, Hanson, and Mills 1993). Arhin's findings corroborate that the primary reason for not purchasing a prepayment plan was financial affordability. She reports that in Burundi 27 percent of survey respondents did not purchase the health card because they could not afford it (Arhin 1994). 78 Health Financing for Poor People: Resource Mobilization and Risk Sharing Financial Protection In this section, we explore whether CF schemes are effective in providing protec- tion from the impoverishing effects of catastrophic health care events. At this point, only imperfect measures are available to approximate this question. Specif- ically, we looked for the following indicators to assess whether community- financing schemes reduce the financial burden of seeking care. What is the level of out-of-pocket payments of members relative to nonmembers? What is the uti- lization of CF scheme members relative to nonmembers? Analyzing utilization and out-of-pocket expenditure patterns together allows us to take into account forgone use due to the high cost of seeking health care. Assessing financial protection based only on point-of-service spending informa- tion does not allow delayed or forgone care due to high costs to be factored in. Twenty studies present evidence regarding the financial protection impact of the CF schemes they reviewed. In 13 studies, scheme members are more likely to use health care services than nonmembers, and 2 report no difference between members and nonmembers. One study compared user fees with prepayment schemes (Diop, Yazbeck, and Bitran 1995) and found a slight decrease in the use of health care for the user-fee modality compared with the prepayment scheme of CF. In 9 of these, mem- bers pay less out-of-pocket. These findings do not appear to systematically vary with the modality of the scheme. We summarize these findings in table 2.8. Findings for modality 1 schemes: · Soucat and others (1997) have reported the increased utilization of health ser- vices after the introduction of the Bamako Initiative in Benin and Guinea. The authors attribute this development to the availability of drugs and improved quality of services brought about by community involvement. The poor in Guinea had fewer alternative sources of care compared with what they could find in the illegal drug market in Benin, which led Guinea's poor to opt out of seeking care. This study emphasized that improvements in qual- ity, access to care, availability of drugs, and community involvement play an important role in increasing utilization of schemes that rely on user fees as the predominant health-financing mechanisms. Findings for modality 2 schemes: · Pilot studies conducted in Niger compared three resource-mobilization meth- ods: newly introduced user fees, prepayment scheme plus user fees, and the control district where health services remained "free" (Diop, Yazbeck, and Bitran 1995). Revenues from fees were managed by local providers and by local health committees organized by the population. Revenues were pooled at the district level and used mainly to purchase drugs. In both intervention districts, quality improvements and availability of drugs stimulated use of health care while utilization continued to decline in the control district. The authors con- clude that the "positive effects of the quality improvements cancelled out the negative effects of the introduction of use fees." A few details: Review of the Strengths and Weaknesses of Community Financing 79 TABLE 2.8 Summary of Findings (2): Does CF Reduce the Burden of Seeking Health Care? Utilization of Level of OOPs for members relative members relative to nonmembers to nonmembers Modality 1 Soucat and others (1997), Benin and Guinea Higher Diop, Yazbeck, and Bitran (1995), Niger Lower Lower McPake, Hanson, and Mills (1993), Burundi Higher Gilson and others (2000), Benin, Kenya, Zambia Higher Modality 2 Desmet, Chowdhury, and Islam (1999), Bangladesh Same Gumber and Kulkarni (2000), India Lower Arhin (1995), Ghana Higher Diop, Yazbeck, and Bitran (1995), Niger Higher Lower Schneider, Diop, and Bucyana (2000), Rwanda Higher Jütting (2000), Senegal Same Liu and others (1996), China Lower Carrin and others (1996), China Lower Xing-yuan and Xue-shan (2000), China Higher Lower Chen, Ma, and Guan (2000), China Lower Modality 3 Criel, van der Stuyft, and van Lerberghe (1999), Dem. Rep. of Congo Higher Roenen and Criel (1997), Rwanda Higher Atim (1999), Ghana Higher Modality 4 Arhin (1994), Burundi Higher Lower DeRoeck and others (1996), Ecuador Higher Lower Musau (1999), Tanzania Higher Total number of studies confirming selected finding 16 9 Note: OOPs = out-of-pocket payments. · People using improved services in the fee-for-service district saved 40 per- cent of the amount they spent on health care for an episode of illness before the intervention. · In the prepayment district, out-of-pocket health spending declined by 48 percent, and total health spending (including the tax component) declined by 36 percent. · The number of initial visits to the health care facility increased by about 40 percent in the prepayment district. Utilization among the poorest quartile doubled. Utilization decreased slightly in the fee-for-service district. · Even for short travel distances, utilization in the fee-for-service district decreased from 45 percent to 37 percent and increased from 36 percent to 43 percent in the prepayment district. 80 Health Financing for Poor People: Resource Mobilization and Risk Sharing · The SEWA Scheme in India improves financial protection for its members. Among the rural population, the total cost of seeking care for SEWA members was significantly less than for ESIS members and the uninsured (Rs 295 versus Rs 380 and Rs 401, respectively, for acute morbidity; Rs 451 versus Rs 644 and Rs 697, respectively, for chronic morbidity). However, the burden of seeking care on the household budget continued to be higher among SEWA members than among those insured by other mechanisms (Gumber and Kulkarni 2000). · Arhin (1995), in assessing the viability of rural health insurance as an alterna- tive to user fees, also found that the scheme in Ghana removed a barrier to admission and led to earlier reporting of patients and increased utilization among the insured. · There is no convincing evidence that the two reviewed Bangladesh schemes fare strongly in terms of improving access to hospital care for the poor. Desmet, Chowdhury, and Islam (1999) report that the use of hospital services among members shows a significant income gradient. Hospital admissions per 100 persons per year amount to 2 for the destitute, 2.3 for the poor, 3.72 for the middle income, and 10.7 for the rich. Whether this is due to overuse by the higher income groups or underuse by the lower income groups needs to be tested. The Grameen Bank scheme does not include hospital care, and the lack of coverage for hospital care is the most frequently raised complaint in the implementation committee. · Jütting (2000) finds no significant difference among the contact rates between members of three Senegal schemes and nonmembers. An interesting finding in one of the schemes is the low contact rate, which Jütting attributes to the availability of malaria medication. · In China, various attempts to revive the Cooperative Medical System (CMS) are described in detail in Hsiao (2001). A number of studies assess the success of these experiences in terms of reducing out-of-pocket payments and increas- ing utilization by its members. A few examples are provided below: · In Shandong Province, a study was conducted to determine the level of disease- induced poverty. It was measured by calculating average medical expendi- tures for those diseases classified as contributing to a high economic burden based on income level and disease type. Disease-induced poverty was found to have decreased from 23 percent to 3.7 percent in Shougang County, and from 30 percent to 3 percent in Pingdu County after the intro- duction of CMS coverage (Chen, Ma, and Guan 2000). · Carrin and others (1999) assessed "ratios of insurance protection" in China's Rural Cooperative Medical System (RCMS). The authors measured the ratio of average health insurance contribution (destined for reimburse- ment of health care costs) per capita to average health care expenditure per capita. They found wide variation in the level of insurance protection across counties, from less than 10 percent in Lingwu and Xiaoshan Coun- Review of the Strengths and Weaknesses of Community Financing 81 ties to more than 30 percent in Yihuang County. However, the authors also observed that "on average" health insurance contributions are not enough to offer RCMS members a reasonable health insurance benefit because out- of-pocket expenditures are still associated with seeking care. · Another study based on household data compared out-of-pocket expenses and utilization by members and nonmembers of the Cooperative Health Care Scheme (CHCS) pilot study. The average fee per outpatient visit was 10.1 yuan for CHCS members compared with 21.7 yuan for nonmembers. At the same time, higher utilization of medical care among CHCS members was observed in the two pilot sites. Hospital admission rates were 60 percent among mem- bers, compared with 43 percent in the control group. In the township of Wuzhan, 17.3 percent of CHCS members used outpatient services compared with 7.4 percent in the control group (Xing-yuan and Xue-shan 2000). · Liu and others (1996) compared households covered by community-financing schemes and the uninsured in China's poverty regions. They show that the cost per visit is twice as high for the uninsured as for the insured (3 yuan per visit for the uninsured as compared with 1.5 yuan for a member of a house- hold insured under community-financing schemes). They also find that the higher average charge per outpatient visit for the uninsured can be attrib- uted to the fact that these "schemes can exercise their bargaining power in demanding discounted prices or the providers can be paid on a partial-capi- tation basis" (Liu and others 1996; Hsiao 2001). Findings for modality 3 schemes: · Criel, van der Stuyft, and van Lerberghe (1999) looked at the utilization of hospital services associated with the Bwamanda hospital insurance scheme in the Democratic Republic of Congo. They found that hospital utilization was significantly higher among the insured population than among the unin- sured. The innovative aspect of their study was assessing whether the addi- tional utilization was justified or the result of insurance coverage in terms of moral hazard and induced demand. They concluded that the impact of insur- ance increased access to justified care in the case of caesarian sections and her- nias. Thus the Bwamanda scheme succeeded in increasing utilization of high-priority hospital care services (strangulated hernias and C-sections). · A distance gradient was observed in both insured and uninsured populations, suggesting that insurance can overcome the financial barriers to use but not necessarily the geographic barriers. The indirect costs of travel and hospital- ization time in rural areas may outweigh the direct costs of hospitalization. When looking at specific high-priority interventions (strangulated hernias and C-sections), the distance gradient is reduced, suggesting that the insur- ance scheme improved equity in the district. The same impact is not observed for nonurgent care. This suggests that the impact of geographic barriers was more successfully compensated for in the case of high-priority service use than in the case of low-priority service use. In addition, these findings suggest 82 Health Financing for Poor People: Resource Mobilization and Risk Sharing inelastic demand for high-priority services as well as effective resource alloca- tion practices (Criel, van der Stuyft, and van Lerberghe 1999). · In 1993, three-fourths of the consultants at the hospital-clinic level and half of the hospitalized patients were members of the Kanage cooperative scheme in Rwanda. Members used the hospital services 8.5 times as often as non- members. Although utilization of the services was high among the members, there was a lack of equality, which could have contributed to the failure of the system. The size of the premium was independent of income or distance from the hospital. It was not an integrated system and lacked quality care and ser- vices, which led to the failure of the scheme (Roenen and Criel 1997). Findings for modality 4 schemes: · The Health Card Scheme (CAM) in Burundi studied by Arhin illustrates that, in the month preceding the study, 27.9 percent of the households that held valid CAM cards had incurred out-of-pocket expenses for medical consulta- tions, drug purchases, or both, while the corresponding figure for those households without valid cards was 39.9 percent. The mean expenditure per treatment was also lower for scheme members (Arhin 1994). · The formal treatment rate (modern, or western, care sought outside the home) was more than 50 percent higher for the CAM group than for the non-CAM members. This high rate for the CAM group may be explained by the fact that some government health centers gave incomplete treatments, delaying recovery and requiring visits to collect the remaining drugs. It is also possible that this high utilization rate among CAM households was the outcome of "supplier- induced demand," that is, an increase in the demand and consumption of health care by patients as a result of the providers' actions. In addition, house- holds participating in the CAM scheme were three times more likely to use the government facilities than non-CAM households (Arhin 1994). · The Ecuador Seguridad Social Campesino (SSC) rural health facility significantly increased financial protection for its members: out-of-pocket expenditure for health care of SSC members was only one-third of that for nonmembers. Mem- bers of the SSC rural health facility were more likely to seek care for illness than nonmembers (80 percent versus 66 percent). Demand analysis conducted demonstrated that improving the quality of care and increasing the referral rates and availability of drugs would increase the utilization rates of SSC health ser- vices. The analysis also showed that there was no significant relationship between household income and distance and travel time to reach the health facility and the decision to seek care outside the home. Lower income house- holds were more likely to belong to the scheme (DeRoeck and others 1996). Discussion of Performance Results Our review found that community-financing arrangements--regardless of the modality--contributed a significant amount to the resources available to local Review of the Strengths and Weaknesses of Community Financing 83 health systems. At the same time, there was a large variation in terms of the share of community financing in the local health revenues. It is also apparent that community-financing arrangements alone can rarely support hospital-level care in full, and thus other mechanisms of health financing are frequently used in conjunction with such arrangements. It has to be noted, however, that the evidence base regarding the resource- generation capacity of community-financing schemes requires further strengthen- ing. Currently, the total amount of resources mobilized through community-based health financing is anyone's guess. What is the share of health financing through CF arrangements in terms of total health care financing? How many resources are mobilized through CF arrangements relative to general taxation, social insurance, private insurance, and out-of-pocket payments? In the absence of more comprehensive information and improved method- ologies, it is difficult to assess the global impact and the potential of community financing as resource mobilization instruments to finance health care for the poor. Having a systematic assessment of the volume of resources raised through community financing would allow exploration of the following issues: · Comparability with other sources of health financing would allow assessment of effectiveness, efficiency, and equity of community financing as health- financing mechanisms. · Assessment of the impact of community financing on the amount of govern- ment funding. A critical question is whether community financing comple- ments or displaces government funding. Does the existence of CF make governments allocate fewer resources to a region with a lot of CF initiative, or is the reverse true? · Assessment of sustainability of CF arrangements over the long run. More evidence can be found about CF's impact on social inclusion and financial protection. The targeting outcome of community-financing schemes is impressive although there are indications that the very poorest are not automatically included. In terms of financial protection, CF reduces financial barriers to access through increased utilization by members, as opposed to nonmembers, and reduced out-of- pocket spending. There were no studies suggesting an inverse relationship, and two studies found no difference between use by members and nonmembers.1 However, a number of methodological concerns warrant some caution in interpreting the reported results. The most important one is that selection into membership status is nonrandom. People with higher risk for illness and higher propensity to use health care will be more likely to purchase insurance. Thus the impact of membership on utilization and out-of-pocket payments cannot be validly discerned by looking at sample averages of members and nonmembers. Such measures are biased and magnify the impact of community financing on utilization because these individuals would also have had higher use in the absence of membership. 84 Health Financing for Poor People: Resource Mobilization and Risk Sharing DETERMINANTS OF SUCCESSFUL RESOURCE MOBILIZATION, SOCIAL INCLUSION, AND FINANCIAL PROTECTION The key determinants that contribute to successful resource mobilization, com- bating social exclusion, and financial protection include: (a) ability to address adverse selection, accommodate irregular revenue stream of membership, pre- vent fraud, and have arrangements for the poorest; (b) good management with strong community involvement; (c) organizational linkages between the scheme and providers; and (d) donor support and government funding. Table 2.9 sum- marizes successful and unsuccessful design features. TABLE 2.9 Determinants Associated with Effective Revenue Collection and Financial Protection Design features Supporting effective Undermining revenue revenue collection and collection and financial protection financial protection Technical design characteristics Addressing adverse selection Noncompliance, evasion of through group membership membership payments Accommodating irregular income Adverse selection stream of members (allow in-kind Lack of cash income contributions, flexible revenue No cash income at collection time collection periods) Sliding fee scales and exemptions for the poor make schemes more affordable Management characteristics Community involvement in Provider capture--high salary of management can exert social providers at the expense of pressure on member compliance service-quality improvement with revenue collection rules Weak supervision structures Extent of capacity building increase the chance of fraud with Information support membership card Poor control over providers and members contributes to moral hazard and cost escalation, and undermines sustainability of the scheme Organizational characteristics Linkages with providers to Fragmentation between inpatient negotiate preferential rates and outpatient care leads to raises attractiveness of schemes inefficiency and waste ultimately and contributes to successful resulting in loss of membership membership Institutional characteristics Government and donor support make the schemes more sustainable and pro-poor. Review of the Strengths and Weaknesses of Community Financing 85 Technical Design Characteristics Revenue Collection Most community-financing schemes rely on a combination of revenue sources including prepayment, user charges, government subsidy, and donor assistance. For example, in the Kanage Cooperative Scheme in Rwanda, the sum of the premiums collected covered only a fraction of what members spent on care, and the scheme was largely financed by the revenues of the hospital. The share of each these sources varied widely by scheme (Roenen and Criel 1997). Despite large variations in the composition of revenue sources, it appears that schemes can rarely raise enough resources from prepayment only. As a result, user charges are often utilized in conjunction with other resources, and most schemes rely on some form of external financing (government subsidy, donor support). For example, the CLAISS study (1999) compared 11 schemes in Latin America and found that 9 received significant external financial support. Most prepayment schemes collect membership fees on the basis of annual premium rates, which are typically a flat rate (community rated). Annual collec- tion is consistent with the agrarian-based structure of income-generating capac- ities (Diop, Yazbeck, and Bitran 1995; Roenen and Criel 1997; Atim and Sock 2000; Bennett, Creese, and Monasch 1998; Arhin 1994). Sliding scales and exemptions for the poor are rare and reported more frequently from Asia than Africa (Desmet, Chowdhury, and Islam 1999; Dave 1991). Flat rate contributions simplify the collection procedure and are less subject to manipulation. At the same time, such contribution schemes may prevent the poorest from joining. Revenue collection appears to be more successful when the contribution scheme takes into account the nature of the membership population's revenues. Synchronizing the contribution collection period with cash-earning periods makes a difference in terms of the ability of the schemes to raise resources. Some examples include: · In Central and West Africa, 73 percent of the 22 MHOs reviewed had already designed their contribution scheme to coincide with a more cash-endowed period (Atim 1998). · The Kanage Cooperative Scheme in Rwanda scheduled registration in the coffee-harvesting period between June and September. Roenen and Criel (1997) suggested that this may have been too short a time period and, along with low membership levels, may have contributed to the scheme's failure. · The Bwamanda Hospital Insurance Scheme also has a community-rated sys- tem of premium collection during the crop-selling season. The scheme offers voluntary membership to the family as a subscription unit (Criel, van der Stuyft, and van Lerberghe 1999). · The Nkoranza community-financing scheme in Ghana was found to have a low coverage rate of 30 percent. The registration period did not coincide with the cash-earning period of the community. This was one of the main reasons behind the low enrollment (Atim 2000). 86 Health Financing for Poor People: Resource Mobilization and Risk Sharing BOX 2.4 TURNING POTATOES AND LABOR INTO CASH REVENUES IN BOLIVIA Toonen argues that allowing farmers to provide in-kind contributions either in terms of farm products or labor increases the ability of the scheme to attract members and thus resources. He presents the example of a rural prepayment plan from Bolivia in which membership dues were in the form of contributing seed potatoes to the community organization. In addition, at least one family member had to work on the community lot for production of potatoes. Some of the harvest was kept for use as seed potatoes for the following year, and the remainder was sold on the market. Proceeds were used to pay for drugs, to pay a bonus to the auxiliary nurse, and to renovate the health centers. Source: Toonen (1995). In-kind contributions are rarely allowed (Atim 1998; Bennett, Creese, and Monasch 1998; Musau 1999; Dave 1991). There are a few exceptions, and in all cases the authors appear enthusiastic about the potential of generating resources from in-kind contributions. For example, Preker (2001) found that in the Philip- pines' Pesos for Health community scheme, when people fell ill and had to visit the hospital, or for members who could not pay the premiums, they could mon- etarize agricultural produce such as chickens into cash in the hospital and were able to pay for health care. See also box 2.4. Dave (1991) also found schemes in India where membership payments were accepted in the form of rice and sorghum. In-kind contributions, however, are accepted as payment for prepayment insurance scheme membership and not as an on-going payment option--RAHA scheme (rice), Sewagram (sorghum), Goal- para (community labor)--to ensure that the poor are not excluded. Schemes such as Sewagram employ a community health worker (CHW) to collect the contribu- tions once a year, usually at harvest time; the CHW then sells the grain in the open market. From the funds generated the CHW purchases drugs, pays Sewa- gram for mobile support, and retains the difference as his or her salary. Pooling There is wide variation in the size and number of risk-pools. At the two extremes are schemes that start up with a few dozens households (such as Guatemala and ASSABA) and schemes that operate with several million members (such as Burundi and the health card scheme). From the database of 82 schemes com- plied by Bennett, Creese, and Monasch (1998), the following conclusions emerge about the size of community-financing schemes: · The population covered varied between 40 households and 700 million (Niger). · The share of eligible population in the total local population also varied from less than 1 percent to 90 percent. Review of the Strengths and Weaknesses of Community Financing 87 · The average level of coverage of eligible population in the sample amounted to 37 percent. In addition to small size, pooling is further undermined by adverse selection. As community-based schemes are mostly voluntary and charge a flat premium, adverse selection is often reported as a key difficulty in ensuring financial sus- tainability (Atim 1998; Bennett, Creese, and Monasch 1998; Arhin 1994; Ekman 2001). For example, in the Burundi Health Card Scheme the overall low mem- bership rate (23 percent of households) was associated with community-rated premiums that discouraged low-risk individuals from purchasing CAM cards. Non-CAM members referred to higher care often purchased cards prior to the treatment or at the time of the referral to reduce their financial barrier to expen- sive curative care without having to participate in the risk sharing. Larger house- holds were also more likely to be current or past cardholders (Arhin 1994). Prepayment schemes apply several mechanisms to increase and diversify their risk pool. These include waiting time between registration and eligibility for ben- efits; (mandatory) group-based membership at the family or enterprise or profes- sional association levels; and incentives to register entire families (Atim 1998; Musau 1999; Bennett, Creese, and Monasch 1998; Dave 1991; Atim and Sock 2000). For example, membership in the Chogoria Hospital Insurance Scheme in Kenya is open, but the premium structures favor group memberships. Coffee and tea cooperative societies and schools were the target groups in the commu- nity. However, coverage fell to only a group of hospital employees in 1998­99 because of an inability to attract group memberships, high premium rates for individual memberships, slow services, and insufficient benefits (Musau 1999). Purchasing and Resource Allocation The purchasing and resource allocation function of community-financing schemes is less extensively discussed in the literature than other aspects of their operations. Some schemes rely on third-party reimbursement to members; others pay providers directly. Often, sustainable financing is associated with the community's ability to negotiate preferential rates with providers. Through several mechanisms, the community aims to ensure social control over the doctor-patient relationship and prevent unjustified overuse of services. There are examples of mandatory referrals to use higher level care, treatment guidelines, and various limitations and caps on utilization to prevent moral haz- ard and induced demand. For example, the MHO in Senegal is community-based and covers only hospi- talization. The membership fee is per person insured, although, in general, the household is a member of the mutual and participates in the decisions. If a member needs surgery, he or she has to pay 50 percent of the costs of the opera- tion. Any excess stay at the hospital (beyond 10 to 15 days) is initially covered by the mutual, and then the member has to reimburse it. This seems to keep overutilization of services in check (Jütting 2000). 88 Health Financing for Poor People: Resource Mobilization and Risk Sharing Management Characteristics Besides getting the technical characteristics of the contribution scheme right, good management can also contribute to the success or failure of resource mobi- lization mechanisms. Good management is often as visible for members as the care they receive. For example, if the claims settlement process and other admin- istrative measures prove cumbersome and lengthy, members may be less willing to join the scheme. Perhaps the strongest agreement among the reviewed articles concerns the issue of community involvement in management of schemes as a prerequisite for success (Atim 1999; Atim and Sock 2000; Arhin 1995; Roenen and Criel 1997; Gumber and Kulkarni 2000; Carrin and others 1999; Desmet, Chowdhury, and Islam 1999). Community involvement in scheme management leads to improve- ments in not only revenue collection but also in cost containment, access, and quality of services. These performance measures, in turn, are prerequisites for sus- taining membership levels and thus revenue flows (Hsiao 2001). For example, Atim (1998) identifies the fact that MHOs owe their success to democratic governance, which is one of the original contributions of these schemes. MHOs represent their communities or members before the health authorities, including the providers influencing decisions of resource allocation, and ensure responsiveness of the health authorities to community concerns, which enhances the sustainability of the schemes. The absence of community involvement in management may lead to provider capture and monopoly pricing. For example, the Kanage Cooperative Scheme described by Roenen and Criel (1997) did not have any community par- ticipation in its governance. There was a lack of adequate and relevant technical information to help in the decisionmaking process. There was no dialogue between the population and the hospital, leading to a dominant position for the health service. This may have rendered the system fragile and led to its failure. Even when the community is involved in scheme management, representa- tional issues might arise. Gilson and others (2000) reviewed the experience of Kenya, Zambia, and Benin. Community structures often were not seen as reflect- ing the views of the wider population, critical decisions often did not take into account the interest of the poorest, and the poorest were rarely involved directly in decisionmaking. The authors conclude that the voice of the poorest within the communities often is not heard or is not influential. As a result, community mechanisms by themselves may not adequately address the poor's lack of finan- cial protection. The problem encountered by most schemes, however, is that community- based schemes lack the management and administrative skills necessary for the successful design and operation of prepayment schemes. Such skills would include being able to calculate premium rates, determine benefits packages, mar- ket the scheme, communicate with members, negotiate with providers, keep accounting and bookkeeping records, work with computers, and monitor and evaluate the scheme (Atim 1998). See also box 2.5. Review of the Strengths and Weaknesses of Community Financing 89 BOX 2.5 POOR MANAGEMENT IN THE NKORANZA SCHEME The Nkoranza community-financing health insurance scheme was launched in 1992 in Ghana. The scheme is a hospital-based (modality 3) scheme. It was designed in association with Memisa, a Dutch Christian NGO. The scheme is affiliated with a private district hospital, St. Theresa's Hospital, for which the hospital bills are paid and which is paid on a fee-for-service basis. The scheme has voluntary membership based on a community-rated premium. The found- ing NGO offered to meet any deficits in the first three years of the scheme's oper- ation. The scheme has a low coverage rate of 30 percent of the area's population. Poor management of the Nkoranza scheme affected the enrollment and attrac- tiveness of the scheme. Staff members were not aware of their formal job descrip- tions. There was a lack of training for staff in marketing and community participation methods, general management skills, risk-management techniques, negotiation skills, accounting and bookkeeping, computing skills, scheme-monitoring and eval- uation of the scheme, and some of the hospital staff displayed negative attitudes toward the patients. The management approach was also top-down, with no supervisory body, reflecting the nonparticipatory design of the scheme. The community was not effectively involved in the governance of the project. The community's participation was restricted to education and information cam- paigns. It was a hospital-based scheme, run as another department of the hospi- tal. However, there are rules and regulations governing membership and access to services, which are revised annually and circulated within the community. Source: Atim and Sock (2000). Yet hiring someone with adequate skills to run and manage the scheme may cost the scheme too much. For example, in the Kanage Cooperative Scheme, the salary of the person hired to manage the scheme and register members proved too much for the scheme to bear (about US$670 a year), so the hospital took over. The scheme's total administrative costs were evaluated at 12 percent of its revenues, which may have been grossly underestimated as the rent on the leases was never included (Roenen and Criel 1997). Organizational Characteristics Of the organizational characteristics reviewed, linkages between schemes and providers are reported to be an important determinant of the performance of community-based schemes. Schemes that have a durable partnership arrange- ment or contractual arrangement with providers can negotiate preferential rates for their members. This, in turn, increases the attractiveness of the scheme to the population and contributes to sustainable membership levels. For example, the schemes in the Thiès region of Senegal negotiated preferen- tial rates with the nearby private St. Jean de Dieu Hospital. The hospital is run by a religious organization that is driven by altruistic objectives and has been very 90 Health Financing for Poor People: Resource Mobilization and Risk Sharing supportive of the activities of the Mutual Health Organizations. The negotiated rates allow the schemes to offer considerable benefits with acceptable contribution rates. This makes the schemes very attractive to the population and explains the high penetration rate among the target groups (Atim 1998). Close ties with providers also allow the community to monitor provider behavior and exert social pressure on providers. This can lead to efficiency gains, allowing the schemes to use the resources for noticeable service improvement, which again increases the schemes' attractiveness to the population and is the cornerstone of sustainability. Conversely, inefficiencies due to weak gatekeeping, for example, may lead to moral hazard and wasted resources. In this case, mem- bership may drop if service and quality do not improve and if the costs of the membership are higher than the perceived value of the benefits. The Nkoranza health insurance scheme is an example of this (Atim and Sock 2000). Another level of organizational linkages is the relationship of the scheme to other schemes, in particular to the national government health system and/or social security system. In the Thai Health Card Scheme, the beneficiaries were allowed to use the health provider units under the Ministry of Public Health via the health center or community hospital and follow the referral line. Providers were compensated for the care they provided to health cardholders on a per case basis. They were also reimbursed for administrative expenses incurred by being part of the health card program (Supakankunti 1997). Institutional Characteristics Information is available about certain institutional characteristics of community- financing schemes, such as stewardship and regulation, ownership forms and related governance structures, and markets. However, better understanding is nec- essary to assess how various institutional characteristics contribute to scheme per- formance. This is particularly the case regarding issues of ownership in which modalities are well-formulated, but their impact on performance is less understood. Stewardship. The role of government-level stewardship is often hypothesized as a critical determinant of sustainable health financing through community structures. Some researchers argue that government and donor support are criti- cal for successful and sustainable community-based health financing. This can be financial support but it can also be a supportive policy environment and the provision of training and information opportunities. In Thailand, for example, the government subsidizes half of the cost of the health card. The household contributes half the price of the insurance card dur- ing the income-earning season, while the other half is subsidized by general tax revenue through the Ministry of Public Health. The Ministry of Public Health decentralized the management and decisionmaking to the provincial level, allowing provinces to define their own policies. The premiums, however, remained the same. The health card officers helped increase access to the scheme by providing clear information to the community (Supakankunti 1997). Review of the Strengths and Weaknesses of Community Financing 91 In some provinces of China, the Rural Cooperative Medical System is a joint effort between the government, the villages, and the rural population. Counties and townships played a vital role in the design of the scheme, which was adapted to different local situations (Carrin and others 1999). At the same time, there are community-financing schemes that were created in response to a vacuum in government stewardship--and managed to survive. Criel, van der Stuyft, and van Lerberghe (1999) offer the example of the Bwamanda scheme in the Democratic Republic of Congo, which "succeeded in generating sta- ble revenue for the hospital in a context where government intervention was vir- tually absent and where external subsidies were most uncertain." A more systematic assessment of the various forms of government support (financial, nonfinancial) for community financing and their performance im- pact would make a much-needed contribution toward our understanding of what makes certain schemes work and others fail. Ownership. There are various forms of ownership and related forms of gover- nance: by members (that is, cooperative), by providers, by nongovernmental organizations, by microfinance organizations, and by churches. (See Bennett, Creese, and Monasch 1998 for a comprehensive discussion of ownership forms of community financing.) Each form of ownership can demonstrate successful as well as unsuccessful resource-mobilization schemes. Thus the same conclusion holds as for government regulation--linking alternative ownership forms with performance measures is a much-needed contribution to the field. Markets. Community-financing schemes compete on the factor markets (par- ticularly labor and supplies) with other organizations involved in financing and providing health care. Attracting physicians to remote rural areas where most community-financing schemes work is difficult. Community-financing schemes' effective demand for factors of health services production is hampered by their low ability to pay due to their predominantly poor contributing population. In the health services markets, community-financing schemes often fill a vac- uum, and thus competitive forces do not necessarily apply: community-financing schemes are often initiated in response to the complete absence of other income- protection instruments for the poor against the cost of illness. Thus their mem- bers often do not have a meaningful choice of alternative schemes or other health-financing modalities. At the same time, competition is more likely when the scheme is involved in active purchasing of health services from providers and employs selectivity in the resource allocation process and performance rewards. This is again ham- pered by the geographic monopoly of providers in poor rural areas where many of the schemes operate. Further understanding of how market mechanisms apply to community-financing schemes and how they affect performance would be helpful. In conclusion, the reviewed literature is very rich in terms of describing various technical, managerial, organizational, and institutional features of community- financing schemes. At the same time, better understanding is needed to assess 92 Health Financing for Poor People: Resource Mobilization and Risk Sharing how these structural characteristics contribute to scheme performance. This is particularly true where modalities are now well-formulated but their impact on performance is less understood. CONCLUDING REMARKS This review of 45 studies on community-based health financing has found a number of interesting observations. Perhaps the most obvious conclusion is that the literature on community-based health financing is growing exponentially. This reflects enthusiasm among policymakers and researchers alike about the potential of these schemes to mobilize resources for the health care of the poor. Although this growing literature is varied in terms of focus, content, scope, and approach, the following observations emerge: · The reviewed literature is very rich in describing the nature of community financing and its variants. There is plenty of information about the design of various schemes and also about the implementation process. · Evidence regarding the performance of community financing is building up. In particular, there is rather convincing evidence that community involve- ment in resource mobilization increases access to health care for those cov- ered by these programs while reducing the financial burden of those seeking health care. · At the same time, the need persists for further evidence about the perfor- mance of community-based health-financing arrangements along various measures. Most striking is the lack of knowledge about the number of people covered globally, the extent of their coverage, and the volume of resources mobilized. In the absence of these indicators, assessment of the potential of community financing at a global scale is difficult. · There are a number of definitions and typologies presented in the literature, and this chapter is guilty in adding an additional one. It would be an impor- tant step, however, to arrive at a common definition so that individual stud- ies and presented schemes could be compared more easily. · Accepting that community financing comes in many shapes and forms, a key unanswered question is, What form of community financing is more effective in terms of mobilizing resources for the health of the poor and providing financial protection against the cost of illness? See page 114 for acknowledgments, notes, and references. APPENDIX 2A. PERFORMANCE VARIABLES REPORTED IN THE REVIEWED STUDIES In this section, we list the reviewed 45 studies, grouped according to the modal- ity of the scheme or schemes it reviews (table 2.10). Sections 1 through 4 are the four modalities, Section 5 summarizes studies that address multiple modalities and were large comparative papers, and Section 6 lists the conceptual papers. The performance variables include resource generation, social inclusion, and financial protection. TABLE 2.10 Performance Variables Reported in the Reviewed Studies Countries Resource Social Financial Modality 1: Community cost sharing reviewed generation inclusion protection Others 1. McPake, B., K. Hanson, and Burundi A. Mills. 1993. "Community Guinea Financing of Health Care in Kenya Africa: An Evaluation of the Uganda Bamako Initiative." 2. Ogunbekun, I., and others. Nigeria 1996. "Costs and Financing of Improvements in the Quality of Maternal Health Services through the Bamako Initiative in Nigeria." 3. Soucat, A., T. Gandaho, and Benin others. 1997. "Health Seeking Guinea Behavior and Household Expenditures in Benin and Guinea: The Equity Implications of the Bamako Initiative." 4. Soucat, A., D. Levy-Bruhl, and Benin others. 1997. "Local Cost Guinea Sharing in Bamako Initiative Systems in Benin and Guinea: Assuring the Financial Viability of Primary Health Care." 5. Gilson, L. 1997. "The Lessons Africa Efficiency of User Fee Experience in Equity Africa." Sustainability 6. Supakankunti, S. 1998. Myanmar "Comparative Analysis of Various Community Cost Sharing Implemented in Myanmar." (continued) 93 TABLE 2.10 Continued Countries Resource Social Financial Modality 1: Community cost sharing reviewed generation inclusion protection Others 7. Fiedler, J. L., and R. Godoy. Honduras 1999. An Assessment of the Community Drug Funds of Honduras. 8. Fiedler, J. L., and J. B. Wight. Honduras Quality 2000. "Financing Health Care at the Local Level: The Community Drug Funds of Honduras." 9. Gilson, L., and others. 2000. Benin "The Equity Impacts of Kenya Community Financing Activities Zambia in Three African Countries." 10. Okumara, J., and T. Umena. Vietnam 2001. "Impact of Bamako Type Revolving Drug Fund on Drug Use in Vietnam." Modality 2: Community prepayment Countries Resource Social Financial or mutual health organizations reviewed generation inclusion protection Others 11. Arhin, D.C. 1995. Rural Health Ghana Insurance: A Viable Alternative Guinea-Bissau to User Fees. Burundi 12. Toonen, J. 1995. Community Bolivia Financing for Health Care: A Case Study from Bolivia. 13. Hsiao, W.C. 1995. "The China Chinese Health Care System: Lessons for Other Nations." 14. Ron, A., and A. Kupferman. Philippines 1996. A Community Health Insurance Scheme in the Philippines: Extension of a Community-Based Integrated Project. 15. Liu, Y., and others. 1996. "Is China Community Financing Necessary and Feasible for Rural China?" 16. Desmet, A., A. Q. Chowdhury, Bangladesh and K. Islam. 1999. "The Potential for Social Mobilization in Bangladesh: The Organization and Functioning of Two Health Insurance Schemes." (continued) 94 Appendix--Performance Variables Reported in the Reviewed Studies 95 TABLE 2.10 Continued Modality 2: Community prepayment Countries Resource Social Financial or mutual health organizations reviewed generation inclusion protection Others 17. Ron, A. 1999. "NGOs in Guatemala Community Health Insurance Philippines Schemes: Examples from Guatemala and Philippines." 18. Carrin, G., and others. 1999. China `'The Reform of the Rural Cooperative Medical System in the People's Republic of China: Interim Experience in 14 Pilot Counties." 19. Chen, N., A. Ma, and Y. Guan. China 2000. `'Study and Experience of a Risk-based Cooperative Medical System in China: Experience in Weifang of Shandong Province." 20. Xing-yuan, G., and F. Xue-shan. China Drug use 2000. "Study on Health behavior Financing in Rural China." 21. Gumber, A., and V. Kulkarni. India 2000. "Health Insurance for Informal Sector: Case Study of Gujarat." 22. Jütting, J. 2000. "Do Mutual Senegal Health Insurance Schemes Improve the Access to Health Care? Preliminary Results from a Household Survey in Rural Senegal." 23. Schneider, P., F. Diop, and Rwanda Quality S. Bucyana. 2000. Development and Implementation of Prepayment Schemes in Rwanda. 24. Preker, A. 2001. "Philippines Philippines Mission Report." Modality 3: Provider-based Countries Resource Social Financial community health insurance reviewed generation inclusion protection Others 25. Roenen, C., and B. Criel 1997. China Effectiveness The Kanage Community Rwanda Efficiency Financed Scheme: What Can Sub-Saharan Be Learned from the Failure? Africa (continued) 96 Appendix--Performance Variables Reported in the Reviewed Studies TABLE 2.10 Continued Modality 3: Provider-based Countries Resource Social Financial community health insurance reviewed generation inclusion protection Others 26. Criel, B., P. van der Stuyft, and Democratic Efficiency W. van Lerberghe. 1999. "The Republic Bwamanda Hospital Insurance of Congo Scheme: Effective for Whom? A Study of Its Impact on Hospitalization Utilization Patterns." 27. Atim, C., and M. Sock. 2000. Ghana An External Evaluation of the Nkoranza Community Financing Health Insurance Scheme, Ghana. Modality 4: Community-driven prepayment scheme attached to social insurance or government- Countries Resource Social Financial run system reviewed generation inclusion protection Others 28. Arhin, D. 1994. "The Health Burundi Benefit to Card Insurance Scheme in women Burundi: A Social Asset or a Non-Viable Venture?" 29. DeRoeck, D., and others. 1996. Ecuador Cost and Rural Health Services at demand Seguridad Social Campesino analysis Facilities: Analyses of Facility and Household Surveys. 30. Supakankunti, S. 1997. Future Thailand Quality of Prospects of Voluntary Health care Insurance in Thailand. Studies that address Countries Resource Social Financial multiple modalities reviewed generation inclusion protection Others 31. Dave, P. 1991. "Community India and Self-Financing in Voluntary Health Programmes in India." 32. Diop, F., R. Bitran, and Niger Quality M. Makinen. 1994. Evaluation of the Impact of Pilot Tests for Cost Recovery on Primary Health Care in Niger. 33. Diop F., A. Yazbeck, and Niger Quality R. Bitran. 1995. "The Impact of Alternative Cost Recovery Schemes on Access and Equity in Niger." (continued) Appendix--Performance Variables Reported in the Reviewed Studies 97 TABLE 2.10 Continued Studies that address Countries Resource Social Financial multiple modalities reviewed generation inclusion protection Others 34. Atim, C. 1998. Contribution of Benin Efficiency Mutual Health Organizations to Burkina- Quality Financing, Delivery, and Access Faso Sustainability to Health Care. Synthesis of Cameroon Research in Nine West and Côte d'Ivoire Central-African Countries. Ghana Mali Nigeria Senegal Togo 35. Bennett, S., A. Creese, and Bangladesh Efficiency R. Monasch. 1998. Health Burundi Insurance Schemes for People Cameroon Outside Formal Sector China Employment. Dem. Rep. of Congo Ecuador Guatemala Guinea-Bissau India Indonesia Kenya Madagascar Mali Mexico Nepal Nigeria Papua New Guinea Philippines Tanzania Thailand Vietnam 36. Musau, S. 1999. Kenya Efficiency Community-Based Health Tanzania Quality Insurance: Experiences and Uganda Sustainability Lessons Learned from East and Southern Africa. 37. Atim, C. 1999. "Social Cameroon Efficiency Movements and Health Ghana Insurance: A Critical Evaluation of Voluntary, Nonprofit Insurance Schemes with Case Studies from Ghana and Cameroon." (continued) 98 Appendix--Performance Variables Reported in the Reviewed Studies TABLE 2.10 Continued Studies that address Countries Resource Social Financial multiple modalities reviewed generation inclusion protection Others 38. CLAISS. 1999. "Synthesis of Argentina Equity Micro-Insurance and Other Bolivia Financial Forms of Extending Social Colombia Sustainability Protection in Health in Latin Dominican Quality America and the Caribbean." Republic Ecuador Guatemala Honduras Nicaragua Peru Uruguay 39. Hsiao, W. C. 2001. Unmet China Needs of Two Billion: Is Indonesia Community Financing a Solution? 40. Narula, I. S. and others. 2000. Cambodia Quality Community Health Financing China Financial Mechanisms and Sustainability: Indonesia Sustainability A Comparative Analysis of 10 Lao PDR Asian Countries. Mongolia Myanmar Papua New Guinea Philippines Thailand Vietnam Conceptual papers that did not address any schemes classified Countries Resource Social Financial under the modalities reviewed generation inclusion protection Others 41. Dror, D., and C. Jacquier. 1999. "Micro-Insurance: Extending Health Insurance to the Excluded." 42. Brown, W., and C. Churchill. 2000. Insurance Provision in Low-Income Communities: Part II--Initial Lessons from Micro-insurance Experiments for the Poor. 43. Ziemek, S., and J. Jütting. 2000. Mutual Insurance Schemes and Social Protection. (continued) Appendix--Performance Variables Reported in the Reviewed Studies 99 TABLE 2.10 Continued Conceptual papers that did not address any schemes classified Countries Resource Social Financial under the modalities reviewed generation inclusion protection Others 44. Criel, B. 2000. Local Health Insurance Systems in Developing Countries: A Policy Research Paper. 45. Ekman, B. 2001. Community-Based Health Insurance Schemes in Developing Countries: Theory and Empirical Experiences. APPENDIX 2B. CORE CHARACTERISTICS OF COMMUNITY FINANCING SCHEMES FROM THE REVIEW OF LITERATURE In this section, we list 21 schemes reviewed in the literature grouped by their modality (table 2.11). The design characteristics of the schemes are detailed: technical design characteristics, management characteristics, organizational characteristics, and institutional characteristics. TABLE 2.11 Core Characteristics of Community Financing Schemes, from the Review of Literature Modality 1: Community cost sharing Authors (year) Soucat and others (1997) Name of the scheme Bamako Initiative in Benin and Guinea Technical design characteristics Revenue collection mechanisms · User fee · Voluntary Pooling and risk-sharing arrangements Purchasing and resource allocation · Curative care covered in revitalized health centers · Reduced prices or free care for the poor provided based on a case-by-case basis interview and visual inspection · Highly utilized by children, and low SES exclusion only due to financial reasons · Low price for preventive care due to cross-subsidization of long-term curative care Management characteristics Staff · Large proportion of operating costs covered through user fee funds · Funds retained at health center level and managed locally Culture Access to information Organizational characteristics Organizational forms Incentive regime Linkages Institutional characteristics Stewardship · Community involved in monitoring and budgeting, increases accountability and autonomy Governance · Community sense of ownership Insurance markets Factor and product markets (continued) 100 TABLE 2.11 Continued Modality 2: Community prepayment scheme or mutual health organization Authors (year) Gumber and Kulkarni Desmet, Chowdhury, Desmet, Chowdhury, (2000) and Islam (1999) and Islam (1999) Name of the scheme Self Employed Women's Grameen Health Gonosasthya Kendra, Association (SEWA), program, Bangladesh Bangladesh India Technical design characteristics Revenue collection · Voluntary membership · Prepayment with a · Sliding scale fee mechanisms for families form of scaling in fee structure of premiums · Women beneficiaries structure and copayments · Fixed premium, which · Members are · Voluntary per is low as assets of the beneficiaries of the household based on NGO assist the running Grameen Bank signing of contract of the scheme cooperative Pooling and risk-sharing arrangements Purchasing and resource allocation Management characteristics Staff · Preference for management at the panchayat level · Easy and quick settlement of claims by administrative staff Culture · Top-down approach of · Power struggle in management management scaling down of the interaction with the community to family and individual Access to information Organizational characteristics Organizational forms Incentive regime Linkages Institutional characteristics Stewardship · No active subscriber · No active subscriber involvement involvement Governance Insurance markets Factor and product markets (continued) 101 102 Appendix--Core Characteristics of Community Financing Schemes TABLE 2.11 Continued Modality 2: Community prepayment scheme or mutual health organization (continued) Authors (year) Jütting (2000) ATIM (1999) Arhin (1994) Name of the scheme Mutual Health Mutuelle Famille Abota Village Health Organization (MHO), Babouantou de Insurance Scheme, Senegal Yaoundé, Cameroon Guinea-Bissau Technical design characteristics Revenue collection · Fee per member · Individual or family · Revenue collection mechanisms insured membership--high varies from village to · Generally household is premiums village, from individual a member · Members of same to household basis, in- · 50 percent of costs to be ethnic group kind contribution in the paid in case of surgery form of agricultural to check overutilization, produce accepted any excess stay in the · Prepayment hospital of more than contributions collection 10­15 days initially paid time varied from once by the mutual and to twice a year eventually reimbursed by the member Pooling and risk-sharing · 3 months probationary · Social cohesion arrangements period to check for responsible for adverse selection reducing adverse · Family registration selection and moral incentives to check for hazard adverse selection Purchasing and resource · Covers only · Association pays a · If Abota scheme allocation hospitalization lump sum to member member, referred in the event of patients to the public hospitalization for a health facilities specified time, surgery exempt from for at least 15 days consultation fees · Cannot claim benefits more than once a year · As check on moral hazard, scheme pays fixed amount per person per year (continued) Appendix--Core Characteristics of Community Financing Schemes 103 TABLE 2.11 Continued Modality 2: Community prepayment scheme or mutual health organization (continued) Management characteristics Staff · Mutual aid organization · Decreasing capacity of draws on voluntary government health labor of its members for workers to train and management and other supervise village tasks health workers · No full-time paid staff · Abota funds · No external grants in misappropriated by the income village health workers · Potentially good or staff of the Ministry management staff-- of Health skilled in their own · Drug shortages workplaces, and stiff · Village health workers sanctions exist for attend refresher dereliction of duty training courses Culture · Household participates · Community in the decisionmaking participation in meetings and elections of management · Social control to check fraud, moral hazard, etc. Access to information Organizational characteristics Organizational forms Incentive regime Linkages · Contract with nonprofit · Supplier of drugs is the St. Jean de Dieu Central Medical Store hospital provides a in the capital to reduction of up to 50 government health percent for treatment centers and sectoral hospitals · Government obligated to train and supervise village health workers, supply essential drugs · Support also provided by NGOs such as GVC and WHO/UNICEF evaluation teams (continued) 104 Appendix--Core Characteristics of Community Financing Schemes TABLE 2.11 Continued Modality 2: Community prepayment scheme or mutual health organization (continued) Institutional characteristics Stewardship · No enforcement of · Government essential drug list involvement apart from policy or generic drugs management by both for refunds traditional and political leaders through the village committees · Individual communities develop financing system based on local appropriateness · However, no control of community in purchasing of inputs Governance · Social solidarity is · Has the characteristics prominent of a social institution · Democratic · Community accountability, involvement beyond participation and a mobilization of local sense of ownership is material and labor strong resources Insurance markets Factor and product markets Authors (year) Musau (1999) Musau (1999) Schneider, Diop, and Bucyana (2000) Name of the scheme Mburahati Health Trust Atiman Insurance Community-Based Fund, Tanzania Scheme, Tanzania Health Insurance-- Prepayment Schemes, Rwanda Technical design characteristics Revenue collection · Two types of · Monthly premiums · Annual premium per mechanisms payments: registration paid directly to the family fees to cover Parish Office · Copayment is paid per operational costs · Family or individual episode of care related to start-up of membership · In the pilot project, two scheme and regular · Voluntary membership districts had voluntary contributions (daily) in subscription and one cash or in-kind for daily subscription was income earners through health · Membership based on solidarity fund a nuclear family, flat fee per day per person (continued) TABLE 2.11 Continued Modality 2: Community prepayment scheme or mutual health organization (continued) Pooling and risk-sharing · To prevent moral hazard, · Schemes practice a · One-month waiting arrangements there is social control as short probation or period the group is small waiting period for a · On a health center · There is a 3-month month, in practice level, risk is shared probation period, and varies and adverse within the community; the whole family must selection exists on a hospital level, the enroll in scheme to · Moral hazard risk is risk is shared on a prevent adverse minimized by social district level selection control Purchasing and resource · Includes outpatient · Includes outpatient · Covers basic health allocation care in designated care at local church center package of dispensary, and covers dispensary, no limit to services, drugs, and 10 percent of costs of cost ambulance referral to hospitalization in · Primary care available district hospital public hospital at St. Camillus · Subsidization of · No MCH services Dispensary premiums by included · Members have an ID employers and · Family photograph in card with photograph religious authorities dispensary is required to minimize fraud · Prepayment schemes to prevent fraud, and reimburse health centers the patient signs for by capitation payment treatment received · District hospital reimbursed by district federation on a per episode basis from the schemes' monthly disbursement Management characteristics Staff · Manual recordkeeping · Manual records kept in · Provided regular by different officials of church office, training before and the scheme incomplete after theft after launch of the · All members receive · Weak management of prepayment scheme on training from the the dispensary resulted scheme modalities, SSMECA regarding need in irregularities in accounting tools, for social protection and leadership and administration, characteristics of mutual accountability of the organizational and health insurance schemes dispensary over financial issues, etc. · On-the-job training prescription of drugs, · In order to strengthen related to administration and poor quality care financial and and management · No fraud check organizational received systems in place management · Health care provider also · Scheme's leaders, capacities on the receives training staff, and health care provider side, regarding administration provider have no members prepay for requirements and training on care, and schemes pay adherence to established management of health a capitation rate procedures prior to insurance instead of fee-for- medical treatment service payment (continued) 105 106 Appendix--Core Characteristics of Community Financing Schemes TABLE 2.11 Continued Modality 2: Community prepayment scheme or mutual health organization (continued) Culture · Operates through an · Staff receives regular elected Health feedback on service Committee composed utilization, financial of a chairperson, standing, and secretary, and membership status treasurer, and monthly · Contractual meetings and annual relationship with the general meeting partners of scheme lends democratization in Rwanda Access to information · Population informed about introduction of prepayment schemes via radio spots, newspaper articles, and community and church meetings Organizational characteristics Organizational forms Incentive regime Linkages · With public hospital · With local church · Technical assistance dispensary that reports from the SSMECA to the diocese medical (Strengthening Small director and the and Micro Enterprise medical board and their Cooperatives/ · Linkages with the Association) Christian Mutual · Contract with the Association in Belgium Harlem Agape to develop control Dispensary to provide measures such as health care treatment guidelines · The Medical and official agreement Department of the between scheme and Catholic Secretariat of dispensary the Tanzania Episcopal Church assists the group in checking the treatment forms on a regular basis (continued) TABLE 2.11 Continued Modality 2: Community prepayment scheme or mutual health organization (continued) Institutional characteristics Stewardship · The Scheme · The Parish Office and · Two districts chose for Management Scheme Executive the schemes to be Committee is elected Committee manage managed by providers by members the scheme and population, while · There are formal links · Consultations with one chose to be with the local MOH, and ILO's managed directly by government; Kinondoni SSMECA STEP project the population district cooperative · Some sort of subsidy officer provides reliance exists training on aspects related to cooperative management Governance · Members run the · Community scheme, active participation, members involvement in the attend general meeting design and and elect their implementation of representatives in the scheme Executive Committee Insurance markets Factor and product markets Modality 3: Provider-based community health insurance Authors (year) Atim and Sock (2000) Roenen and Criel (1997) Criel, van der Stuyft, and van Lerberghe (1999) Name of the scheme Nkoranza Community Kanage Cooperative Bwamanda Hospital Health Financing Scheme, Rwanda Insurance Scheme, Scheme, Ghana Democratic Republic of Congo Technical design characteristics Revenue collection · Community-rated · Community-rated · Voluntary scheme mechanisms premiums premiums · Community-rated · Collected Dec.­Jan. · Collected June­Sept. premiums · Voluntary membership · Collected during the · Entire families covered crop-selling season Pooling and risk-sharing · Scheme insists at the · Inverse relationship-- · Family is the arrangements time of admission of poor ends up financing subscription unit patient, that whole the services offered to family be registered the more affluent · Medical officer members of the determines access to cooperative benefits to prevent moral hazard (continued) 107 108 Appendix--Core Characteristics of Community Financing Schemes TABLE 2.11 Continued Modality 3: Provider-based community health insurance (continued) Purchasing and resource · No good surveillance · 20 percent copayment allocation system leading to in case of hospital fraudulent use of admission, which services helps reduce adverse selection Management characteristics Staff · Lack of training in · One staff initially community participation managed enrollment, skills, negotiation skills, but hospital took over accounting and book- due to high costs keeping, computing skills, monitoring and evaluation of scheme Culture · Top-down approach of · No community management involvement Access to information Organizational characteristics Organizational forms · Contract with St. Theresa's Hospital, admission costs are covered Incentive regime · Scheme pays the · Subsidized by Murunda hospital on a fee-for- Hospital service basis Linkages · With private district · Linked to the Murunda · Linked to hospital as hospital--St. Hospital health care provider Theresa's Hospital Institutional characteristics Stewardship · Hospital based · Hospital based Governance · No community · Hospital played a · Managed by the involvement dominant role, no District Health Team community participation Insurance markets Factor and product markets (continued) TABLE 2.11 Continued Modality 3: Provider-based community health insurance (continued) Authors (year) Musau (1999) Musau (1999) Name of the scheme Kisiizi Hospital Health Society, Chogoria Hospital Insurance Uganda Scheme, Kenya Technical design characteristics Revenue collection mechanisms · Premium rates depend on family · Fixed premiums based on size and time period for which individual or family enrollment premiums are paid and benefits included · The scheme is for those who can · All members should also belong afford it, access for the poor is to the Kenya National Hospital not considered Insurance Fund (NHIF) · Voluntary membership to scheme Pooling and risk-sharing · To prevent moral hazard, · To prevent moral hazard, out- arrangements copayments are charged for out- patient visits have a copayment and in-patient services · There is a two-week waiting · At least 60 percent of the group period, exclusion of preexisting has to be enrolled for the scheme conditions, and discount for those to work and prevent adverse who join as a group to prevent selection adverse selection · There is also a waiting period before coverage commences to stop people from joining scheme when they have just fallen sick Purchasing and resource allocation · Includes out-patient care and · Includes outpatient and inpatient in-patient care in general ward care subject to annual limits bed, and has no annual limit · To prevent fraud and abuse, there · Member ID cards are used to is a member ID card with prevent fraud photograph Management characteristics Staff · Manual data management with · Computerized data management no regular reports kept in scheme with monthly reports kept in the office scheme office · Good internal control over use of · Good internal controls over use of hospital services, and external service and external audit along audit to prevent fraud with monthly reports on · Some hospital staff have a utilization help prevent fraud and negative attitude toward scheme abuse of scheme members · Not enough staff members · Delay in processing claims so that they can collect drugs from pharmacy · Hospital has a computerized financial accounting system Culture Access to information · Scheme conducts education meetings to help prospective members understand the scheme (continued) 109 TABLE 2.11 Continued Modality 3: Provider-based community health insurance (continued) Organizational characteristics Organizational forms Incentive regime Linkages · No separation between the · With the Chogoria Hospital under scheme and the hospital, and the the Presbyterian Church of East scheme is part of the hospital Africa and hence no contractual · Current members are all agreement exists employees of the hospital Institutional characteristics Stewardship · The Kisiizi Hospital Committee · The Hospital Committee manages Consultative group manages the scheme along with the community · Technical assistance from the members MOH and USAID funded Kenya · Scheme recognized and Health Care Financing Project supported by the MOH and the Ugandan Community-Based Health Financing Association · The scheme falls under the Community-Based Health Care program of the hospital Governance · Community participation in · No community participation design and implementation of scheme, and management of scheme Insurance markets Factor and product markets Modality 4: Community-driven prepayment scheme attached to social insurance or government-run system Authors (year) Xing-yuan and Xue-shan Supakankunti (1997) Carrin and others (1999) (2000) Name of the scheme Cooperative Health Care Thai Health Card Rural Cooperative Scheme (CHCS), China Scheme, Thailand Medical System (RCMS), China Technical design characteristics Revenue collection · Funded by peasants · Voluntary prepaid · Voluntary mechanisms and government scheme · One-time registration, · Half the price of the contributions collected insurance card is paid once a year by the household · Subsidy by government during the cycle, depending on seasonal fluctuations, and the other half is subsidized by general tax revenue through the MOPH (continued) 110 TABLE 2.11 Continued Modality 4: Community-driven prepayment scheme attached to social insurance or government-run system (continued) Pooling and risk-sharing · Problem of adverse · All funds pooled into arrangements selection and over- one account except in 8 utilization of services townships where risk sharing was limited due to separate accounts for farmers and workers Purchasing and resource · Provides curative and · 80 percent of the funds · Provides hospital care allocation preventive care from the health card is at the township and allocated to compensate county level providers, and 20 percent for administrative costs Management characteristics Staff · Effective as most of the · Technical support funds were spent on provided by a Central health care and only 6­7 Technical Team, percent on management comprising representatives from the MOH, medical universities, and WHO Culture · Decentralized by the MOPH to the provincial level to define their own policies Access to information · Health card officers effective in providing clear information to the community Organizational characteristics Organizational forms Incentive regime Linkages · Beneficiaries used health provider units under the MOPH via health center or community hospital and referral line Institutional characteristics Stewardship · Joint financial effort by the government, villages, and the rural population Governance · Counties and townships played a vital role in the design of the scheme adapted locally (continued) 111 TABLE 2.11 Continued Modality 4: Community-driven prepayment scheme attached to social insurance or government-run system (continued) Insurance markets Factor and product markets Authors (year) DeRoeck and others Arhin (1994) Musau (1999) (1996) Name of the scheme Seguridad Social La Carte d'Assurance Community Health Fund Campesino (SSC), Maladie (CAM), Burundi (CHF), Tanzania Ecuador Technical design characteristics Revenue collection · Urban payroll tax and · CAM card purchased · Voluntary participation mechanisms subsidies from by household entitles except for civil government's general two adults and all servants employed by budget and investment children younger than the Ministry of Local income pay for the 18 to free health care Government rural population at all public health · Pricing of benefits enrolled in the scheme facilities package based on · SSC members contribute · Fixed price out-patient department a small monthly fee that (community-rated health services makes up less than 5 premium) percent of the program's · Valid for one year and budget purchased at any time · Voluntary membership for whole family · Scheme study found user fees being charged largely for drugs, even to members Pooling and risk-sharing · From urban workers to · Adverse selection of · There is adverse arrangements the rural poor households was a selection in the scheme major problem due to · Fixed premiums do not larger households recognize the ability of being more likely to the community to pay purchase card · There is a mechanism · Moral hazard also a for the very poor to be huge problem exempt from paying for participation in the scheme · However, mandatory user-fee program together with CHF, has eliminated inappropriate use of services Purchasing and resource · Provides medical and · If no card, user charges · Includes outpatient allocation dental outpatient determined by health care and has no limit services, maternity, pre- worker · Member ID cards are and postnatal care, · Names of members used to prevent fraud outreach activities, written on card, thus health education, and preventing fraudulent follow-up home visits use 112 (continued) Appendix--Core Characteristics of Community Financing Schemes 113 TABLE 2.11 Continued Modality 4: Community-driven prepayment scheme attached to social insurance or government-run system (continued) Management characteristics Staff · Shortages of drugs and · Shortage of drugs is a · Manual recordkeeping fulltime medical staff problem at facilities and district led to the 50 percent · Health worker headquarters, and decrease in utilization discriminated against computer spreadsheets of clinics CAM holders in favor at headquarters · Medical staff stresses of cash payers · Friendly staff the need for frequent in- · Few female medical · Good drug availability service training; training technicians (poor ante- in specialized in-patient natal care) care appropriate to urban area problems is often provided Culture · Top-down management · Revenue retained by local · Top-down approach from central office to committees that have from MOH to DMO to regional community financial responsibilities, CHF Ward Committee clinics although in practice only and community a small fraction used for health Access to information Organizational characteristics Organizational forms Incentive regime Linkages · Provides primary · Health worker salaries · No contract between health care outpatient and drugs funded by the CHF and service services through a government providers network of 549 small · Public health facilities health clinics in remote and health centers and rural areas and coastal dispensaries participate and mountain regions in the CHF Institutional characteristics Stewardship · Administered through the · National health · The management is by government division of insurance scheme the District CHF Board, Instituto Ecuadoriano implemented by Ward Health Committee, Seguro Social (IESS), government and facility staff including procurement of · Initiated by the MOH medicines, hiring government initiative, employees, management, and receives full budget recognition (continued) 114 Appendix--Core Characteristics of Community Financing Schemes TABLE 2.11 Continued Modality 4: Community-driven prepayment scheme attached to social insurance or government-run system (continued) Governance · Community participation in the management of the fund and running of the public health facilities Insurance markets Factor and product markets Acknowledgments: The authors of this chapter are grateful to the World Health Organiza- tion (WHO) for having provided an opportunity to contribute to the work of the Commission on Macroeconomics and Health and to the World Bank for having published the material in this chapter as an HNP Discussion Paper. Many individuals from various organizations helped us compile an extensive list of pub- lished and unpublished studies for this review. We are grateful to all who helped us with this exciting investigative work. We would particularly like to thank Chris Atim (Abt Associates) for his insightful comments and suggestions. We are grateful to Christian Baeza (International Labour Organisation) who shared with us ILO's extensive experience and research results. Andrew Creese (WHO) and Sara Bennett (Abt Associates) greatly helped our work by making available the database of their research on health insurance in the informal sector. NOTE 1. There may be some bias in the above conclusion as a result of "publication bias." It could be that research that found no difference on performance is less likely to be pub- lished. In addition, successful schemes are more likely than failed ones to make their way into studies. Review of the Strengths and Weaknesses of Community Financing 115 REFERENCES Arhin, D. C. 1994. "The Health Card Insurance Scheme in Burundi: A Social Asset or a Non-Viable Venture?" Social Science and Medicine 39(6): 861­70. ------. 1995. Rural Health Insurance: A Viable Alternative to User Fees. London School of Hygiene and Tropical Medicine, Discussion Paper. London. Atim, C. 1998. Contribution of Mutual Health Organizations to Financing, Delivery, and Access to Health Care: Synthesis of Research in Nine West and Central African Countries. Technical Report 18. Partnerships for Health Reform Project, Abt Associates Inc., Bethesda, Md. ------. 1999. "Social Movements and Health Insurance: A Critical Evaluation of Volun- tary, Non-Profit Insurance Schemes with Case Studies from Ghana and Cameroon." Social Science and Medicine 48(7): 881­96. Atim, C., and M. Sock. 2000. An External Evaluation of the Nkoranza Community Financing Health Insurance Scheme, Ghana. Technical Report No. 50. Partnerships for Health Reform Project, Abt Associates Inc., Bethesda, Md. Bennett S., A. Creese, and R. Monasch. 1998. Health Insurance Schemes for People Outside Formal Sector Employment. ARA Paper 16. World Health Organization (WHO), Geneva. Brown, W., and C. Churchill. 2000. Insurance Provision in Low-Income Communities: Part II--Initial Lessons from Microinsurance Experiments for the Poor. Microenterprise Best Prac- tices Project, Development Alternatives Inc., Bethesda, Md. Carrin, G., R. Aviva, H. Yang, H. Wang, T. Zhang, L. Zhang, S. Zhang, Y. Ye, J. Chen, Q. Jiang, Z. Zhang, J. Yu, and L. Xi. 1999. "The Reform of the Rural Cooperative Medical System in the People's Republic of China: Interim Experience in 14 Pilot Counties." Social Science and Medicine 48: 961­72. Chen, N., A. Ma, and Y. Guan. 2000. "Study and Experience of a Risk-Based Cooperative Medical System in China: Experience in Weifang of Shandong Province." Paper pre- sented at the International Conference on Health Systems Financing in Low-Income African and Asian Countries, November 30­December 1, Centre d`Etudes et de Recherches sur le Développement International (CERDI), Clermont-Ferrand, France. CLAISS (Latin American Center for Health Systems Research). 1999. "Synthesis of Micro- Insurance and Other Forms of Extending Social Protection in Health in Latin America and the Caribbean." Under the supervision and guidance of the ILO and PAHO coun- terparts, for the ILO-PAHO initiative of extending social protection in health in Latin America. Paper presented to the Mexico City tripartite meeting of ILO with the collab- oration of PAHO, Mexico City. Criel, B. 2000. Local Health Insurance Systems in Developing Countries: A Policy Research Paper. Departement Volksgezondheid, Instituut voor Tropische Geneeskunde, Antwerp. Criel, B., P. Van der Stuyft, and W. Van Lerberghe. 1999. "The Bwamanda Hospital Insur- ance Scheme: Effective for Whom? A Study of Its Impact on Hospitalisation and Utili- sation Patterns." Social Science and Medicine 48(7): 879­911. Dave, P. 1991. "Community and Self-Financing in Voluntary Health Programmes in India." Health Policy and Planning 6(1): 20­31. DeRoeck, D., J. Knowles, T. Wittenberg, L. Raney, and P. Cordova. 1996. Rural Health Ser- vices at Seguridad Social Campesino Facilities: Analyses of Facility and Household Surveys. Health Financing and Sustainability Project, Technical Report 13. Partnerships for Health Reform Project, Abt Associates Inc., Bethesda, Md. 116 Health Financing for Poor People: Resource Mobilization and Risk Sharing Desmet, A., A. Q. Chowdhury, and K. Islam, M.D. 1999. "The Potential for Social Mobi- lization in Bangladesh: The Organization and Functioning of Two Health Insurance Schemes." Social Science and Medicine 48: 925­38. Diop, F., R. Bitran, and M. Makinen. 1994. Evaluation of the Impact of Pilot Tests for Cost Recovery on Primary Health Care in Niger. Technical Report 16. Health Financing and Sus- tainability Project, Abt Associates Inc., Bethesda, Md. Diop, F., A. Yazbeck, and R. Bitran. 1995. "The Impact of Alternative Cost Recovery Schemes on Access and Equity in Niger." Health Policy and Planning 10(3): 223­40. Dror, D., and C. Jacquier. 1999. "Microinsurance: Extending Health Insurance to the Excluded." International Social Security Review 52(1): 71­97. Ekman, B. 2001. Community-Based Health Insurance Schemes in Developing Countries: Theory and Empirical Experiences. Lund University Center for Health Economics (LUCHE), Department of Economics. Lund University. Fiedler, J. L., and R. Godoy. 1999. An Assessment of the Community Drug Funds of Honduras. Tech- nical Report No. 39. Partnerships for Health Reform, Abt Associates, Inc., Bethesda, Md. Fiedler, J. L., and J. B. Wight. 2000. "Financing Health Care at the Local Level: The Com- munity Drug Funds of Honduras." International Journal of Health Planning and Manage- ment 15: 319­40. Gilson, L. 1997. "The Lessons of User Fee Experience in Africa." Health Policy and Planning 12(4): 273­85. Gilson, L., D. Kalyalya, F. Kuchler, S. Lake, H. Oranga, and M. Ouendo. 2000. "The Equity Impacts of Community Financing Activities in Three African Countries." International Journal of Health Planning and Management 15: 291­317. Gumber, A., and V. Kulkarni. 2000. "Health Insurance for Informal Sector: Case Study of Gujarat." Economic and Political Weekly, September 30, 3607­13. Hsiao, W. C. 1995. "The Chinese Health Care System: Lessons for Other Nations." Social Science and Medicine 41(8): 1047­55. ------. 2001. Unmet Health Needs of Two Billion: Is Community Financing a Solution? World Bank, HNP Dicussion Paper. Washington, D.C. Jütting, J. 2000. Do Mutual Health Insurance Schemes Improve the Access to Health Care? Pre- liminary Results from a Household Survey in Rural Senegal. Paper presented at the Interna- tional Conference on Health Systems Financing in Low-Income African and Asian Countries, November 30­December 1, CERDI, Clermont-Ferrand, France. Liu, Y., S. Hu, W. Fu, and W. C. Hsiao. 1996. "Is Community Financing Necessary and Fea- sible for Rural China?" Health Policy 38: 155­71. McPake, B., K. Hanson, and A. Mills. 1993. "Community Financing of Health Care in Africa: An Evaluation of the Bamako Initiative." Social Science and Medicine 3(11): 1383­95. Musau, S. 1999. Community-Based Health Insurance: Experiences and Lessons Learned from East and Southern Africa. Technical Report 34. Partnerships for Health Reform Project, Abt Associates, Inc., Bethesda, Md. Narula, I. S., J. T. G. Tan, R. Knippenberg, and N. Jahanshahi. 2000. Community Health Financing Mechanisms and Sustainability: A Comparative Analysis of 10 Asian Countries. Paper presented at the International Conference on Health Systems Financing in Low-Income African and Asian Countries, November 30­December 1, CERDI, Clermont-Ferrand, France. Review of the Strengths and Weaknesses of Community Financing 117 Ogunbekun, I., O. Adeyi, A. Wouters, and R. H. Morrow. 1996. "Costs and Financing of Improvements in the Quality of Maternal Health Services through the Bamako Initia- tive in Nigeria." Health Policy and Planning 11(4): 369­84. Okumara, J., and T. Umena. 2001. "Impact of Bamako Type Revolving Drug Fund on Drug Use in Vietnam." Http://www.who.int/dap-icium/posters/3D3_TXTF.html. Preker, A. S. 2001. "Philippines Mission Report." Internal document. The World Bank. Preker, A. S., G. Carrin, D. Dror, and M. Jakab. 2001. Health Care Financing for Rural and Low-income Populations: The Role of Communities in Resource Mobilization and Risk Shar- ing. A draft background report submitted to the Commission on Macroeconomics and Health. WHO, Geneva. Roenen, C., and B. Criel. 1997. The Kanage Community Financed Scheme: What Can Be Learned from the Failure? Antwerp: Institute of Tropical Medicine. Ron, A. 1999. "NGOs in Community Health Insurance Schemes: Examples from Guatemala and Philippines." Social Science and Medicine 48: 939­50. Ron, A., and A. Kupferman. 1996. A Community Health Insurance Scheme in the Philippines: Extension of a Community-Based Integrated Project. Geneva: WHO; London: International Cooperation­World ORT Union. Schneider, P., F. Diop, and S. Bucyana. 2000. Development and Implementation of Prepayment Schemes in Rwanda. Technical Report 45. Partnerships for Health Reform Project, Abt Associates Inc., Bethesda, Md. Soucat A., T. Gandaho, D. Levy-Bruhl, X. de Bethune, E. Alihonou, C. Ortiz, P. Gbedonou, P. Adovohekpe, O. Camara, J. M. Ndiaye, B. Dieng, and R. Knippenberg. 1997. "Health Seeking Behavior and Household Expenditures in Benin and Guinea: The Equity Impli- cations of the Bamako Initiative." International Journal of Health Planning and Manage- ment 12(S1): 137­63. Soucat, A., D. Levy-Bruhl, P. Gbedonou, K. Drame, J. P. Lamarque, S. Diallo, and R. Osseni. 1997. "Local Cost Sharing in Bamako Initiative Systems in Benin and Guinea: Assuring the Financial Viability of Primary Health Care." International Journal of Health Planning and Management 12(S1): 109­35. Supakankunti, S. 1997. Future Prospects of Voluntary Health Insurance in Thailand. Takemi Research Paper 13. Takemi Program in International Health, Harvard School of Public Health. Harvard University, Cambridge, Mass. Supakankunti, S. 1998. Comparative Analysis of Various Community Cost Sharing Imple- mented in Myanmar. Paper presented to the Workshop of Community Cost Sharing in Myanmar, November 26­28. Toonen, J. 1995. Community Financing for Health Care: A Case Study from Bolivia. Bulletin 337. Royal Tropical Institute, Amsterdam. Xing-yuan, G., and F. Xue-shan. 2000. "Study on Health Financing in Rural China." Paper presented at the International Conference on Health Systems Financing in Low- Income African and Asian Countries, November 30­December 1, CERDI, Clermont- Ferrand, France. Ziemek, S., and J. Jütting. 2000. Mutual Insurance Schemes and Social Protection. Bonn: Cen- ter for Development Research (ZEF) Bonn. CHAPTER 3 Experience of Community Health Financing in the Asian Region William C. Hsiao Abstract: One of the most urgent and vexing problems around the world is how to finance and provide health care for the more than two billion peasants and ghetto dwellers in low- and middle-income countries. The first part of this chapter develops a conceptual framework for community financing and uses it to clarify and classify the variety of community-financing schemes. This section of the chapter discusses the impact of community-financing schemes on outcomes and compares them to several scheme outcomes in African countries. The second part of the chapter uses the concep- tual framework developed above to explain why some community-financing schemes in Asia have been successful and why some have failed. The review points to a number of measures governments could take to strengthen such community financing. Those mea- sures include subsidizing the premiums of the poor, providing technical assistance to improve scheme management capacity, and forging links with formal health care net- works. Satisfaction with a particular scheme was often related to the nature of direct com- munity involvement in its design and management. A critical factor was the matching willingness and ability of individuals to pay, with the expectation of benefits to be received at some later time. The review also highlights several areas of government actions that appear to have a negative impact on the function of community-financing schemes. Top-down interference with scheme design and management appear to have a particularly negative impact on their functioning and sustainability. A most urgent and vexing problem around the world is how to finance and provide health care for the more than two billion peasants and ghetto dwellers in low- and middle-income countries.1 Most of them are poor. Today, these two billion people do not have adequate health care to meet their basic needs (World Bank 2001). Most countries try to serve this population by operating public clinics in rural areas, but getting qualified practitioners to staff the clinics is often difficult. Practitioners frequently evade or refuse assignment there or do not attend the clinics regularly; when they are there, they often pro- vide poor customer service. In addition, the facilities lack drugs and supplies. Thus when people become ill, they turn first to home remedies. Unsuccessful self-treatment frequently leads to big bills for extensive use of outpatient services by traditional healers, private practitioners, and pharmacists. When serious ill- ness strikes, the poor flood into, and overcrowd, the public and charity hospi- tals. In numerous countries, the patients have to pay for inpatient hospital 120 Health Financing for Poor People: Resource Mobilization and Risk Sharing services, and many of them have to bankrupt their families to pay for the services or forgo the treatment and die. Studies found higher proportions of women and children forgoing medical treatment. Studies also consistently found that poor households use a very large part of their income for health care, even when the government theoretically provides free, or nearly free, services.2 Studies in several countries, including China, found that large medical expenditure (for example, inpatient hospital services and costly outpatient drugs) is the major cause of poverty (Liu 2001). These facts raise at least three serious questions. First: Is a nation spending a reasonable amount for its health? Many countries do not spend enough for the health care of their rural residents and urban poor. Can the governments spend more? It depends. Most low-income nations have a tax base that is too narrow and a tax collection system that is too ineffective to yield large sums of general revenue. The result is inadequate public funding for basic health care for the rural and ghetto households.3 Other well-known financ- ing modalities are unfeasible or undesirable. Social and private insurance are unviable. User fees are inequitable and create high barriers for the poor who need access to health care. (Whether foreign sources and domestic governments can allocate additional funds to support the health care of this population is being addressed in another chapter.) Second: Does a nation have the capacity to transform money into effective services for the rural and poor population? In many countries where the govern- ment funds and provides free, or nearly free, services for rural residents and the poor, the target population is not utilizing those public health services. These households use their meager income to pay for services and drugs from the pri- vate sector. Why? Detailed country studies have consistently found a disturbing fact. In most low-income countries, although governments fund public provi- sion of primary care at the village and township (subdistrict) levels,4 they cannot manage and monitor these publicly funded services at the grassroots level. Whatever the amount of funds spent, it does not produce the services people want and value, though facilities are built and staffed (Bitran 1995; Gilson 1995; Zere, McIntyre, and Addison 2001). As a result, when people become ill, they pay to see the private practitioners and buy drugs themselves. (The findings of these studies are also summarized in another chapter on efficiency.) Third: Can these services be organized so they will be used more efficiently and effectively? Out-of-pocket payments to private providers have some serious drawbacks. For one, there is no risk pooling. Patients also have to pay whatever private practitioners and drug peddlers charge. At the village level, the prices can be high since the population size is not likely to be able to support competing providers. At the subdistrict (township) level, the competition is also limited because of population size. In addition, the health service market suffers from well-documented market failures that can result in price gouging, poor quality of medical care, and induced demand for drugs sold at a high profit.5 If households are willing to prepay the amount that they now pay out of pocket into an orga- nized financing scheme, collective gains can be obtained. The organized fund Experience of Community Health Financing in the Asian Region 121 could pool risks, improve quality and expand the delivery of health care, using the same amount of money. Combinations of the above three problems lead to unmet health needs around the world. The different combinations that cause the problem must be identified since they need different policy remedies. For many very low-income countries in Africa, the causes of their unmet health needs are clear: underfund- ing as well as the inability of their health systems to transform their money into effective health care for rural and poor populations. China, Egypt, India, and Kenya spend reasonable amounts, but their health systems cannot transform the money into effective services for rural and poor populations. In contrast, Sri Lanka, spending a modest amount, has produced enviable results in health sta- tus and risk protection. As a result, we have some confidence that additional public spending by Sri Lanka could yield significant gains, but we cannot say that about India. Throughout the world, community financing has been used to mobilize resources to fund and deliver health care for rural and ghetto communities. Some types of community-financing schemes have been successful in addressing all of the three issues discussed above, while others are primarily income-generating schemes for providers (for example, the Community Health Fund in Tanzania). In recent years, community financing has become a term used loosely by health- financing specialists to label any financing scheme that may involve some com- munity contribution or participation. Schemes range from the Drug Revolving Funds, which rely on user fees to fund a continuous availability of drugs, to government-managed prepayment schemes that require a community's residents to contribute to funding for public facilities, to hospital-sponsored and -managed insurance schemes that principally cover only that hospital's services. These schemes are very different in nature and purpose, in population covered, in ben- efit structure, in the extent of risk pooling, and in management.6 Labeling any scheme that involves the community as a community-based health financing (CF) scheme has confused health policy leaders about which types of schemes are viable and how CF can alleviate the health needs of two billion people. It also impairs researchers in their investigations of the key common characteristics of community financing that explain the success or failure of such a scheme so that countries around the world can have a generalized concept of CF to assess when they can take the individual successful cases of it to scale. This chapter has two parts. The first develops a conceptual framework for com- munity financing and uses it to clarify and classify different types of schemes. The framework is intended to clarify the ambiguity, variations, and perplexity of community-financing schemes to gain some insight into the characteristics that make community financing a success or failure. To do that, we have to ask, Suc- cess or failure in what? This chapter offers two criteria: the potential community financing has to cover a significant percentage of the target population and mobi- lize resources and the value CF adds to the outcomes that matter to a society. We evaluate the schemes' impact on outcomes and compare them to several schemes 122 Health Financing for Poor People: Resource Mobilization and Risk Sharing in Africa. The comparative analysis provides some insight into which character- istics matter the most for establishing and sustaining community financing and what benefits schemes bring. The second part of the chapter uses the conceptual framework to explain several community-financing schemes in Asia and reasons for their success or failure. WHAT IS COMMUNITY FINANCING? Community-based health funds have existed for centuries. The earliest ones were largely sponsored by local religious organizations such as churches and syna- gogues. In the last century, community cooperatives, local mutual aid societies, and local funeral funds have sponsored and managed local health funds. The ini- tiation of a nationwide community-based and -managed program in China--the Cooperative Medical System (CMS)--in the late 1950s captured the world's atten- tion with the potential of community-based efforts to mobilize resources and provide cost-effective health care for the rural population. Other well-known, successful community-based financing and provision programs include the Thai Health Card Scheme and Indonesia's Dana Sehat. Each scheme covers millions of rural people for primary care and some secondary hospital services. Other local schemes such as Grameen Health Program, Dhaka Community Hospital Insur- ance Program, and the Self-Employed Women's Association (SEWA) have been successfully established and cover thousands of low-income households. Several major studies have reviewed and summarized the numerous community- financing schemes around the world. Stinson (1982) was the first to compile an inventory and brief description of close to 200 financing schemes. Bennett, Creese, and Monasch (1998) analyzed the risk-pooling characteristics of more than 80 schemes globally. Most recently, Atim (1998) conducted a study of 22 mutual health organizations in Africa. These three comprehensive studies cov- ered schemes ranging from local prepaid user fee plans and church-sponsored, traditional, third-party insurance schemes to universal compulsory social insur- ance for target populations. The schemes' wide variations make it almost impos- sible to understand what constitutes a community-financing scheme much less grasp what made these schemes successful. Community financing can be broadly defined as any scheme that has three features: community7 control, voluntary membership, and prepayment for health care by the community members.8 This definition would exclude financ- ing schemes such as regional compulsory social insurance plans and community- managed user fee programs. However, the definition here is still too vague for analytical purposes when we try to understand what makes a community- financing scheme a success or failure. Two analytical definitions of community financing can be derived from our strategic framework in mobilizing domestic resources for health outlined in chap- ter 1 of the Report of Working Group 3, Commission on Macroeconomics and Experience of Community Health Financing in the Asian Region 123 Health. Our framework argued that community financing is one of several financ- ing modalities to raise funds. Then one way to differentiate community-financing schemes is by examining their capability to mobilize resources and population coverage. Our framework also suggests another way to differentiate CFs. Financing is an instrument used to achieve societal goals. We are ultimately interested in a financing scheme's impacts on the outcomes of a health system. Therefore, another way to differentiate CFs is by the final outcomes they produce. We explain in greater detail about the two approaches in classifying and studying CFs below. Classify and Analyze Community-Financing Schemes by Their Potential to Mobilize Fiscal Resources and Attract a Large Percent of Target Population to Enroll A CF scheme's ability to raise funds from households depends on their ability to pay. A scheme's capacity to raise funds from community businesses depends on the number of rural enterprises and cooperatives. Poverty households have to be heavily subsidized, and poor households also require some subsidizing. What would motivate households that have some ability to prepay to do so? Their willingness to prepay must be the primary determinant. Economic and social factors influence their willingness to prepay (Hsiao 1995, Bennett, Creese, and Monasch 1998). From the economic perspective, the expected economic and quality gains have to be equal or greater than the prepayment. Social norms and close relationships, however, may shape people's preference for prepayment that involves elements of income transfer. The economic and social considera- tions are discussed below. The chapter then discusses which specific gains mem- bers are likely to value. In examining a CF scheme's ability to produce economic and quality gains, we argue that it is dependent on the management's motiva- tion. Management's motivation and competence are manifest in certain organi- zational and incentive characteristics known to improve efficiency and quality of health care. We identify and discuss them in detail. Major Determinants of Community Members' Willingness to Prepay Economic theory suggests that households' willingness to prepay depends on their belief that by doing so they will gain economically or in health care or both. In other words, the expected benefit has to be greater than the cost. That could happen in three ways. First, the existing facilities could produce the patient-valued services more efficiently (including reducing corruption) so that the prepayment would buy more than it presently does. Second, the prepayment could purchase something new that is valued by the household, such as risk pro- tection. Third, the government could provide a direct and visible subsidy to motivate community members to join a CF. For example, when the government matches every dollar the community member prepays, members can easily see the economic gain. Other forms of subsidy can be the discounting of the price CF members pay for services or drugs or both. 124 Health Financing for Poor People: Resource Mobilization and Risk Sharing As for the social characteristics of a community that may influence households' willingness to prepay, we hypothesize that social capital could influence people's preference to prepay. Prepayment implicitly involves risk pooling and cross- subsidizing between the healthy and the less healthy, and between the rich and the poor. Young and healthy people will not enroll if they have to prepay a similar amount as the elderly and less healthy people. But sociologists have long argued that social capital is an important determinant of people's willingness to cooperate with each other. This theory has been supported by several empirical economic and political studies (Putnam 1993; Liu 2001; Narayan and Pritchett 1997). We hypothesize that the degree of mutual concern that community members have for each other (social cohesion and solidarity) could have a significant influence on their willingness to prepay, even when an individual household is uncertain that the expected benefit will be greater than the amount to be pre- paid. In economic terms, it means social cohesion and mutual concern shape people's preference for prepayment. We hypothesize that the greater the social capital, the more people are willing to prepay. In a simple diagram, we can illustrate the interactions between economic gains and social capital and that their sum has to reach a threshold--the prepayment required. We show that a CF scheme can be successful even when the prepayment amount is greater than the expected economic or health gain because of the social capital. This hypothesis is illustrated in figure 3.1. While community A can produce the same level of economic and quality gains as community B, A has less social cap- ital; the sum of the two for A did not exceed the prepayment amount required. Con- sequently, community A did not establish a CF scheme. Community B, with its FIGURE 3.1 Feasibility of Establishing Community Financing and the Amount the Average Person Is Willing to Pay as a Function of Expected Gains and Social Capital Total government revenues as % GDP 100 · Governments in many countries often 80 raise less than 20% of GDP in public revenues; and 60 40 · The tax structure in many low-income countries is often regressive. 20 0 100 1,000 10,000 100,000 Per capita GDP (log scale) Experience of Community Health Financing in the Asian Region 125 greater social capital, was able to establish a scheme. Our hypothesis may explain why some CF schemes have been successful while others have failed since they vary in their ability to produce gains, and communities have varied social capital. Supported by the Chinese government and UNICEF, we have been conduct- ing social experiments in 10 poor rural counties in eight different provinces in China, testing whether social capital affects rural households' willingness to pre- pay. Liu (2001) developed several measurements of social capital, including the degree of mutual assistance a household has given and the participation in civic activities in the villages. His preliminary regression analysis found a statistically significant association between these social capital variables and people's will- ingness to pay and their actual enrollment in CF schemes. What Gains May Be Valued by Community Members? Numerous studies have examined the reasons why rural population have volun- tarily enrolled and stayed enrolled in different schemes. Market surveys have also been conducted in China and in Indonesia to glean what kind of health care and risk pooling people prefer to prepay and how much they are willing to prepay. Overall, these studies found that people most often mentioned the following items among the top three products they valued most: availability of close-by and affordable9 primary care and drugs; some protection against high financial risks such as hospital charges; neat and clean facilities particularly outhouses or bath- rooms; reasonably competent practitioners; and good customer service. Various studies conducted in Africa, Asia, and Latin America also had similar findings. We summarize the products valued by community members in table 3.1. When Would a CF Scheme Give Priority to Producing Economic Gains and Improving Quality of Services for the Patients? Though we can understand what products community members value highly, the producers may still not have supplied them, particularly under public provision. Under what circumstances would CF schemes be motivated to produce the eco- nomic and quality gains valued by the community members? Management control is one critical determinant. Household interview surveys consistently found com- munity members' key concern was whether their funds would be used exclusively TABLE 3.1 How Community Members Valued Service Availability, Quality, Risk Protection, and Costs Quality of services (competence, cleanliness, Extent of Availability of affordable services and custom service) risk protection Costs Primary Primary Charges care and care and at time Preventive drugs Hospital Preventive drugs Hospital Travel of use High High Modest Modest High High Modest High High 126 Health Financing for Poor People: Resource Mobilization and Risk Sharing for their benefit. Corruption is a major worry together with excessive spending on staff compensation and services that have less value to the patients. Consequently, to be willing to pay, people must trust and have confidence in the organization that manages the fund. In most low-income countries, the government has not earned the trust and confidence of the people at the village and township levels (Gilson and others 1994; Hsiao 1995). When this is the case, nongovernmental organizations (NGOs) that have the people's confidence must manage the fund. Other managers could include local agricultural cooperatives, churches and mosques, funeral funds, or a newly formed community organization. This aspect--people's confidence and trust in the organization managing the fund--is a precondition for a scheme's success. We believe that in general, governments are less capable than the local com- munity in managing the services for the patients' benefit at the village and sub- district levels. The reasons are straightforward. Governments face financial and human resource constraints in managing thousands of clinics at village and township levels. But CF scheme members have a self-interest in seeing that their prepayment is used wisely and efficiently for services they value. When the local community has significant control over the scheme, members can have a greater voice in deciding how the funds should be spent. Members can manage the effi- ciency and quality of services much more effectively because they can readily monitor the staff's regular attendance at the clinic and the availability of drugs and supplies. They can directly experience a practitioner's technical competency and the service quality and can observe daily the cleanliness of the health facili- ties. Therefore we can expect greater economic and quality gains as community control and management increases. This relationship is illustrated in figure 3.2. The curve in figure 3.2 represents the hypothetical relationship between the gains and community control. Their relationship is nonlinear because we assume the community members have limited education, management know- how, and knowledge of medical affairs. At some point, the gains reach an asymp- totic point because of the community's limited ability to manage. Thus the best combination of control and management may be a combination of community, government, and health professionals. Experience from local community-controlled schemes and household inter- view surveys conducted in China and Indonesia support our hypothesis. As described in the second part of this chapter, one large Chinese survey conducted in five provinces found that more than two-thirds of the community residents want significant control over the CF schemes before they will enroll. In examin- ing the CF experience, we have consistently found that agricultural and lumber cooperatives managed schemes in Indonesia. Community-managed Dana Sehat and CMS have increased the availability and access of health care at the village and subdistrict levels. Moreover, the Indonesian experience also illustrates the limited managerial capability of the community members. The most successful Dana Sehat plans were the ones in which villages grouped themselves by subdis- trict. The board members were chosen by community members, but full-time qualified managers were hired to manage the plan at the subdistrict level. Experience of Community Health Financing in the Asian Region 127 FIGURE 3.2 Plausible Relationship between Locus of Control and Economic and Quality Gains Level of Competence Government Community Managing Managing Managing daily technical quality customer service village post operations Magnitude of service quality Magnitude of technical efficiency Government control Community control What Operational Characteristics of a CF Scheme Represent Managerial Efforts to Improve Efficiency and Quality? Four key factors emerged in our examination of which organizational and man- agement characteristics and incentives structures are more likely to yield efficiency and quality gains valued by community members. First, market surveys in China and Indonesia and household interview surveys consistently found that availabil- ity and proximity of reasonably competent practitioners matter the most to com- munity members. They have a paramount concern about the distance they have to travel for basic primary care and drugs. They also highly value home visits by health practitioners such as midwives. Second, studies found that economic gains can be produced by organizational arrangements such as integrating financing and provision of preventive and primary care at the village level,10 using an essen- tial drug list, centralizing the bulk purchase of drugs, and having an organized drug distribution network (Carrin and Vereecke 1992; Saurborne, Nougtara, and Latimer 1994). Quality gains can be obtained through organizational arrange- ments such as establishing a formal referral system, regular monitoring of clinical performance of practitioners at lower level facilities by higher level facilities, and cleanliness and hygiene at the clinic, particularly in the bathrooms. Efficiency and 128 Health Financing for Poor People: Resource Mobilization and Risk Sharing quality gains can also be obtained by using patient and provider incentives such as imposing a copayment on drugs to reduce moral hazard, allowing patients some choice to create competition, paying practitioners separately for health education and prevention to improve self-care and prevention, and paying salary plus bonus to practitioners to improve working hours and custom services (McPake, Hanson, and Mills 1992). Using the factors gathered from various CF experiences and marketing studies, we can classify and examine a CF by the local community's relative control and managerial power over it and its organizational, managerial, and incentive char- acteristics. These are important because they may indicate the relative gains that a CF can produce, gains valued by those who have to prepay. Once we can exam- ine the magnitude of the gains and the kind of gains a CF can produce, we can infer what a CF's likelihood of success is in raising funds from the community and the size of the population likely to enroll. In table 3.2, we summarize the major characteristics of CFs that enable them to produce significant economic and quality gains. It is important to stress that most of the cost reductions pro- duced by CFs came from purchasing drugs in bulk and using them more appro- priately (Carrin and Vereeke 1992; Saurborne, Nougtara, and Latimer 1994). When Households Are Willing to Prepay: Lessons from Specific Community-Financing Schemes in Asia By closely examining the CF schemes mentioned in this chapter, we determined that there are several types, based on the key factors that can explain the success or failure of the various schemes. We summarize them here. Direct government subsidy to individuals. The Thai Health Card represents this generic model. Its success seems to come from four factors: (1) patients have to pay high user fees unless they enroll, (2) the government directly and visibly matches the premium paid by the enrollee, (3) patients have free choice of public providers, and (4) most people can readily calculate that the benefits would exceed the pre- mium they pay. This model should not even be included as a community-financ- ing scheme since it has little grassroots community involvement. Cooperative health care. China's CMS and Dana Sehat represent this generic model. Financing and provision are integrated at the village and subdistrict levels. The original success of CMS was largely due to its extremely efficient and low-cost health care delivery system, which brought clear benefits to the enrollees and was also compulsory. Its major weakness was in public control and management. The local officials in many communities abused their power and misused CMS funds for their own gain. When the compulsory feature was removed after agriculture reform in the early 1980s, only the uncorrupted CMSs (between 50 and 60 percent of them) had any chance of surviving on a voluntary basis. The CMSs that contin- ued, despite the government's effort to abolish them in the mid-1980s, can be divided into two groups. The first were those found in better-off communities with enough middle-class rural households to be able to give some subsidy. The second Salaried Increased efficiency practitioners of drugs Incentives reduce quality Copayment to services overuse and Increased eficiency and Subsidy Greater participation Allow and choice competition Increased efficiency Features Operational hospitals Contracting Increased efficiency use drugs quality Organized purchase and of Increased efficiency and Schemes Organization by level system, quality Organized continuous education upper Increased referral monitoring by level Community-Financing of financing provision Integration of village Increased efficiency and at of Characteristics Control control Degree community Selected of Fund Health ypologyT A County Name Card Counties Counties 3.2 Community Jiangsu 14 10 ibetT Health Sehat ABLET Thai anzaniaT CMS-- CMS-- CMS-- CMS-- Dana 129 130 Health Financing for Poor People: Resource Mobilization and Risk Sharing were those found in poorer communities that had local government leaders who were enthusiastic about continuing with the scheme and were able to mobilize sub- sidies from either local government or rural enterprises. In addition, the CMSs in these poorer communities had to account to the enrollees for use of the fund and were able to integrate financing and provision at village and township levels to pro- duce efficient, quality health care. The experience of Dana Sehat was similar to that of CMS. Dana Sehat's slow expansion can be explained by two factors. First, the government gives no subsidy, and the poor and near-poor households could not pay. Dana Sehat succeeded largely in areas where a religious charity or rural indus- tries such as a lumber cooperative were able to provide some subsidy. Second, a vil- lage is too small a unit to support an entire scheme and to provide management know-how and risk pooling. Community-sponsored third-party insurance. A social experiment designed and operated by the Rand Corp. was conducted in China. The experiment aimed to assess how many people were willing to join a voluntary community-based health insurance plan, what premiums the Chinese peasants were willing to pay without a government subsidy, and the price elasticity of demand. The Rand experiment set the premiums at 1.5 percent of average income. The system was designed as an insurance plan, and the insured paid a significant copayment or coinsurance when they sought health care from village, township, and county health facilities. They could visit county hospitals only in an emergency or with the approval of the township health center. More than 90 percent of households in the test areas voluntarily joined the program, and 95 percent reenrolled after the first year. Administrative costs were kept low (8 percent of total reimbursements). Provider-sponsored prepayment plan. The Dhaka Community Hospital (DCH) Plan represents this generic model. The DCH system operates a health insurance scheme at its clinics, known as the Health Card Program. There are five types of health cards. The Family Health Card, intended for rural households, costs 40 taka per month (about US$1) for an initial enrollment and 20 taka for renewal, covering up to 12 members per household, including servants living there. This plan entitles the whole household to consult the clinic doctor at any time and to monthly home visits by DCH-trained health workers. Patients with the health card do not pay additional fees for consulting a doctor but have to buy medicines outside the clinic. The School Children Card, free to schoolchildren living near the health clinics, offers children free physical examinations and health educa- tion. The Worker Health Card, for workers in enterprises near the clinics, costs 2 taka per month per worker. Premiums are paid by the companies or the owners' associations. The benefit package includes free consultation but no monthly home visits by health workers. The Sports Card, for professional sports players, is intended mainly to publicize the clinics. No premium is charged for enrollment, and medical consultations are free. Poor families in the communities received a special Destitute Card at no cost, which allows members of poor households to visit the clinic at 5 taka per visit. The community committees decide which fam- ilies in the village qualify as "poor" and should receive Destitute Cards. Experience of Community Health Financing in the Asian Region 131 Consumer or producer cooperatives. The Grameen Bank (GB) represents this generic model. GB is internationally known for its successful group-based credit program and as a provider of credit to the rural poor, particularly women, who own less than a half acre of land or whose assets do not exceed the value of one acre of land. Grameen has 2.3 million members and, through its 1,167 branches, covers almost half of the villages in the country. GB established the Grameen Health Program (GHP) to provide basic health care services to its members as well as nonmembers living in the same operational area and to pro- vide insurance to cover the cost of basic care. The GHP functions as an insurer as well as a health care provider. The GHP's prepaid health insurance program is open to everyone covered by a GB branch regardless of whether they are GB members. The insurance scheme utilizes the organization structure of the GB credit program. At the GB branch operational level, which normally has 2,500 GB families and 3,500 non-GB families, GHP organizes health centers to provide outpatient ser- vices, routine pathological tests, and basic drugs. Each center is staffed with a doc- tor, who also acts as the center director, a paramedic, a lab technician, and an office manager. Some centers have subcenters, usually staffed by one paramedic and two health workers (Grameen Bank 1995). The number of Grameen health centers gradually expanded from 5 in 1993 to 10 in 1997. The families enrolled in the insurance increased from 13,000 in 1994 to 25,935 in 1996. About 85 percent of the subscribers are GB members. This ratio has not changed much over the years. In summary, we can classify community-financing schemes into five types: · Direct Subsidy to Individual Thai Health Card Tanzania Community Health Fund · Cooperative Health Care High-income communities: Jiangsu Province Middle-income communities: 14-county (WHO) Low-income communities: 30-county study 10-county experiment Tibet · Community-Based Third-Party Insurance Rand Experiment in Sichuan Province Dana Sehat · Provider-Sponsored Insurance Dkaha Community Hospital Gonoshasthaya Bwamanda · Producer or Consumer Cooperative Grameen 132 Health Financing for Poor People: Resource Mobilization and Risk Sharing General Findings The general findings can be summed up as follows: · Rural households and urban poor households are willing to prepay a portion of their health services. The amount they are willing to prepay depends on economic and social factors. The economic factors include the household's ability to pay, the size of out-of-pocket payments they have to make if not enrolled in a CF scheme, direct subsidy given, and who controls the funds and delivery of basic primary care services. The social factor includes the sense of kinship and mutual concern for each other in a community (social capital). · The poor and near poor need simple and direct government or donor subsi- dies to make the economic gains very visible to motivate them to prepay. The subsidy could be as low as half the prepayment amount to be paid by the peo- ple. Poverty households need almost full subsidy. · The revealed preference seems to show that people prefer to have both pri- mary care and insurance and are willing to make a tradeoff. Among the CF schemes in Asia, people most wanted coverage of primary care and drugs. This is logical because people want to have a direct payoff. Understanding and appreciation for risk pooling is rudimentary.11 Furthermore, greater coverage of primary care reduces adverse selection by including individuals who have no serious health problems, but they may drop out once they learn their pre- payment has little immediate direct payoff. · In most communities, a CF scheme must have its members' trust and confi- dence. This means the community must have reasonable control over the scheme and the services delivered at the village level. · CMS and many Dana Sehat plans have demonstrated that they can produce measurable economic gains and improvements in service quality. · The government or an established NGO must initiate the scheme and con- duct training. · Community-sponsored third-party insurance schemes have seldom suc- ceeded in covering a significant percentage of the target population. · Some nations rely on community cooperatives and have found that con- sumer cooperatives have done better than producer cooperatives in looking out for the patients' interests.12 Classify and Assess Community-Financing Schemes by Outcomes Another approach in analyzing and assessing CF schemes is by their impacts on the health system outcomes. Outcome data by community have seldom been collected. A few studies have found measurable improvements in health out- Experience of Community Health Financing in the Asian Region 133 FIGURE 3.3 The Trade-Offs between Health Gains and Risk Protection by Type of Service Funded Resources Allocation of resources with priority given to maximizing health outcomes Allocation of resources with priority given to maximizing financial risk protection Prevention Primary care Tertiary care come and financial risk protection after the introduction of certain types of CF at the regional or national level. The Chinese experience is a good example. In the absence of reliable data, we can examine the health services that are pro- vided under a scheme and infer the potential impacts on health and risk pool- ing. The relationship between covered health services and health outcomes and risk pooling are shown in figure 3.3. It also illustrates the trade-offs when resources are very limited. The most painful trade-off is between improvement of health status and prevention of impoverishment. The CF schemes can be compared by their likely impacts on health status and risk pooling. Table 3.3 summarizes the outcomes by which the schemes should be assessed; the use of this framework is illustrated with several selected CF schemes. A SUMMARY OF THE VALUE ADDED BY TYPES OF COMMUNITY-FINANCING SCHEMES For low-income countries, CF schemes have only modest ability to increase the total amount of funds for health care. The reason is straightforward. The target population consists of largely poor and low-income households whose ability to pay is modest. The major value added by such schemes is their organization of what households and government are already spending directly and their use of the funds to buy more and better services. Properly structured CF schemes can significantly improve efficiency, reduce the cost of health care, improve quality and health outcomes, and pool risks. risk risk pooling Greater protection Financial Improve quality health outcomes Schemes Better Improve access and cost Community-Financing efficiency of reduce Consumer satisfaction ypesT Increase Selected in by tyiuqE Equity financing Added alueV mobilize to funds Potential of Potential CF Card Assessment of Sehat CMS 3.3 Health Name Dana Thai ABLET 134 Experience of Community Health Financing in the Asian Region 135 Community-financing schemes could improve preventive services and reduce the incidence of disease. They could also improve people's access to health care and the quality of services, thus improving their health status. The schemes could improve risk pooling and reduce health-induced impoverishment as well. CF schemes can be grouped by their basic purposes: · Mobilizing additional funds for government facilities and improving access-- prepaid user fees (for example, Community Health Funds in Tanzania, Health Card in Thailand) · Mobilizing funds from rural population and urban poor and improving access, efficiency, and quality of care with modest risk pooling--cooperative healthcare (for example, CMS, Dana Sehat, Grameen Kalyan, Boboye, Abota) · Assuring more stable funding for providers--provider sponsored insurance (for example, Dhaka Hospital, Nkoranza, Bwamanda). We summarize the five types of CF schemes and their potential impact on the final outcomes of a health system. Table 3.4 gives our evaluation, based on stud- ies of a limited number of schemes. A REVIEW OF SELECTED ASIAN COMMUNITY-FINANCING SCHEMES Community financing is an ambiguous term that has been used loosely over the years to describe some level of community involvement in financing health care. Community involvement alone, however, cannot be used as the single dimensional factor for all the various schemes, including copayment for govern- ment services, prepayment schemes for hospital services, financing schemes for immunization, private insurance schemes, and even revolving drug funds. Community financing can be broadly defined as any scheme that has two fea- tures: a community base13 and prepayment into an identifiable fund by the com- munity members14 that entitles them to some health benefits. This definition would exclude some financing schemes such as regional social insurance plans and community-managed user fee programs. Using this definition, we selected several well-known national and local financing schemes for analysis. These case studies may shed light on why some schemes have succeeded and others have failed. National Schemes China's Cooperative Medical System In China, about 800 million people live in rural areas, most of them engaged in farming. Most of them were living on bare subsistence before the 1979 agricul- tural reform. The average disposable income per person was roughly equivalent to US$115 in 1985, ranging from under US$25 per capita for the poorest house- holds to US$175 (1,390 yuan) for those in the highest income quartile. unless pooling Greater risk Low Low Modest government subsidy High High Medium Low Low High Low Low High Improve quality for insured for insured Improve access Low Modest High High those High those High and cost None None High Low Low High Increase reduce efficiency Schemes in Low Low Low Low Low Low Equity financing pop. fund covered families families higher higher member Community-Financing raise be of Potential to and Low Modest High Cover income Cover income Cover ypesT by purpose fund households controls of Added community Who use special alueV Government Individual Local and NGO Community Hospitals Cooperatives Potential fees healthcare insurance consumer 3.4 CF user -sponsored or of - party ABLET ypeT third insurance cooperative Prepay Cooperative Community-based Provider Provider 136 Experience of Community Health Financing in the Asian Region 137 Beginning in the late 1950s, health care for China's rural population was orga- nized and financed through the Cooperative Medical System (CMS), an inte- grated part of the overall system of collective agriculture production and social services. Health care for the rural population was organized into a three-tier structure. Village health stations served an average of 500 to 1,000 residents; township health centers served 15,000 to 20,000 people; and county hospitals served a catchment area of 200,000 to 300,000 people. Village stations were staffed by part-time village doctors whose training consisted of three to six months' basic medical education after junior middle school. Their role was to provide basic preventive care (for example, immunization, prenatal consultation) and simple curative services (treating common illnesses and injuries). Most township health centers were owned and operated by the town government. An average facility had 7 to 10 beds and 10 staff members, led by a physician with a three-year med- ical school education after senior middle school. County hospitals, serving as medical referral centers for rural residents, were owned and operated by the county government and staffed by physicians with four to five years of medical school training. The typical county hospital had 135 beds and 186 staff mem- bers, of whom 8 percent had bachelor of medicine degrees. County hospitals had at least the five basic specialty services--obstetrics-gynecology, pediatrics, internal medicine, general surgery, and radiology. Under this system, village health stations, township health centers, and county hospitals were integrated within the three-tier system by a vertical administrative system. County hospi- tals and township health centers provided regular technical assistance and supervision to the lower level organizations. At CMS's peak, 90 percent of the rural population was covered by its schemes. Health services, financed through CMS, relied on prepayment plans. Most of the villages funded CMS from three sources: · Compulsory prepayment by residents. Depending on the benefit structure of the plan and the local community's economic status, 0.5 to 2 percent of a peasant family's annual income (4 to 8 yuan) was to be paid into the fund as premiums. · Village contributions. Each village contributed a certain portion of its income from collective agricultural production or rural enterprises to a welfare fund, and a portion of this fund was used to finance health care. · Government subsidies. Subsidies from higher level governments funded the compensation of health workers and capital investments. China achieved remarkable health improvements for its rural population before 1985. CMS was characterized by its collective financing, prepayment, and orga- nization of health services through the three-tier system. This community financing and organization model of health care was believed by many observers to have contributed in a significant way to China's success in accomplishing its 138 Health Financing for Poor People: Resource Mobilization and Risk Sharing "first health care revolution," by providing preventive and primary care to almost every Chinese and reducing infant mortality from about 200 per 1,000 live births (1949) to 47 per 1,000 live births (1973­75), increasing life expectancy from 35 years to about 65. In the early 1980s, when China's rural reforms decollectivized Chinese agri- cultural production, most CMS schemes collapsed. Most villages had to dissolve their CMSs after the main source of financing, the welfare fund, supported by collective farming income, disappeared. An ideological shift prompted some high government officials to declare that the remaining CMS programs should be abolished. Thus most communities that still had CMSs were forced to disband their systems by the mid-1980s, much to the dismay of local people. Last but not least, patronage, corruption, and poor management contributed to the downfall of CMS. CMS, though based in local communities, was controlled and managed by local officials who were not held accountable to the people. Some of these officials abused their power for selfish ends. As a result, people lost confidence in the government-run CMS program and refused to make financial contributions once the system became voluntary in the early 1980s. The health status of the rural population in China has deteriorated. This dete- rioration has been closely related to the collapse of CMS. For example, a World Bank study found that China's earlier progress in improving child health appears, in the aggregate, to have come to a stop, despite rapid economic growth since the early 1980s. The analysis found that mortality of children under five years of age (under-5) declined steadily until the early 1980s and then began a slight upward drift. Experiences from other countries suggest that the under-5 mortality rate need not reach a plateau, as China's has done. This indicates that China's performance has deteriorated not only in absolute terms but also rela- tive to other countries. The China Network­Harvard study of 30 poor counties confirmed the World Bank's study. China's poverty areas have experienced steady economic growth since the beginning of economic reform, in 1980; per capita gross domestic product, or GDP, (in real terms) increased from US$56 in the late 1970s to US$88 in the late 1980s. However, the median infant mortality rate in the surveyed counties increased from about 50 per 1,000 live births to 72 per 1,000 live births during the same period of time. Preventive services provided under CMS, financed out of central government and local welfare funds, were essentially free. Now, the vast majority of the rural population obtain their health services on a fee-for-service (FFS) basis instead of the previous prepaid basis. CMS Schemes that Survived: A Natural Experiment Despite the government's policy of abolishing CMS and the absence of central and provincial government support, some CMSs survived. That these were now voluntary schemes underscored the fact that they survived principally because peasants in those communities had chosen to continue the schemes. However, Experience of Community Health Financing in the Asian Region 139 these natural experiments have several biases. The peasants were already familiar with CMS; they knew the scheme and its possible benefits and drawbacks. It is important to bear in mind that this level of awareness makes the experiment unrepresentative. How Many CMSs Survived? A survey of five poor provinces in China, conducted in 1993 in preparation for the World Bank's loan for the Rural Health Workers Development Project, found that among the five provinces, one relatively poor province, Shanxi, had the greatest coverage--close to two-thirds of the villages maintained some form of community financing (see table 3.5). But in another very poor province, Guizhou, few villages had community financing (0.8 percent). In these poor provinces, the schemes were financed largely by household contributions, and benefits covered only primary care services because the very poor households could make only small contributions. The Study of Thirty Poor Counties was conducted in 1993­95 by a network of Chinese universities and Harvard University. The study found that 16.5 percent of the villages surveyed still maintained some type of community-based health- financing scheme, covering 11.6 percent of the sampled population (see table 3.6). About two-thirds of the schemes covered only primary care services at the village level; a third covered comprehensive services, ranging from primary care to inpatient services. The benefit structures all incorporated coinsurance and often set high copayment rates for inpatient services. The study found that the most prevalent type of community fund management was by village committee or by the village and township jointly (see table 3.7). TABLE 3.5 Prevalence and Benefits of Community Health Financing in 5 Provinces, 1991 (percent) Percent of Percent of Number of Percent of schemes with schemes with villages with villages with comprehensive primary care Province community financing community financing coverage services only Fujian 512 6.3 25 75 Guizhou 160 0.8 6 94 Hebei 3,992 13.1 42 58 Henan 1,590 6.2 7 93 Shanxi 4,727 65.6 15 85 Total 10,981 12.2 24 76 Note: Data from the Study of Thirty Poor Counties indicate that almost 80 percent of the category "services and drugs coverage" is comprehensive coverage. Therefore, for the purposes of this table, the Five Province Survey data were recategorized, with services and drugs plans counted as comprehensive coverage and the remaining categories (services only, drugs only, other) counted as primary care services coverage only. Source: World Bank (1993). 140 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 3.6 Prevalence and Benefits of Community Health Financing in 30 Poor Counties, 1993 Number of Percentage of Percentage of Type of benefits villages covered villages covered population covered Comprehensive 29 5.1 4.4 Primary care services only 59 11.4 7.2 Total 88 16.5 11.6 Note: Comprehensive benefits refers to schemes that reimburse 30­100 percent of hospitalization charges for township and county-level hospitals and 50­100 percent of outpatient fees. Primary care services refers to coverage of fees (or discounted prices) for most village-level services with fees at the township and county levels paid out-of-pocket by patients. Source: China Network and Harvard School of Public Health (1996). TABLE 3.7 Management of Community Health Financing in 30 Poor Counties, 1993 (percent) Type of benefits Form of management Comprehensive Primary care services only Township government 17.2 3.4 Township health center 20.7 6.8 Village and township jointly 20.7 10.2 Village committee 34.5 47.5 Village and township doctors 6.9 32.1 Source: China Network and Harvard School of Public Health (1996). A summary of results: Overall, about 13 percent of the rural villages continued with CMS. More affluent areas with rural industries and a higher tax base got financial support from both the government and rural industries. Why some poorest villages con- tinued with CMS remains a puzzle. Despite the government edict in the mid- 1980s to abolish the Cooperative Medical System, close to 10 percent of the rural villages continued to maintain their systems. This empirical evidence suggests that there is a significant amount of public support in many villages for estab- lishing community-financing schemes. However, voluntary development of CMS is very limited even in affluent areas. For example, only about 6 percent of the villages have established CMS in Fujian, a high-income province in China. In our 30-county poverty study survey, 70 percent of villages that reinstated CMS did so at the request of the government. Experience from China and other countries demonstrated that government could play a significant role as initiator and enabler of community financing. Experience of Community Health Financing in the Asian Region 141 Why Do People Prefer to Have CMS? Several large household surveys have found that a majority of Chinese peasants want cooperative financing established in their communities. The study of 30 poor counties involving 11,044 randomly selected households found strong sup- port for reestablishing community-financed health care. Of the households not covered by community financing or maternal and child immunization prepay- ment schemes, 70 percent responded that they would like to see an improved scheme established, similar to the Cooperative Medical System. Of those covered by community-financing schemes, 88 percent stated that they would like those schemes to continue. Several possible reasons could explain why people wish to maintain the CF schemes. We briefly enumerate the major ones. Financial barriers reduce access to primary health care for rural populations. Another reason is physical supply especially since 80 percent of the rural poor live in mountainous areas. The number of medical personnel, including barefoot doctors and township and village health workers, have been reduced signifi- cantly. The 30-county study also showed that the number of villages with func- tioning health stations dropped from 71 percent in 1979 to 55 percent in 1993. Between 1983 and 1993, the costs of health care also rose at an average of 10 to 15 yuan per year, twice the average growth rate of farmers' disposable income in China. The change to a fee-for-service system brought several changes, which fac- tored heavily into increased costs, limiting access in the process. Under the cur- rent fee-for-service system, the income of village doctors and health facilities depends on the profits they can earn on drugs. The government sets very low prices for services; in compensation, practitioners are allowed to mark-up the wholesale price of drugs by 15 to 20 percent, which provides an incentive to overprescribe drugs. Moreover, user fees charged for previously free preventive services have had detrimental effects on public health through reduced demand for, and supply of, preventive services. The Epidemic Prevention Service workers shifted their attention to services for which high fees could easily be charged-- such as cosmetic product inspection and food safety--which were not necessar- ily the highest priority or the most cost-effective activities. The average charge per outpatient visit for uninsured patients is almost three times that of the patients under CMS. Community-financing schemes can exer- cise their bargaining power in demanding discounted prices or providers can be paid on a partial capitation basis. The increase in cost from the new fee-for- service system created a financial access barrier for people. In the 30-county study, 28 percent of seriously ill farmers did not seek health care, and 51 percent of rural patients refused hospitalization, mainly for financial reasons. The finan- cial burden can be illustrated by one fact. A poor farmer would have to spend 1.2 years of his disposable income to pay for one episode of hospitalization at a county hospital. 142 Health Financing for Poor People: Resource Mobilization and Risk Sharing The higher cost of health services and drugs impoverished many families. Eighteen percent of the households using health services incurred health expen- ditures that exceeded their total household income in 1993. Of the households interviewed, 24.5 percent borrowed or became indebted to pay for health expenses. Another 5.5 percent sold or mortgaged properties to pay for health care. High health expenses are a major cause of poverty in rural areas. In our 30- county survey, 47 percent of the medically indebted households reported having suffered from hunger. This interaction between health and income could start a vicious cycle of illness, poverty, and more illness. Quality of care also suffered. The interconnection and cooperation among dif- ferent rural health facilities weakened or disappeared after reforms. Under CMS, county hospitals and township health centers provided regular technical assis- tance and supervision to the lower level organizations, with a referral system that managed patients at a lower level when possible. After the collapse of CMS, however, these health care organizations became independent institutions, often competing for patients to increase revenue. This disintegration of the three-tier system may also have implications for the quality of services provided by uncoordinated rural health workers. In addition, the collapse of CMS led to the overprescribing of drugs and the overuse of injections and profitable tests. Who Should Control and Manage CMS? There was a clear preference for community control in the management of CMS. Of the households without coverage that favored reestablishing community- financing schemes, about 62 percent of the peasants wanted the village residents to have a strong voice. Only 17 percent trusted the government to manage it independently. About one-fourth preferred that the scheme be managed by the village, one-fourth preferred township management, and the rest preferred joint management by the township or village and the health facility. How Much Are People Willing to Pay and for What? Responses to the household surveys indicated a wide range of preferences for ser- vices, from drugs and village doctors to township health centers. There was also a preference for coverage of hospital inpatient services (catastrophic expenses), with a willingness to accept coinsurance. World Health Organization study. To learn how to improve organization, financing, and service delivery, the World Health Organization studied commu- nity-financing schemes in 14 counties in Beijing, Henan, Hubei, Jiangsu, Jiangxi, Ningxia, and Zhejiang. In each county, a research team interviewed 540 house- holds and surveyed health services. The study found that a typical community fund might collect 5 yuan (US$0.63) per person from families, 1 yuan per person from the village's social welfare fund, and 1 yuan per person from the township. Patients typically must pay a deductible (for example, 100 yuan) and make a copayment on expendi- Experience of Community Health Financing in the Asian Region 143 tures above the deductible. The schemes limit drug coverage to 120 kinds of medicines, including traditional Chinese medicines, and set a limit on reim- bursement for diagnostic tests. Rand experiment. In the Rand experiment, premiums were set at 1.5 percent of average income. Insured individuals could freely visit village and township facil- ities but could visit county hospitals only in an emergency or with the approval of the township health center. More than 90 percent of households in the test areas voluntarily joined the program, and 95 percent reenrolled after the first year. Administrative costs were kept low (8 percent of total reimbursements). The study also found that · Coinsurance (or copayment) exerted a significant negative effect on demand for care across different population groups. There were no interactions between the effect of coinsurance and age, income, or health status. · Users in all but one village reported high satisfaction with the insurance arrangement (Mao 1995, p. 16). · Services were used less when there was no functioning village health station, underlining the importance of an adequate supply of basic services (Mao 1995). As in other countries, a small share of the population (about 11.5 percent) accounted for a large share of total health expenditures (70 percent), underscor- ing the need for catastrophic insurance. In most rural areas, particularly poor areas, adequate revenues for any orga- nized financing scheme cannot be derived solely from households. Funding must come from multiple sources. According to the Study of Thirty Poor Counties, about half the financing for existing community-financed health plans came from household contributions, about 20 percent from village social welfare funds, and about 16 percent from the government (see table 3.8). Affordability of Hypothetic Basic Benefit Packages Whether community financing is feasible for rural China will depend first of all on its benefit structure: a low benefit package (for example, covering only cost- effective preventive services) is affordable but may not meet the rural population's TABLE 3.8 Community Health Financing by Source in Selected Counties and Provinces, 1991 and 1993 (percent) Government Village social welfare fund Households Other Funds surveyed in Study of Thirty Poor Counties (1993) 16.1 20.3 48.1 15.5 Funds surveyed in Five Province Survey (1991) 8.0 30.3 58.7 3.0 Source: China Network and Harvard School of Public Health (1996); World Bank (1993). 144 Health Financing for Poor People: Resource Mobilization and Risk Sharing need for protection from catastrophic medical expenses. Yet a high benefit pack- age, though desirable, may not be feasible because people's willingness and abil- ity to pay is limited. Illnesses are uncertain, and thus the payoff from participating in community-financing schemes is also uncertain for the house- holds. We found that about 11 percent of the rural population consumed 70 per- cent of the total medical expenditure. This finding illustrates the need for catastrophic insurance and the potential problems of adverse selection and risk selection under a voluntary insurance program. The core issue in designing an appropriate and feasible basic benefit package is the balance among three con- siderations: the cost-effectiveness of the services covered, people's desired cover- age, and the financial constraints on those paying for the coverage. For illustrative purposes, we developed several basic benefit packages for the low-income rural population, based on data from the 30-county poverty survey. We used the following principles and assumptions in designing the benefit pack- ages: (a) first cover the most cost-effective services, but take into account the fact that health care delivery is not organized by disease; (b) coinsurance should vary for different services depending on demand elasticity; and (c) people are risk- averse and demand coverage for catastrophic expenses. From a societal perspec- tive, the coverage of catastrophic medical expenses also reduces the poverty rate. The simulation results are shown in table 3.9. Depending on the coinsurance level, the estimated per capita cost is between 28 yuan and 31 yuan (about US$4 to US$5) to provide a basic benefit package that includes specified maternal and child health care services and a stop-loss provision of about 500 yuan to 600 yuan for patients in 1993. TABLE 3.9 Two Prototype Benefit Packages for China's Rural Poor (benefit structure and costs) Type of expenses covered Level of coinsurance High Low Village post-- service fees 20% 10% drug expenses 50% 40% Township health center-- outpatient service fees 30% 30% outpatient drug expenses 50% 40% inpatient 35% 30% County hospital-- outpatient service fees 40% 35% outpatient drug expenses 50% 40% inpatient 45% 35% Catastrophic protection-- cap patient's payment at 600 yuan 500 yuan Estimated per capita costs 31 yuan 28 yuan Experience of Community Health Financing in the Asian Region 145 TABLE 3.10 Current Financing of Health Spending by Source in China's Poverty Regions Source of payment Expenditure per capita in 1993 (yuan) Percent Household 23.41 59.32 Government for public employees 13.18 33.40 Community financing 2.31 5.85 Welfare fund 0.31 0.80 Other 0.25 0.63 Total 39.46 100.00 Can the low-income population afford these basic benefit packages? In poverty areas, households already spend a significant amount of their income on health care. According to our survey, annual medical expenditures by such households were 23 yuan per capita in 1993 (see table 3.10). However, it would be unrealistic to expect that people are willing to prepay this amount to support an organized financing scheme. Although the majority of individuals surveyed expressed their support for CMS, their willingness to prepay into the system was only about 5 yuan per capita. Per capita contributions to existing community- financing schemes range from 1.05 yuan to 6.14 yuan. With effective social mar- keting, it might be expected that 10 yuan per capita could be obtained from households. Therefore potentially 12 yuan could be collected from individuals and local communities, covering less than half the expected costs of a comprehensive package. The rest of the resource gap would have to be filled by public assistance. Without government support, community-financing schemes in the poverty regions can finance only a limited package for low-income households. Why Did Most Communities Not Have a CMS? When asked about the major reasons for the lack of community-financing ini- tiatives, 53 percent of the community leaders cited financial difficulties. How- ever, about two-fifths of the interviewed leaders listed nonfinancial reasons, such as lack of organizational capacity and lack of policy support from higher level government (see table 3.11). TABLE 3.11 Percentage of 2,236 Surveyed Community Leaders Citing Major Reasons for Lack of Rural Community Financing Inadequate organizational capacity 22 percent Inadequate policy support 12 percent No mass support 8 percent Inadequate financial resources 53 percent 146 Health Financing for Poor People: Resource Mobilization and Risk Sharing Indonesia's Dana Sehat and Health Card Schemes Indonesia has the world's fourth largest population, around 210 million people, living on 5 major islands and 30 groups of small islands. In 1990, the urban-to- rural ratio was 30:70, placing the number of people living in the rural areas of Indonesia close to 145 million. In 1996, GDP per capita reached US$1,155; how- ever, after the economic crisis of 1997, income per capita fell to US$380, impair- ing the people's ability to pay basic needs, including health services, family planning services, and food. Since the early 1970s, the Ministry of Public Health (MPH) has encouraged the Dana Sehat (the Village Health Fund) program. The objective of this program has been to improve the coverage of health services in Indonesia by accelerating community participation in financing and maintaining its own health. In 2000, the total membership of Dana Sehat was 23 million people, about 11 percent of the total population. The Dana Sehat, a voluntary community-based, prepaid health care program is most common in the rural areas of Indonesia. The prime movers are health centers, local government, and NGOs such as cooperatives and pesantrens (Muslim teaching units). The people covered are primarily farm- ers, fishermen, and students. Contributions come from local economic activities; some schemes generate funds through co-ops of crops or handicrafts, while others are paid in cash. Every household is obliged to pay the premium either in-kind or in cash to the bank or the committee of the Dana Sehat. Open management, trust, and com- munity leadership form the basic culture of the Dana Sehat. Vision, mission, objectives, and program identification are based on deliberation and agreement among community members. Community control comes primarily from its members through periodic meetings to discuss the program, for which the gov- ernment provides tools and guidance such as Dana Sehat operation, monitoring, and supervisory procedures. There are many levels of Dana Sehat. On the smallest scale, the Dana Sehat operates with simple management organized by the village and often run by var- ious local institutions. Membership is between 50 and 499 households, the pre- mium is relatively low, and the highest contribution is Rp 100 per household per month. The health benefits are limited to Health Center Services because of the low contribution of the community. On a larger scale, Dana Sehat has a larger membership and is run by a consolidated organization, organized by several vil- lages (that is, a township). It has an organizational structure and job description for the providers. The premium is about Rp 500 per household per month. At this level, Dana Sehat can provide more health benefits, including primary care as well as inpatient services. Payments to the providers are based on the number of consultations (Rp 700 per consultation). Government officials usually initiate the process by organizing meetings between health providers, local authorities, religious organizations, and key per- sons in the community and also community-wide meetings. For community Experience of Community Health Financing in the Asian Region 147 surveys, government officials help to train surveyors and assist with analysis and presentation of the results to the community. At a second round of community meetings facilitated by government authorities, community members choose the services to be covered by the fund, balancing their needs against their ability and willingness to finance the package. Poor rural communities often choose a package of basic outpatient and curative care, combined with free preventive ser- vices. Continued government supervision and monitoring, combined with monthly or bimonthly community meetings, guide fund management. Indonesia Health Card The Health Card program, begun in 1996, is a national commitment to assure health care access for the indigent. The target population is eligible people in vil- lages. In 1999, the estimated number of participants was 11,096. The Health Card program is administered by the Municipality Health Office (MHO). The premium is fully subsidized; the beneficiaries do not have to pay any premiums or fees. Ser- vices at the health center and certain services in public hospitals are covered. Payment to the providers is Rp 10,000 per year by capitation of the designated poor. The basic funding comes from the Social Protection Sector Development Program (SPSDP) and MHO: 580,000,000 per year and public hospitals 600,000,000 per year. The government seems to be in charge of this program. Several factors explain Dana Sehat's success. The community is heavily involved in the supervision and monitoring of health care activities of the Dana Sehat. However, at the village level, there are insufficient management capabili- ties, risk-pooling capacity, and ability to monitor the technical quality of ser- vices. As a result, many village-managed Dana Sehat rely more on fee-for-service than on prepayment. Dana Sehat performs better at the township level. At this level, the manager- ial capacity improves and the larger population base allows greater risk-pooling. Centralized management has more resources and can offer more benefits and more providers. One major barrier to establishing Dana Sehat is the absence of government subsidy. The poor community simply cannot afford the scheme. Only commu- nities with rural enterprises or producer cooperatives have the potential to obtain funds to subsidize the very poor and perhaps the poor and near-poor as well to induce them to join. Thailand's Health Card The population of Thailand is estimated at 59 million, of whom 31.5 percent are urban, hence the ratio of urban to rural population is about 1:2. Income distri- bution has worsened; in 1992, the highest income quintile held 59.5 percent of the national income while the lowest income quintile held merely 3.8 percent. The 1994 per capita income was US$2,410. 148 Health Financing for Poor People: Resource Mobilization and Risk Sharing The Health Card program was implemented in 1983 as a voluntary scheme, pri- marily to promote maternal and child health. Purchase of the card meant prepay- ment of a certain fixed premium, capitation to the provider, in return for free services for one year. Proceeds from the card sale went into the health card fund and was managed by a village committee. The program's primary objective at inception was to improve health among rural populations, with an emphasis on primary health care, including health education, environmental health, maternal and child health, and provision of essential drugs. The system incorporated a referral system from primary care to tertiary care. The program was also intended to involve local villagers in self-help as well as in managing the health card fund. As time went by, various cards were introduced for different purposes, but by 1991 they had been discontinued, and only family cards priced at 500 Baht were offered. A major change in the program since 1994 is the explicit contribution of the MPH: an equal contribution of 500 Baht. In addition, no limits were imposed on the number of episodes or the cost of coverage per visit. Flexibility was built into the referral system, and each province could impose any conditions deemed appropriate for particular situations. Health care funds also became managed by a committee at the district level in coordination with village-level bodies, a move that was aimed at expanding the enrollment base to the district level. As for the share of funds, 80 percent of the card price was earmarked for providers for med- ical care, and the remaining 20 percent was to be retained for marketing and sales incentives. The health card project has evolved over the years and can now be considered a kind of social welfare program, since it now receives an explicit con- tribution from the government equal to the contribution of the cardholder. In 1994, free health cards were given to community leaders and village health volunteers to provide free health care for their families. The voluntary cardhold- ers consumed more health care than other types of cardholders. The compulsory community leader cards and health volunteer cards provided better risk pooling and compensated for the deficit on operating the voluntary health cards. Con- sidering costs per card in relation to population coverage, provinces with low coverage of health cards were more likely to face higher utilization rates and health expenditure per card than provinces with high population coverage. Therefore, the health card fund provided on average only a 50 percent subsidy to the regional and general hospitals while providing 80 percent subsidy to the community hospitals and the full cost to health centers. Initially, individuals with a monthly income below 1,000 Baht were eligible. Now, health card eligibility extends to families with monthly incomes lower than 2,800 Baht per month and individuals with monthly incomes below 2,000 Baht, primarily farmers and informal sector workers at the community level. The cards entitle holders to free medical care at all government health facilities oper- ated by the Ministry of Public Health, the Bangkok Metropolitan Administra- tion, the Red Cross Society, Pattaya City, and the municipalities. Each card is valid for three years. The government provides block grants to health facilities based on the expected distribution of the eligible population and past records. Experience of Community Health Financing in the Asian Region 149 Health centers are the cheapest source of health care, but inpatient care is available at community, general, and regional hospitals. Specifically, the budget allocation is based on the number of low-income people living in the designated less-developed villages. However, the budgetary allocation is invariably insuffi- cient to cover the cost of providing services. The health card fund has been changed so it can be managed like a revolv- ing fund. In 1995, the Ministry of Finance set up an accounting system for the central and provincial health card funds that complies with the regulations of the government's revolving fund. Risk pooling at the central level facilitates portability of benefits and risk sharing among provincial funds, allowing for cross-boundary services used in different provinces and high-cost services within the same or different provinces. Now there are no benefit limits, and the program is targeted to the subgroup of the population with no health benefit coverage. Those factors explain the success of the Thai Health Card: The health card is able to cover a large part of its target population, the near poor, because the gov- ernment's subsidy is given directly to the eligible household. The subsidy is rela- tively large--100 percent matching for what the household pays. Currently, it covers about 3 million people. The enrollees have unlimited access to free services at the public facilities. Thus it also offers a high degree of risk pooling. Local or Target Population Schemes Bangladesh Community-Financing Schemes In Bangladesh, community participation in health care is increasing.15 NGOs and the private sector actively organize and finance health care delivery. Com- munity health insurance schemes have emerged as a mechanism for paying providers and mobilizing resources. Examples in Bangladesh include the Grameen Health Program (GHP), the Gonoshasthaya Kendra (GK) Health Care System, and the Dhaka Community Hospital Insurance Program. This section briefly describes the three schemes. The Grameen Health Program The Grameen Bank, internationally known as a successful group-based credit program, provides credit to the rural poor, particularly women, who own less than a half acre of land or whose assets do not exceed the value of one acre of land. At present, Grameen has 2.3 million members and covers almost half the villages in the country through 1,167 branches. The GB is an institution for financial intermediation, and it also supports social development pro- grams for poverty alleviation. The Grameen has more than 20 years of opera- tional experience as a financial intermediary. It started a health program only in 1993. 150 Health Financing for Poor People: Resource Mobilization and Risk Sharing GB became involved in health care mainly because illness was identified as the single largest cause of loan default. A study found that 44 percent of the loan defaults were due to illness. Poor health prevented borrowers from carrying out their economic activities and therefore from repaying their loans because of lim- ited access to health care and limited capability to pay for health services when health needs occur. Thus illness and its financial consequences are serious threats to the Grameen borrowers and the long-term viability of the GB itself. The Grameen Health Program (GHP) was established to provide basic health care services to its members as well as nonmembers living in the same operational area and to provide insurance to cover the cost of basic care. The GHP thus func- tions as an insurer as well as a health care provider. The GHP started with five Grameen Health Centers in 1993. Each center is staffed with a doctor, who also acts as the center director, and with a paramedic, a lab technician, and an office manager. Each center is attached to a Grameen Bank branch and covers the branch's operational area, which normally has 2,500 GB families and 3,500 non-GB families. Some centers have subcenters, usually staffed with one paramedic and two health workers (Grameen Bank 1995). A health center provides outpatient services, routine pathology, and basic drugs. A television is provided in the waiting room and shows various health education programs. Health workers provide door-to-door services on health education and health promotion. The GHP's prepaid health insurance program is open to everyone covered by a GB branch, regardless of whether they are GB members. The insurance scheme utilizes the organizational structure of the GB credit program. The subscription of the insurance scheme for GB members is based on groups of a minimum of five families, the same requirement as for borrowing. In general, GB members partici- pate in the health insurance program in the same groups as their loan groups. The GHP charges an annual premium of 120 taka (about US$2.60) per family for GB members and 150 taka (US$3.30) for non-GB members. However, if non-GB mem- bers can organize into a group of five, they pay the same rate as members. The benefit package covers unlimited outpatient visits with a copayment of 2 taka (about US$0.05) per episode; 50 percent of the cost of basic pathology tests and 15 essential drugs; 15 percent of other drugs sold at the health center; 50 percent of the cost of a specialist consultation and more sophisticated tests in referral hospi- tals; and reimbursement of up to 500 to 1,000 taka per year for hospitalization. The number of Grameen health centers grew from 5 in 1993 to 10 in 1997. The number of insured families increased from 13,000 in 1994 to 25,935 in 1996. About 85 percent of the subscribers are GB members. This ratio has changed little over the years. The Gonoshasthaya Kendra Health Care System The Gonoshasthaya Kendra (GK) Health Care System is an NGO-run local health care system. It operates in Savar, a rapidly industrializing area with a population of 271,448, 40 kilometers from Dhaka, the capital city of Bangladesh. GK started as a health project in 1971 with donor support and gradually expanded into a Experience of Community Health Financing in the Asian Region 151 two-tier health care system with a 70-bed hospital and 4 subcenters. Each sub- center covers 25,000 to 30,000 inhabitants with a team of 8 to 10 paramedics. A paramedic usually provides door-to-door services, including preventive and sim- ple curative care and health education, to inhabitants under his or her coverage, usually 600 to 700 families. A doctor from the GK hospital visits a subcenter twice a week to see patients referred by the paramedics, and the severely ill are referred to the hospital. The GK system also runs other programs: a pharmaceu- tical factory, a workshop producing furniture, and a small credit program. The GK initiated the first community insurance scheme in 1975 to increase the poor's access to health care. GK subscribers pay a premium for a package of benefits, including primary health care and some portion of hospital care. The GK scheme classifies the population into four socioeconomic groups, and the premium and copayment schedules are set based on these groups. The crite- ria for determining households' socioeconomic status are not accurate measure- ments of income, but communities' perceptions of poverty. The communities in the catchment areas participated in the exercises for defining socioeconomic groups, and the final classifications were generally accepted by the population in the communities. The sliding fee structure was designed to reflect ability to pay. A household subscribing to the insurance scheme receives a registration card that indicates the household's socioeconomic group. The premium and copay- ment are charged according to the fee structure for that group. People who are not insured would pay fees based on market prices. The sliding premium rates are differentiated according to the four socioeco- nomic status groups in the area. Group 1 encompasses destitute single-headed households (most of them widowed or divorced women) and the disabled. Group 2 consists of households that cannot afford two meals a day for all household members, landless farmers (less than an acre of land), and daily wage earners. Group 3 includes households that can afford minimum needs but have no savings, such as farmers with small landholdings (two to three acres), small shop owners, and industry labor workers. Group 4 covers households with savings, farmers with more than three acres of land, owners of big shops or businesses, middle- and upper-class civil servants, and professionals (Desmet and Chowdhury 1996). The GK insurance scheme covers 12,393 member families, about 33 percent of the target population. Dhaka Community Hospital Health Insurance Program Dhaka Community Hospital (DCH) offers an innovative approach in health ser- vice provision. The DCH differs from other private hospitals in its mission and setup. A group of devoted senior medical practitioners organized a nonprofit trust in 1989, which later developed into a system with a 24-bed referral hospi- tal, 19 rural health clinics, 12 school health clinics, and 24 industrial health clin- ics. Their mission is to provide quality health care services at low cost, so most of the poor people can afford them. In its system, DCH attempts to integrate the provision of primary, secondary, and tertiary care. 152 Health Financing for Poor People: Resource Mobilization and Risk Sharing The DCH itself operates on a fee-for-service basis and provides walk-in ser- vices at fixed rates lower than those of equivalent private, for-profit hospitals. The DCH offers a special program called After Payment, which allows patients who cannot pay at the time of treatment to pay the medical bill in installments after treatment. The patient's community guarantees payment. The After Pay- ment Program is very small, only two to three cases per clinic per year. Since communities take the responsibility for making sure the fees are paid, no default has thus far occurred. The DCH is self-reliant and receives no funds from the government or donors. At its clinics, the DCH system operates as a health insurance scheme, known as a health card program. There are five types of health cards. The Family Health Card, intended for rural households, costs 40 taka per month (about US$1) for an initial enrollment and 20 taka for renewal, and covers up to 12 members per household, including servants living there. This plan entitles the whole household to consult the clinic doctor at any time and to receive monthly home visits by health workers who are trained by the DCH. Patients with the health card do not pay additional fees for consulting a doctor but have to buy medicines outside of the clinic. The School Children Card, free to schoolchildren living near the health clinics, offers children free physical exam- inations and health education. The Worker Health Card, for workers at enter- prises near the clinics, costs 2 taka per month per worker. Premiums are paid by the companies or the owners' associations. The benefit package includes free consultation but no monthly home visits by health workers. The Sports Card, for professional sports players, is intended mainly as a means of publicizing the clinics. No premium is charged for enrollment, and medical consultations are free. Poor families in the communities receive a special Destitute Card at no cost, which allows household members to visit the clinic at 5 taka per visit. The community committees decide which families in the village are considered poor and should receive Destitute Cards. Some services are provided by field health workers: preventive care, health education, and simple checkups such as measuring blood pressure and urine sugar levels. Specialists from the DCH visit clinics periodically to treat villagers. The rural health card scheme does not cover inpatient care or the costs of drugs or medical tests. Doctors in the clinics refer patients to the DCH, and patients pay for hospital care at DCH rates. The DCH has begun to provide health cards covering inpatient care at the DCH to some companies in urban areas. As presented above, many communities in Asia have long-established community-based financing schemes for health care. Their sponsors vary, rang- ing from the government to the community itself, when it perceived the need for an organized way to finance and provide basic health care. These varied schemes offer different benefits and cover different populations; some are more affordable than others. After first classifying the different types of community- financing schemes, then evaluating them, we can discover the major factors that determine the success or failure of each type. Experience of Community Health Financing in the Asian Region 153 Acknowledgments: The authors are grateful to the World Health Organization (WHO) for having provided an opportunity to contribute to the work of the Commission on Macroeco- nomics and Health and to the World Bank for publishing the material in this chapter as an HNP Discussion Paper. NOTES 1. This chapter was prepared as a part of the Commission on Macroeconomics and Health's Working Group 3 studies on mobilizing resources for households not employed in the formal sector (hereafter referred to as the "informal sector"). To state the obvious, they are not a homogenous group. Their occupations range from peasant, peddler, day laborer, taxi driver, and informal sector employee to shop owner and self-employed pro- fessional, such as a physician or a lawyer. Some of them are rich, but most are poor. Some live in the city, but most live in rural communities. This chapter focuses on mobilizing resources for the residents of rural communities, who make up more than 70 percent and 50 percent, respectively, of the population in low- and middle-income nations. The chapter also gives some attention to mobilizing resources for the urban poor. 2. Close to 50 percent of the total national health expenditure for most low-income nations comes from direct out-of-pocket payment by patients (WHO 2000). 3. The industrial nations (except the United States) use general revenue or compulsory social insurance to pay for health care for citizens working in the informal sector. 4. Funding maybe inadequate, but public funds usually go first to pay health workers, regardless of whether they deliver satisfactory services and whether drugs and other supplies are adequate. This practice has created a public employment program, not a health delivery program to meet patients' needs and demands. 5. There are exceptions, situations in which private nonprofit providers charge reason- able prices and deliver quality primary health care, such as the PROSALUD does in Boliva. 6. This confusion is exacerbated by studies that examine the community-financing schemes from a particular point of interest and then label them by that single unitary factor. Several widely circulated documents have used new terms such as rural health insurance (U.S. Agency for International Development and the World Health Organi- zation) and microinsurance (International Labour Organization) for community- financing schemes. 7. Community is defined as a group of households living in close proximity to each other, such as a village or a neighborhood. Often for risk pooling and managerial purposes, the villages might be grouped. In addition to geographic proximity, a community must include organizations in which people who share common interests come together, as in producer and consumer cooperatives or women's banks. 8. Prepayment can be for two types of health expenses: high cost and low frequency; and low cost and high frequency. The former involves much greater risk pooling (insurance) than the latter. Insurance literature has long documented that most peo- ple lack the appreciation for the benefit of insurance, a fact that led to a common say- ing: "Insurance is sold not bought." In advanced economies, voluntary private health insurance is being sold to the affluent risk-averse households, but this is not the case in low-income countries. 154 Health Financing for Poor People: Resource Mobilization and Risk Sharing 9. Affordable is defined as reasonably priced as judged by historical precedents or by common sense. 10. The currently favored policy of separating financing from provision can work only when there are several competing providers. This condition seldom exists in villages and towns. In the United States, the major industry in many isolated towns organized the staff-model health maintenance organization (HMO) in which financing and pro- vision are integrated. They have proven that this organizational form, controlled pri- vately, can produce high quality health services at lower cost. Low-income countries have had similar experiences. More important, low-income countries have few quali- fied practitioners working in villages and towns. Where public health services are absent or operate inefficiently, peasants rely on indigenous doctors, drug peddlers, or private practitioners whose competence varies widely but who charge high prices. The local community can improve the efficiency and quality of basic health care by orga- nizing health posts and clinics and by recruiting qualified health workers and practi- tioners and assuring them that they will have a steady, reasonable income. This requires, however, a prepayment financing arrangement. 11. Even in advanced economies in which people are better educated and buy other types of insurance, insurance companies initially found little willingness to buy health insurance. From the 1930s to the 1960s, it was often said that health insurance has to be sold, not bought. 12. In several countries, particularly in Africa, church-sponsored and -managed hospitals have enjoyed the people's confidence and have been successful in starting hospital- based prepayment plans. 13. Community is defined as a group of households living in close proximity to each other, such as a village or a neighborhood. Often for risk pooling and managerial purposes, the villages might be grouped. A community can also be a group of people formally organized to advance some common interest (for example, agricultural and consumer cooperatives). 14. Prepayment can be for two types of health expenses: high cost and low frequency; and low cost and high frequency. The former involves much greater risk pooling (insurance) than the latter. In advanced economies, voluntary private health insur- ance is being sold to affluent risk-averse households, but this is not the case in low- income countries. 15. This section is taken largely from the paper by Shiyan Chao (1998), Community Health Insurance in Bangladesh: A Viable Option? REFERENCES Atim, C. 1998. Contribution of Mutual Health to Financing, Delivery, and Access to Health Care: Synthesis of Research in Nine West and Central African Countries. Technical Report 18. Partnerships for Health Reform Project; Abt Associates Inc., Bethesda, Md. Bennett S., A. Creese, and R. Monasch. 1998. Health Insurance Schemes for People Outside Formal Sector Employment. ARA Paper 16. WHO, Geneva. Bitran, R. 1995. "Efficiency and Quality in the Public and Private Sectors in Senegal." Health Policy and Planning 10 (3): 271­83. Experience of Community Health Financing in the Asian Region 155 Carrin, G., and M. Vereecke. 1992. Strategies for Health Care Finance in Developing Countries with a Focus on Community Financing in Sub-Saharan Africa. New York: St. Martin's Press. Chao, Shiyan. 1998. Community Health Insurance in Bangladesh: A Viable Option? Unpub- lished report. World Bank. China Network and Harvard School of Public Health. 1996. Study of 30 Poor Counties. Gilson, L. 1995. "Management and Health Care Reform in Sub-Saharan Africa." Social Sci- ence and Medicine 40(5): 695­710. Gilson, L., P. D. Sen, S. Mohammed, and P. Mujinja. 1994. "The Potential of Health Sector Non-Governmental Organizations: Policy Options." Health Policy and Planning 9(1): 14­24. Grameen Bank. 1995. Preparatory Report on the Establishment of the Grameen Health Program. Dhaka. Hsiao, W. C. 1995. "The Chinese Health Care System: Lessons for Other Nations." Social Science and Medicine 41(8): 1047­55. Liu, Y. 2001. Demand for Community Pre-Payment Schemes in Health Care. Unpublished. Mao, Z. 1995. "Findings from the Rand Experiment in China. Chinese Health Economics Journal 11(3): 16. McPake, B., K. Hanson, and A. Mills. 1992. Experience to Date of Implementing the Bamako Initiative: A Review and Five Country Case Studies. London: London School of Hygiene and Tropical Medicine. Narayan, D., and L. Pritchett. 1997. Cents and Sociability: Household Income and Social Cap- ital in Rural Tanzania. Washington, D.C.: World Bank. Putnam, R. 1993. Making Democracy Work: Civic Traditions in Modern Italy. Princeton, N.J.: Princeton University Press. Saurborne, R., A. Nougtara, and E. Latimer. 1994. "The Elasticity of Demand for Health Care in Burkina Faso: Differences across Age and Income." Health Policy and Planning 9: 186­92. Stinson, W. 1982. Community Financing of Primary Health Care. Washington, D.C.: Ameri- can Public Health Association. World Bank. 1993. World Development Report 1993: Investing in Health. Washington, D.C. ------. 2001. World Development Report 2000­2001: Attacking Poverty. Washington, D.C. WHO (World Health Organization). 2000. The World Health Report 2000: Health Systems-- Measuring Performance. Geneva. Zere, E., D. McIntyre, and T. Addison. 2001. "Technical Efficiency and Productivity of Pub- lic Sector Hospitals in Three South African Provinces." South African Journal of Econom- ics 69(2): 336­58. CHAPTER 4 Experience of Community Health Financing in the African Region Dyna Arhin-Tenkorang Abstract: Studies and literature reviews of health insurance schemes targeting rural or informal sector populations in developing countries (often called community insurance schemes) frequently conclude that schemes have design weaknesses, yet do not explore in detail the effect of design features on performance. This chapter presents a conceptu- alization of how performance in the areas of risk protection and resource mobilization is determined by the interaction of design features with institutional and technical factors. Design features refer to scheme specifications (for example, required contribution) and to operating modalities (for example, procedures for enrollment or obtaining benefits). Performance, with respect to risk protection and resource mobilization, of several poten- tial high-population schemes for the informal sector in Africa, is assessed. The outcome suggests that the design of community health insurance schemes may be improved by (a) design specifications that utilize data on willingness to pay (WTP) of the target popu- lation and projected health care costs; and (b) incorporating modalities of operations that facilitate cost-effective exchange between a formal organization and individuals act- ing in an informal environment. I ncreasing the access of African populations to health care is one of the formi- dable challenges facing the global community. During the 1980s and 1990s, African governments, with the endorsement of their international and bilat- eral donor partners, implemented health sector reforms intended to improve the efficiency of health systems and the quality of care. In many countries, these reforms included the introduction or consolidation of cost-recovery mecha- nisms, in particular out-of-pocket fees, paid at the time of illness (user fees), which had the unintended effect of decreasing the poor's access to health care (Bethune, Alfani, and Lahaye 1989; Booth and others 1995; Nyonator and Kutzin 1999). Since the mid-1990s, the increasing incidence and prevalence in low-income countries of HIV/AIDS, tuberculosis (TB), and other communicable diseases have contributed to the widening of the gap between the need for, and the utilization of, health services among poor individuals. At the time of ill health, households in Africa do not have recourse to mecha- nisms that will protect the financial resources required for basic consumption needs, such as transportation, education, and food not produced by the house- hold. As most functional health insurance schemes in Africa are associated with formal sector employment--requiring regular contributions compatible with 158 Health Financing for Poor People: Resource Mobilization and Risk Sharing formal sector earnings--the majority of individuals are not insured. Vogel (1990) and Abel-Smith and Rawal (1994) conclude that the formal sector schemes effec- tively cover only members of the relatively small upper and middle classes. Uncertainty about the timing of illness, the unpredictability of health care costs during illness, and the low and irregular income of individuals mean that it is virtually impossible for households to make financial provision for illness- related expenditures. User fees constitute a major part of such expenditures. As a consequence, user fees have been, and still are, a major contributing factor to the high incidence of out-of-pocket payment by individuals and households at the time of illness. Furthermore, most households cannot obtain credit from the formal banking system. Thus user fees, in addition to having been largely unsuc- cessful in raising significant resources, have contributed significantly to increas- ing the exposure of poor households to financial risks associated with illness. Individuals are subject to illness-related financial risks correlated with health care prices and their disposable incomes. As ratios of health care prices to incomes rise, households' probabilities of illness-related loss of wealth and assets increase. Health care must often be consumed in complete packages and is therefore a dis- crete, rather than a continuous, variable in the health production function. Fur- thermore, components of packages (for example, consultations, laboratory tests, prescribed drugs) will vary in quantity and type, giving rise to complex relation- ships between the quantities consumed, the costs, and the health outcomes. As a result of the complexity of these relationships and the variations in the type and course of illnesses, identical household budget constraints often have disparate impacts on the consumption of effective health care. In poor commu- nities, this complex relationship also leads to identical health status outcomes for households, irrespective of the income groups to which they belong. Although people in the high-income groups, obtained by ranking, have the eco- nomic means to purchase a greater proportion of health care packages, providers are often unwilling to offer an incomplete package--as in the example of "half a surgical procedure." If an incomplete package is offered, it is usually ineffective in improving health, as in the case of a partial course of antibiotics. Conse- quently, in many situations of low per capita incomes, ranking households into income groups is of little use for policy formulation aimed at providing univer- sal access to effective health care. Rather, public provision of financial protection becomes a crucial element of strategies to reduce poverty for all households in poor communities such as those in rural areas and slums, irrespective of their incomes relative to others in those areas. Over the long term, health investments that make preventive health care avail- able will lead to improved health and productivity of the people, hence to higher incomes. In the short term, provision of access to curative health care is needed to limit income shocks from illness that might otherwise push people into poverty. Households can frequently prevent illnesses, but they are unequipped to treat many illnesses effectively. To prevent malaria, for example, a farming mother in a village in Ghana can build up her immunity system and Experience of Community Health Financing in the African Region 159 her child's by eating high-protein, homegrown foods and by breastfeeding. She can also take actions to repel mosquitoes. These measures enable her to dramat- ically influence the frequency and prognosis of malaria episodes suffered by her child. In contrast, because households cannot treat severe malaria, society con- siders accessible treatment of this disease a high priority among the functions expected of its health system. A premise of this chapter is that in poor African countries, individuals in the informal sector--regardless of their income rank--cannot access appropriate health care, particularly curative care, at the time of need. In one study, expen- diture for 70 percent of inpatient episodes exceeded 6 percent (in the fourth income quintile) and 4 percent (in the highest quintile) of the average annual income for individuals (Arhin 1995b). In this environment, insurance schemes that provide financial protection to households in the informal sector would constitute an important poverty-reduction measure. Another premise is that attempts to stratify informal populations by income, so as to target financial protection just to those in the lower ranks, will be only partially effective in achieving access to health care for the very poor. Establishing subnational health insurance schemes, each targeting house- holds in defined poor communities such as villages or districts, is an option for providing immediate financial risk protection (FRP) to a significant number of households. It also offers the potential of eventually achieving universal cover- age and high cross-subsidization between high- and low-income households through the future linking of schemes for the informal sector to each other and to schemes for the formal sector. The alternative strategy of attempting to pro- vide universal coverage on a national scale by implementing a single national scheme at the outset would be problematic because of the diversity in African populations and the absence of appropriate administrative infrastructures. Studies and literature reviews of subnational health insurance schemes (often referred to as community-financing schemes or microfinancing schemes) targeting rural or informal sector populations in developing countries frequently conclude that such schemes have design flaws that impair their performances (Bennett, Creese, and Monasch 1998). As the reviews do not explore in detail the relation- ships between design features and specific dimensions of performance, this chap- ter focuses on the design-related issues of subnational health insurance schemes for the informal sector in sub-Saharan Africa. It examines, in particular, the func- tions of risk protection against the financial consequences of ill health and the mobilization of significant resources for the health sector. Resource mobilization is examined with reference to the fiscal impact of the insurance on health care pro- vision rather than from an accounting perspective focusing on meeting the finan- cial obligations of a given scheme. Therefore cost recovery is considered a variable related to resource mobilization rather than a prime indicator of performance. The first part of the chapter presents a conceptual framework intended to inform readers about the design of a health insurance scheme that focuses on the characteristics of schemes and factors influencing their performance. It presents 160 Health Financing for Poor People: Resource Mobilization and Risk Sharing key technical and institutional factors that, together with the design features, determine scheme performance in resource mobilization and financial protec- tion. Emphasis is placed on dimensions of these factors that are particularly rel- evant to schemes targeting individuals and households in the informal sector in low-income countries. The technical factors relate to the data (its nature and quality) and the methodology used to specify the contribution levels, the bene- fit package, and the level of external financial subsidy. The institutional factors include the congruence between principles underlying the scheme's operations and norms of the participating population. Other institutional factors include experiences of health providers with third-party contractual arrangements and payments. These institutional factors have a crucial influence on the nature and extent of community participation in the scheme, and on the quality of scheme management and monitoring. Although regulatory factors--such as the guide- lines produced by responsible government agencies and laws governing insur- ance and health care provision--also determine the insurance and health care quality, they are not considered here. In the second part of the chapter, evidence is reviewed about the effectiveness and efficiency with which selected health insurance schemes for the informal sector population have achieved resource mobilization and risk protection. The evidence is presented mainly from schemes judged to have the necessary-- though often insufficient--design and implementation features that would per- mit expansion to achieve high primary or secondary participation rates.1 Logically these features exist because, at the minimum, the calculation of contri- bution levels reflects the ability and willingness to pay (economic demand) of the target population. The proposed scheme is described to potential buyers in terms of benefit package and management structure, and the demand is the ability and willingness to pay (WTP) the annual contribution in the required installments. Although relatively few schemes for Africa's informal sector populations have based their contribution calculations on WTP data, some, through a variety of approaches, have arrived at affordable or near-affordable premiums for their target populations. Some schemes have also obtained sufficient external resources to potentially fund the type and quality of health care benefits their enrolled members expect. Such potentially large population schemes include the Carte d'Assurance Maladie (CAM) program (Burundi), Community Health Fund (CHF, Tanzania), Abota Village Insurance Scheme (Guinea-Bissau), Nkoranza Community Financing Health Insurance Scheme (Ghana), Bwamanda Hospital Insurance Scheme (Democ- ratic Republic of Congo), and Dangme West Health Insurance Scheme (Ghana). Of the examples of potentially large population schemes studied in this chap- ter, two of them, CAM in Burundi and the Health Card Fund in Tanzania, are in some respect national schemes. These two schemes have significant central gov- ernment involvement. Other schemes may be considered district-based schemes from the outset, while the Abota Scheme, for example, evolved into a form of a district scheme and is still expanding, through an institutionalization process, toward national dimensions and characteristics. Criel (1998) uses two main fea- Experience of Community Health Financing in the African Region 161 tures to characterize three schemes in Africa as district-based health insurance schemes (DBHIS): first, a prominent role of the local health service administrators in its operations, and second, an expectation that the scheme will eventually cover the entire population residing in the health administrative area or district. Evi- dence is presented from several schemes (including the Nkoranza Scheme in Ghana and the Bwamanda Scheme in Democratic Republic of Congo), which under Criel's framework are examples of DBHIS. These are effectively single-facility provider schemes (in both cases the facility being a hospital) and are susceptible to the problems of low participation because benefits are accessible at only one loca- tion, discouraging enrollment by people who live far from the provider. When the inclusion of all levels of facilities (health post, health center, and district hospital) as providers is used in addition to Criel's two characteristics to categorize DBHIS, the only true example in the chapter is the Dangme West Health Insurance Scheme (Dangme Hewami Nami Kpee) in Ghana (Arhin and Adjai 1997). The outpatient (OPD) benefit package provided by lower level providers (health centers) in the Dangme Scheme encourages insured patients to present illnesses early, when treatment resource requirements are minimal. In addition, the absence of financial incentives for insured patients to obtain inap- propriate admissions has the potential to foster proper functioning of the refer- ral system. Finally, the availability of OPD and emergency care as part of the benefit package has proven a critical factor in decisions to participate in a scheme. Although the schemes studied have large target populations, provide a comprehensive package in some cases, and are geographically readily accessible to enrolled members, the evidence shows that enrollment is often relatively low. Many institutional factors interact with design features either to enhance or to limit the performance of all schemes. Based on the evidence from the selected schemes, the third part of the chap- ter discusses those institutional and technical factors that appear to influence performance. In the absence of established best practices in the scheme design, an argument is made for experimentation, guided by lessons from these schemes. Suggestions are also made of broad national and international policy measures to support the implementation of risk-protection schemes for popula- tions in the informal sectors. CONCEPTUAL FRAMEWORK Health Insurance, Risk, and Willingness to Pay Health insurance is a mechanism for spreading the risks of incurring health care costs over a group of individuals or households. This definition is not dependent on the nature of the administrative arrangements employed, but on the out- come of risk sharing and subsequent cross-subsidization of health care expendi- tures among the participants. An arrangement designed to provide risk sharing 162 Health Financing for Poor People: Resource Mobilization and Risk Sharing for illness-related events, and which is accessible to households in the informal sectors in low-income countries, is a health insurance scheme regardless of the orthodoxy of its operational modalities. In such an arrangement, an insured individual acquires "a state-contingent income claim" before the state of the world is known and is entitled to resources, income, or both to address the event for which he or she is insured if the event occurs. Some studies have reported that low-income households are initially reluctant to join insurance schemes because they do not readily accept the idea of "paying" for services they might not use (Brown and Churchill 2000). Interpreting such findings as evidence that these households have risk attitudes nonsupportive of insurance (risk neutral or risk-loving attitudes) would predict limited potential for insurance schemes targeting these households. In contrast, three studies in Ghana, Burundi, and Guinea-Bissau suggest that households in rural areas are risk-averse with regard to health care (Arhin 1996a, p. 629). Such differences in population attitude and WTP for health insurance would theoretically lead to predictable variation in insurance scheme enrollment. Therefore WTP informa- tion for a target population would facilitate scheme design and implementation. Currently, limited theoretical constructs and empirical evidence are available to guide WTP studies undertaken in developing countries for the purpose of pricing goods and services, as in the case of providing pricing inputs for the design of goods supplied publicly rather than privately. Consequently, published data on the demand of population groups for health insurance, as indicated by their willingness and ability to pay premiums, pertain to a limited number of countries (Arhin 1996b; Mathiyazhagan 1998; Asenso-Okyere and others 1997). In many developing countries, ensuring the reliability and validity of WTP studies of insurance goods presents many problems, partly because of the popu- lation's limited experience with insurance policies. For example, this limited experience increases the probability that inappropriate discourse may be used to ascertain perceptions and result in an erroneous conclusion of reluctance to pay for uncertain consumption. As a consequence, inept strategies of basic education may be adopted, instead of marketing approaches, to provide information on modalities and build trust. Exploratory discussions before introducing a scheme in a rural part of Ghana found that the term health insurance was not associated with risk sharing and instead referred to an unfamiliar product purchased mainly by the urban elite. Risk-sharing arrangements that were familiar to the rural communities were described as solidarity groups, associations of people who assist each other when events associated with specific needs occur (Arhin 1995c). Relevant Scheme Models Health insurance schemes are arrangements in which officials formally hold funds that consist of payments by insured participants and use the resultant resource pool to finance all or part of members' health care costs. In African countries that have schemes for the informal sector, most plans fall into the first three of Experience of Community Health Financing in the African Region 163 the following four models. Where the officials are members of an identifiable group whose contributions make up the pools, and are responsible for manage- ment activities such as determining benefits and contributions, the model is a mutual benefit society model. Atim provides an example and defines a mutual health organization as "a voluntary, nonprofit insurance scheme, formed on the basis of an ethic of mutual aid, solidarity, and the collective pooling of health risks, in which the members participate effectively in its management and func- tioning" (Atim 1998). In provider insurance models, the officials originate from the health care provider institution (or from the ultimate provider organization such as the government or mission health administration) and manage both the insur- ance and the health care aspects of the scheme, similar to health maintenance organizations (HMOs). In a variant of these mutual and provider models, the offi- cials are responsible for managing the insurance product and providing health care, and are drawn from members of a mutual society as well as a health care provider organization. Such a model may be termed a mutual-provider partnership model and correlates in general to the concept of community-based insurance put forward to test the hypothesis of feasibility of insurance for households in the informal sector (Arhin 1991). Third-party insurance has not been a feature of insur- ance schemes for the informal sector in Africa.2 Each of the four design models is associated with incentive structures that influ- ence the behavior of the actors in the models. For example, mutual models encourage greater accountability to the individuals who make up the pool but also generate a significant requirement for committed time, skill, and knowledge from these individuals. Mutual-provider partnership models reduce the level of these requirements from the informal insured group. The integration of insurance man- agement and health care provision in a scheme offers strong motivation for pro- viding health promotion services and preventive care to limit benefit claims. Yet where health care is provided to members by a separate insurance entity, the threat of nonpayment or contract termination may result in a higher quality of service. The skills and management capacities of the different actors, resource availability, the nature of the existing power balance, and the prospects for posi- tive change will influence the appropriateness of a model for a given setting. In Africa, schemes intended for the informal sector are confronted with the target populations' low and irregular incomes and consequently negligible profit-making potential. Of necessity, therefore, schemes for the informal sector have social welfare dimensions rather than commercial characteristics. This is made more apparent in mutual models since they involve actions by social insti- tutions, communities, and the state (the latter through regulation and legisla- tion). The mutual schemes represent public action taken to reduce human deprivation and eliminate vulnerability (Burgess and Stern 1991). They facilitate explicit or implicit participation by communities in scheme design and imple- mentation. Depending on the community's composition and cultural norms, emphasis will be placed either equally or preferentially on achieving the schemes' social and financial functions. Although they are intrinsically linked, 164 Health Financing for Poor People: Resource Mobilization and Risk Sharing the social and financial functions performed by a social welfare-oriented health insurance scheme may be considered separately. The social function affects risk protection for the individual, whereas the financial function leads to resource mobilization for the group. Risk-Protection Function of Schemes Risk protection, in the health context, is the shielding of an individual from critical income losses as a result of illness or injury. In sub-Saharan Africa, critical income is the resource required for needs such as food not grown on the household farm, housing construction material that cannot be produced by the household (for example, corrugated iron sheets for roofing), and basic formal education. Health- related financial risk protection is inversely related to the percentage of income required to meet expenditures related to treatment for an illness episode. To illus- trate, a household survey finding that out-of-pocket expenditure for 20 percent of inpatient episodes exceeded half the mean annual household income for people in the lowest income quintile suggests that risk protection is nonexistent for a sig- nificant proportion of Africans (Arhin 1997). Insurance schemes that provide 100 percent coverage for illness episodes provide the highest levels of protection, and those involving copayment provide lower levels of protection. Protection is a function of income, the price of health care and other goods, the illness incidence, and the completeness of an insurance benefit package. Therefore data on all these aspects of illness and economic activity are neces- sary to assess the financial protection afforded by a health system. Using the system's ability to safeguard individuals' critical income in the event of illness as a measure of risk protection allows cross-country comparisons of schemes' risk-protection performance under different economic circumstances. Other measures of risk protection such as differences in the out-of-pocket payment between scheme members and nonmembers permit comparisons only between schemes operating in similar economic situations. For example, two schemes whose benefit structures require members to pay the same flat fee per prescrip- tion will provide different risk protection if the incomes and living costs of members of the two schemes are dissimilar. To provide FRP, a scheme must offer an insurance product that is accessible to the target population and either eliminates payments associated with receiving care, as in the case of a zero copayment rate, or reduces the payment to a level that has negligible impact on critical consumption. Accessibility in this context will be high when a scheme's premium does not exceed targeted individuals' noncritical income. The compatibility of the collection schedule with the target households' cash flow patterns, for example, taking into account the seasonality of agricultural workers' cash income, will also enhance accessibility. Since the process of obtaining health care in most low-income settings is a "bad" rather than a "good" (often associated with long journeys on foot and rel- atively expensive and uncomfortable travel by road, long hours of queuing, and Experience of Community Health Financing in the African Region 165 loss of production for most of the day), there is little justification for including measures such as copayments to reduce possible moral hazard. In addition, the completeness of the benefit package is a design feature that has a major influence on FRP. Schemes in which the benefit package excludes commoner expensive care or both will, as in the case of copayments, frequently entail significant pay- ments for the care that is covered, and therefore limited financial risk protection. In many instances, the data on illness incidence, incomes, and prices required for direct assessment of the protection offered by an insurance scheme will not be available, as scheme implementers rarely collect population-based data. As an alternative, from the information commonly collected about community health insurance schemes, a set of "second-best" markers may be combined and used to provide an approximate evaluation of a scheme's financial risk protection. These markers are composed of, first, process indicators of a scheme's accessibility and the likelihood of payment associated with receipt of care, and second, outcome indicators of increased accessibility to care. They include the following: afford- ability of premiums charged by a scheme; appropriateness of payment schedules to their target population, particularly rural dwellers and informal workers; absence of copayments or affordable copayments; completeness of the benefit package; and increases in the population's utilization of health care services as a result of a scheme. In the absence of risk protection, the cost of care becomes a barrier to seeking and obtaining health care. Thus health insurance not only provides protection for the income consequences of ill-health, but also removes financial barriers to obtaining health care at the time of illness, enabling prompt access to treatment. Several African studies have demonstrated that many typical households cannot utilize formal health services effectively for lack of cash to make the immediate cash payments involved (Arhin-Tenkorang 2000). Sick individuals have to post- pone visits to the health facilities until their conditions become critical. Yet delayed emergency treatments can lead to serious health and financial conse- quences resulting in further impoverishment of the household. The financial barriers to care from the formal health system often lead would-be patients to resort to self-medication and other practices that sometimes injure their health. The risk protection provided by insurance also improves health equity in a community. Equity is enhanced as the healthy, at lower risk of illness, subsidize the health care costs of less healthy, higher risk individuals. Although this may be regarded as income redistribution, the more critical interpretation is that health insurance promotes equity because, irrespective of economic status, indi- viduals who have equal health care needs (that is, capacity to benefit from care) are assisted in obtaining comparable care. All participating members of an insur- ance scheme benefit from the removal of uncertainty about their claims to health at the time of ill health. In addition to the private benefits, health insur- ance, by promoting the optimal consumption of health care by individuals in society, maximizes the public benefit accruing from the positive externalities associated with healthier populations. 166 Health Financing for Poor People: Resource Mobilization and Risk Sharing Resource Mobilization Function of Schemes A health insurance scheme functions as a financial arrangement for mobilizing and pooling funds to cover all, or (in government-subsidized schemes) part, of the cost of health care for contributors to the pool. The existence of a pool facil- itates the realization of economies of scale and reduces the tendency to micro- purchase (spend for a few isolated cases as revenue becomes available from, for example, user fees). Thus the pool enables providers to increase efficiency and cost-effectiveness. A scheme's financial performance is a function of contribu- tions, the cost of health care consumed by the insured, the level of external sub- sidy, the size of the pool, and the extent of economies of scale achieved. For a scheme to produce net financial benefit to society in a given period, the pool of funds should represent resource mobilization in excess of that which would otherwise be achieved for health care expenditure by the population. This is true if, in the absence of insurance, some of the funds in the pool would not have been used for health activities (for example, the contributions of members who do not become sick during the period). From the perspective of the public sector, insurance- mediated resource mobilization can also be a consequence of shifts from private sector health expenditure of the participating population. Measuring the resource mobilization effect of a scheme necessitates an estimation of percentage increases in resources for health expenditure among the target group, and this will be heav- ily influenced by people's willingness and ability to pay or contribute. In the absence of direct measures of resource mobilization, an approximation can be obtained by ascertaining if the ratio of health care expenditure to scheme revenue from contributions is positive and greater than unity in the presence of evidence of increased overall utilization rates (that is, for both insured and unin- sured). Alternatively, positive and greater than unity ratios of "average expenditure on an individual" to "the average individual contribution" will also imply net pos- itive resource mobilization. Where the average individual contribution approaches the average expenditure per individual member of the scheme, it suggests that the scheme is functioning as a medical savings scheme. A medical savings scheme enables its members to prepay for their health care and significantly limits the net resource mobilization benefit to the society. From the above conceptualization, net positive resource mobilization will be associated with the following outcome indi- cators of financial performance, which may be used as second-best markers: posi- tive and greater than unity ratios of "health care expenditures" to "revenue from contributions," and significantly greater than unity ratio of "average expenditure on an individual" to "average individual contribution." The financial performance of schemes can be considered adequate if they raise enough revenue from contributions to cover the target or stipulated per- centage of the costs of delivering care, plus all administrative costs. Yet schemes may have poor financial performance because of an excess of high-risk members in the scheme (adverse selection) or as a consequence of targeting to fund from contributions a percentage of costs incompatible with the willingness and ability Experience of Community Health Financing in the African Region 167 of members to pay. Although in theory, irrespective of a net benefit in resource mobilization terms, a scheme may be unable to fund the target percentage of the health care claimed by its insured because members make inappropriately high demands for care as a response to being insured (moral hazard), as already argued, moral hazard is not a major problem among poor populations. Design and Institutionalization A consultative and inclusive process of scheme design will facilitate positive and optimal interactions between the scheme and entities that constitute its stake- holders. Effective consultation requires the specific identities of all the major stakeholders to be established at the outset of the design process. Because a state- ment of the scheme's mission defines its business in terms of client groups, client needs, and implied organizational competence, early adoption of a mis- sion statement will assist early identification. As stakeholders, these entities have a claim, or interest, in the scheme, and in what it does and how well it performs. Among the local entities, some (health workers, health facility managers, and advisory committee members) will be internal stakeholders. Others will make up the external stakeholder group--insured members, suppliers, local supporting government and donor institutions, and the general public. As stakeholders are, or will be, the providers of present and future resources to the scheme as repre- sented in figure 4.1, they form relationships with the scheme in which resources are exchanged in the expectation that interests will be satisfied by inducements (Hill and Jones 1998). Like any other organization, a health insurance scheme must therefore have as part of its objectives the provision of satisfactory returns on stakeholder investments and the satisfaction of their interests to maintain FIGURE 4.1 Relationships between Stakeholders and the Scheme Contributions External Stakeholders The Members Scheme Suppliers Inducements Governments Local communities General public Contributions Inducements Supporting donors Internal Stakeholders Employees Managers Board members Source: Hill and Jones (1998). 168 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 4.1 Scheme Design Options Dimension Choices Strengths Weaknesses Utility functions Access maximization Encourages affordable Discourages affordable Revenue maximization premiums and hence high premiums (based on WTP) participation Cost recovery Encourages community- rated premiums and enhances cross- subsidization between scheme members Operation level Single-facility based Encourages wider Discourages participation (a hospital scheme) geographic spread of by distant populations, and Provider-network based participants and greater limits range of benefits (scheme consisting of range of benefits, and participating clinics and permits effective referral referral hospitals) system Range of providers Restricted to public and Absence of profit Profit margins may lead to not-for-profit private motivation and use of high premiums Includes or limited to existing structures may for-profit private lower premiums Range of benefits Full basic package Encourages high (outpatient plus inpatient enrollment, prompt care) attendance to health Partial packages problems Discourages early treatment until illness is severe or complications arise that are covered by scheme, leading to high costs Note: WTP = willingness to pay the exchange of resources. As a consequence, a critical component of a scheme's governance will be mechanisms to ensure that the actions of scheme officials are consistent with the interests of the different stakeholder groups. By undertaking a situation analysis in which information is collected about proposed or existing target populations and about care providers, the potential contributions of stakeholders to the functions of a scheme can be factored appro- priately into the design. Participation of stakeholders in the design will help ensure that the interpretation of the information obtained is based on an under- standing of societal norms and expectations, as these will have a bearing on the skills and resources that stakeholders may wish to make available to the scheme. The approaches for collecting the relevant detailed information include focus group discussions, household surveys, and health facility costing studies. A situa- tion analysis of a proposed or an existing scheme will reveal for consideration design options in several dimensions. Table 4.1 sets out the main options available Experience of Community Health Financing in the African Region 169 TABLE 4.2 An Example of Goals Matched to Design Options Goal Matching design option Financial access--affordable contributions Public and not-for-profit private No or partial cost recovery Physical access to providers Provider-network based Efficiency Full package benefits when designing a health insurance scheme for the informal sector in Sub-Saharan Africa. Ultimately, a match between the selected goals and options is the intent. For most schemes in Sub-Saharan Africa, an analysis of the situation reveals, in addition to information about stakeholders and design options, the following contextual factors: inadequate budget allocation to the health sector coupled with financial barriers to use resulting from out-of-pocket fees; and inadequacies of physical input and human resources resulting in poor quality of health care. An additional feature of this context is a crisis of management and administra- tive skills within the health sector. This is manifested by micropurchasing and an inadequate referral system, which in turn lead to internal inefficiencies. These contextual factors are the main determinants of the goals that have been selected for schemes. These and other situational findings relating to the target population's economic profiles and their health-related needs have, in most cases, led scheme designers and operators to adapt implicitly or explicitly one or more of the following goals: (a) provision or improvement of financial access with high participation and risk protection, (b) provision or improvement of physical access to health care providers, (c) improvement of supply of services, and (d) improvement of quality of care. Scheme designers' challenge is, in consultation with stakeholders, to identify the design options that will achieve the group's goals. Table 4.2 gives an example of a set of design options that could be used to meet goals in physical and finan- cial access and efficiency. Tasks to be Accomplished by Designs As stated above, in pursuit of its goals, a health insurance scheme for the infor- mal sector--like other organizations--must accomplish the primary task (some- times referred to as "buffering" by Hatch) of resource (material, services, and money) transfer between it and the entities that make up its stakeholders (Hatch 1997). In particular, a scheme relevant to this discussion must first effect the effi- cient transfer of contribution funds from its members in the informal sector to an organization that operates largely in the formal sector. The majority of mem- bers will be limited to transactions in cash, and therefore potentially efficient payment mechanisms are not options (for example, checks, installments by 170 Health Financing for Poor People: Resource Mobilization and Risk Sharing standing bank orders, payments in response to regular bills, and credit card pay- ments). The scheme must then realize transfer of health care resources, in some cases cash from or via the scheme, to health care providers or members of the scheme who have limited mobility or transportation possibilities. Another major task of all organizations is the transfer of information across the boundaries that separate it from its environment (boundary spanning). In a health insurance scheme, information transfer is required between scheme offi- cials and elements in the environment such as insured members, health care providers, and the government. This activity is crucial in establishing an under- standing of the obligations and benefits between members and the scheme. Additionally, by capturing shifts in the scheme members' demand preferences and willingness to pay, transfer of information helps to minimize differences in perceptions of these obligations and benefits over time. A set of economic and financial specifications will be required to facilitate the primary task, and specific administrative and management structures--including business and legal processes--will be the basis for realizing the information- related task. For a given benefit structure, the following economic variables must be specified: amount of contribution (may be cash or in-kind), unit of contribu- tion (may be per individual, household, or village), and schedule of contribution (may be per month, half-year, or year). For financial viability and the meeting of financial obligations, expenditures must be less than the revenue from contributions. These factors have a bearing on resource mobilization, as explained above, while financial protection of the target population will be increased by a unit contribution that does not exceed the target group's WTP. A trade-off is required in setting these specifications when the contribution that leads to desired financial viability is greater than the contribution that opti- mizes risk protection. Alternatively, both financial performance and high-risk protection can be maintained by lowering financial obligations through subsi- dies to the scheme. Administrative and management structures to realize the task of information transfer provide the data to monitor financial performance and risk protection. Their monitoring depends on continuous predictions of expenditure, based on detailed data on the target population and provider activities. With regard to the target population, the specific data include the level of utilization, pattern of ill- ness, and care-seeking behavior. Data required about the care provider include cost per unit of care for different services and the financing or income source. Similarly, monitoring will rely on continuous assessment, prediction of revenue, or both, as determined and influenced by WTP, health insurance perceptions and preferences, quality-of-care perceptions and preferences, and the target pop- ulation's socioeconomic characteristics. Hence this secondary function trans- lates into consumer-oriented research and marketing and also leads to internal analysis of the scheme by its officials. The goal of an internal analysis is to understand the scheme in depth: its strengths, weaknesses, resources, and con- Experience of Community Health Financing in the African Region 171 straints. Information from such an analysis can be used to develop responsive strategies that either exploit strengths or compensate for weaknesses. This process must be ongoing to facilitate the institutionalization process. Institutionalization of the scheme's operational modalities involves a lengthy process of innovation, political lobbying, and geographic expansion. The process permits civil societies to gradually tailor the modalities of operation to local socioeconomic conditions and align their operations with prevailing cul- tural norms relating to collectivism. This is less likely to happen if a scheme starts as a detailed project proposal, created by a development professional. Development initiatives that have detailed plans at the outset often have limited opportunities for communities to participate fully. Participation, as used here, refers to a "process of empowerment of the deprived and the isolated" (Ghai 1988). It implies an increase in the ability of social groups to exercise economic and political power and to make decisions, both private and public. Community involvement in the scheme for the informal sector should therefore extend beyond the mobilization of local material and labor resources. The willingness of the formal health sector to engage in dialogue and collaborate with potential patients as a collective, rather than as individuals, is a crucial determinant of the success of the institutionalization process. The organizational culture of the health providers in the scheme will determine the ease with which effective dia- logue and collaboration evolves between scheme members (through their offi- cial representatives), health managers, and health care workers. In addition, functional collective solidarity schemes in most African societies usually involve the provision of benefits produced in the informal sector and in the control of ordinary people rather than professional groups. Examples of this are associa- tions that provide member households with funeral assistance or periodic pay- ments from member contributions. EVIDENCE This section presents information from the literature and personal experience about the performance of schemes, particularly the provision of financial protec- tion to their target populations and mobilization of resources for health care. Because data on this topic are limited, the focus is on indirect measures of perfor- mance with reference to specific schemes. The literature about existing "poten- tially large population" schemes illustrates a diversity of benefits, organizational structures, administrative arrangements, and paths of evolution (see table 4.3). Carte d'Assurance Maladie, Burundi This is a national health card insurance scheme introduced by the government of Burundi in 1984. The scheme has the characteristics of the provider insurance model in which the government is the organization responsible for both managing 172 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 4.3 Features of "Potentially Large Population" Schemes for Informal Sector Households Policy and legal Organizational Contributions estimation: context structure method and data Carte d'Assurance Part of national strategy Financial control by local Fixed nationally by Maladie (CAM) program, Revenue collected and government (communes) central government Burundi (provider controlled by local Care provided by public No information on insurance model, by government health facilities method or data government) Community Health Fund Governed by district Management and Inconclusive (CHF), Tanzania council bylaws decisions by district information on method (mutual-provider Political support from CHF board and CHF No evidence of use of partnership model) central government ward committees data Abota Village Insurance Village committees Village management Trialanderrorby Scheme, Guinea-Bissau control with oversight Suppliers from central community to limit (mutual benefit society by donors, government, MOH periods of drug stock-outs model) or both Nkoranza Community Religioushealthsector 3-tier: Subjective judgment of Financing Health rules by hospital staff Advisory board affordability by hospital Insurance Scheme, Ghana and external donor Management team staff (provider insurance model, oversight and regulations Voluntary registrars by hospital) Bwamanda Hospital Enrollmentand Selection of premium Insurance Scheme, Dem. accounting by health staff and copayment options Republic of Congo (provider SchemeassistedbyNGO by community insurance model, by hospital) Dangme West Health Government by Community association Based on simulation Insurance Scheme decentralized district and the District applying data on WTP (Dangme Hewami Nami and regional MOH Insurance Management and costs and illness Kpee), Ghana Planned registration of Team rates from survey (mutual-provider association as an NGO partnership model) the insurance scheme and providing health care. Purchase of a Carte d'Assur- ance Maladie (CAM) card by a household entitles its members (restricted to two adults and all children under 18 years of age) to free health care at all public health facilities. The card is sold at a fixed price, irrespective of household size (in June 1992, the price of the card was 500 FBu [Burundi franc] [US$1.85]). Per- sons without cards must pay user charges for government health care. The user charge per episode of illness treated is determined by the health worker at his or her discretion and generally varies with the age of the patient and the quantity and type of treatment received. All health services provided by the government are covered by the CAM scheme, and therefore, in theory, CAM cardholders who seek health care at government facilities should not incur out-of-pocket Experience of Community Health Financing in the African Region 173 Premium level and membership unit Payment schedule Benefits package Purchase of CAM card, Annually All care at government 500FB (US$1.85) per clinic and hospitals household in 1992 Household defined as 2 adults and all dependent children Flat rate per household Annually All care provided by irrespective of size facility where registered Rate carries by ward No upper limits on value Cash or in-kind per Determined by village, Outpatient care by household or individual usually annually or village health worker biannually Free referral care c450 (US$1) in 1991 per Annual registration Inpatient care individual enrolled by October­December Refund of referral family expenses Cash = price of 2 kg soy Annually (normally in All hospitalization and beans + copayment per March) chronic care in health individual centers Membership by Planned to be annually Outpatient care at any households district facility Contribution per Inpatient care when individual referred to regional hospitals expenses. However, due to the shortage of drugs and other inputs, CAM holders, like fee-paying patients, are sometimes given prescriptions to purchase drugs on the open market. The names of household members entitled to use a card are written on the card at the time of purchase, making it difficult for individuals from other house- holds to use it. The card is valid for one year and may be purchased from a com- munity representative at any time of year. This makes it possible for a non-CAM patient to pay a user charge at a health center and, on referral to a hospital, to purchase a CAM card to obtain free hospital care. The cards are not accepted by nongovernmental health facilities, such as mission and for-profit clinics and hospitals. 174 Health Financing for Poor People: Resource Mobilization and Risk Sharing The revenues from CAM card sales and user charges are retained by commune committees.3 These committees have some financial responsibilities for the health centers in their localities and are expected to fund recurrent expenditures such as stationery, fuel for refrigerators, linen, and in some cases, capital projects such as construction of new health centers. However, revenues from CAM and user charges are not designated to be used in the provision of health care, and therefore, in prac- tice, only a small fraction is allocated by communes to health. In 1990, 8 percent of the revenues of communes in Muyinga Province came from the sale of CAM cards, whereas an average of only 1 percent of commune revenues were used to finance health care (McPake and others 1992). The government, through the Ministry of Health budget, funds the salaries of health workers and drug costs. CAM: Risk Protection Affordability of premiums and copayments. A study in Muyinga Province, under- taken in 1993, reported that women had limited access to cash, and therefore, by eliminating cash payment at the point of care, CAM empowered them to decide the need for--and timing of--health care consumption by household members. (Women in CAM households do not require money, and hence permission, from male household heads to seek health care [Arhin 1994].) Cash had become less of a barrier to obtaining curative treatment for cardholders, and it appeared that women, being the main carriers, derived additional utility from the knowledge that, if a child were to fall ill, treatment would be available even in the absence of cash in the household. Large households were more likely to purchase a CAM card than small households, and therefore illness episodes per household were significantly greater for CAM households than for non-CAM households. This situation may be described as "adverse household selection," and possible results would be lowered risk sharing among households. The same study also found that approximately 27 percent of households gave "financial inability" to purchase a CAM card as one of the main reasons for non- membership. However, in practice, non-CAM patients referred from health cen- ters to higher level facilities often purchased cards prior to, or on arrival at, the referral center, thus manipulating the scheme to reduce their financial barrier to expensive curative care without prior participation as members. The conclusion of the study was that a CAM card provided significant financial protection to the communities studied. CAM: Resource Mobilization Relationships between premiums and benefits. A public treatment rate of 0.91 per household per month and a mean value of drugs of 134.1 FBu per formal treat- ment at the time of evaluation by the author implied that an average household consumed 1,464.4 FBu of outpatient drugs annually. The price of the CAM card could therefore cover 34.1 percent of the drug costs, suggesting that the ratio of "average expenditure on an individual" to "average individual contribution" Experience of Community Health Financing in the African Region 175 may have been less than unity. (In practice, revenue from CAM did not appear to be used to purchase drugs or medical supplies required for patient care at the health center.) In general, the target population for the scheme assessed the quality of care provided to members to be inadequate. As a consequence, women who participated in the focus groups were willing to pay a higher CAM contri- bution to improve the benefits provided. In particular, they were willing to pay a higher price for the CAM card on the condition that more drugs would become available at government facilities. Adverse selection and moral hazard are major causes of inefficiencies in the functioning of health insurance schemes and therefore have consequences for financial viability (Arrow 1963; Pauly 1963; Lohr and others 1986; Manning 1987). The rate of illness per person was found to be almost identical for mem- bers of CAM as for non-CAM households, and therefore adverse selection of individuals did not seem to be a problem. In the absence of adverse selection, it can be inferred that capitalizing on WTP for improved benefits would result in a significant increase in resource mobilization. Even so, this chapter omits an eval- uation of the level of resource mobilization actually achieved because other indi- cators of financial performance could not be obtained. Community Health Fund, Tanzania In Tanzania, the Community Health Fund (CHF) strategy for financing rural health services was piloted in Igunga District in 1996, and by 1999 it had been initiated in nine other districts. Currently, schemes comprising the CHF are gov- erned by district council bylaws and are also guided by a coordinator located in the headquarters of the Ministry of Health. Partly as a consequence of receiving funding and external technical assistance through a government­World Bank funded project, the CHF represents a national initiative and thus has the central government's political support. Three financing mechanisms are employed in the CHF: national user fees (introduced by the central government in 1993), insurance contributions, and matching subsidies from the government (funded by a World Bank project). A team of expert consultants designed the CHF and its schemes in 1995. Subse- quently, some aspects of the design were modified following consultations with communities in the first pilot district. In particular, the level of pooling was changed from districts to subdistricts (wards), although a central management role was retained at the district level by the district CHF board. Decisions about the use of funds are made at ward level and are reflected in ward health plans and budgets for their public health facilities and outreach services. Employees of government health facilities have management roles in the CHF. They may be members of the ward health committee, may be selected by the community to collect contributions, or may be part of a consultative group representing com- munity groups and health facility management that meet to review progress. The model of the CHF is therefore a mutual-provider partnership model. 176 Health Financing for Poor People: Resource Mobilization and Risk Sharing All households in a participating district are eligible to enroll in the scheme and are required to pay a flat rate contribution per household irrespective of household size. The level of contribution is determined by the community (at the ward level), and the decisionmaking process is in the hands of the CHF ward committee. Contributing households are registered with a public health facility in the ward where the household resides, and household members are entitled to all the health services provided by the facility. CHF: Risk Protection Affordability of premiums and copayments. One study found that "over 50 percent of members were considered to be poor," suggesting high affordability (World Bank 1999). In addition, the CHF schemes include exemption mechanisms for house- holds that are determined by the ward committee, and certified by the district board, as too poor to pay a flat insurance contribution or the user fees. In contrast, some health workers in Igunga District reported that the majority of participants of the insurance component of the CHF are salaried workers, and the procedure for obtaining exemption status is sometimes inaccessible by those whose incomes and socioeconomic status are very low. The care provided as a benefit is not sub- ject to an upper ceiling of value and copayments are not required. Resulting level of access to health care among the population. About 5 percent of the target population is enrolled, and insured individuals constitute 75 percent of patients who receive care in the provider facilities (Musau 1999). By inference, the increased funding of government health facilities provided by the scheme's revenue (contributions and matching funds) will increase the supply of, and hence the access to, care for the target population. CHF: Resource Mobilization By 1999, the total funds of the scheme for the participating district were US$371,000, consisting of 95 million shillings (35 percent) insurance contribu- tions, 81 million shillings (30 percent) user fees, and 95 million shillings (35 per- cent) matching government funds. The existence of matching funds makes it almost certain that the scheme resulted in increased resources for health care for the target population and hence a high level of resource mobilization. Abota Village Insurance Scheme, Guinea-Bissau The Abota system entails prepayment for essential drugs and the provision of pri- mary health care at the village level by trained villagers. The system comprises many hundreds of autonomous Abota schemes at the village level. Health care is provided voluntarily by members of the village, village health workers known as Agentes de Saude de Base, and by birth attendants at the village health post (Unidad de Saude de Base [USB]). The USBs were constructed from local building materials by the villagers and furnished with basic equipment (such as a metal storage cup- Experience of Community Health Financing in the African Region 177 board, an obstetric stethoscope, a lantern, and a kit of teaching aids) by the Min- istry of Health. Administration of the Abota system in each village is the respon- sibility of the village committee, the lowest level of the country's decentralized political system. This is an example of a mutual benefit society model, in which the officials are community members participating in the scheme. The earliest Abota schemes began in 1980 in a few villages as part of a general village health care program (Chabot and Savage 1984). Villages in the program adopted and modified an indigenous payment mechanism, originally used to collectively finance ceremonies to fund inputs for primary health care. Chabot and others describe the process of trial and error used by these villages over a three- to four-year period, to determine the frequency and level of prepayments that would ensure the availability of drugs throughout the year. Since 1983, patients referred by village health workers to the public health facilities have been exempt from payment of consultation fees upon showing evidence, usually a receipt, of having contributed to Abota. Growth of the scheme was continuous (in 1991, 462 villages participated). By the time of the author's study, the Abota system was widespread, forming an integral part of the country's health system. Furthermore, the Guinea-Bissau government's 10-year health plan (1984­93) emphasized the role of village-based primary health care, thus making the efficient functioning of the Abota system critical to the coun- try's health strategy (World Bank 1987). The Abota revenue is used to purchase essential drugs and bandages from nearby government health centers or sectoral hospitals. The ultimate supplier is the central medical store in the capital. In each village, the village committee decides the procedures for collecting contri- butions, purchasing drugs, and monitoring the system overall. As a consequence of this autonomy, prepayment terms vary substantially from one village to another. In 1988, the annual contributions per adult male varied from PG 20 to PG 500 (at 1988 exchange rates, US$1 = 1,129 Guinea-Bissau pesos). In 2 of 18 villages surveyed by Eklund and Stavem (1990), only men paid; and in another 2 villages contributions were by household. Other villages accepted in-kind con- tributions of agricultural produce. Since the mid-1990s, the country's economic difficulties have threatened the survival of the Abota. Problems such as inadequate recurrent budget allocations, drug shortages, and low health worker salaries plague the public health system (Tanner 1990). Ramifications include a decreasing capacity of government health workers to train and supervise village health workers and difficulties in resupplying village health posts, even when their Abota revenues are sufficient to fund their requisitions. In a few cases, Abota funds have been misappropri- ated (Knippeneberg and others 1991) either by village health workers or Min- istry of Health staff. The published and unpublished literature drawn upon to provide the descrip- tion of the Abota also permits a limited evaluation of the scheme. Again, the crite- ria of affordability, payment schedules, revenue and expenditure, and associated access to health care were used. The results are described below. 178 Health Financing for Poor People: Resource Mobilization and Risk Sharing Abota: Risk Protection Affordability of premiums. Studies of the Abota do not provide direct information on affordability, but since the contribution was set by the members of the vil- lage, it is likely that the majority could afford the amount set. The fact that many individuals were willing to increase their contribution also supports this general conclusion. Appropriateness of payment schedules. The number of times prepayment contri- butions were collected in a year varied: some villages collected twice yearly, others once. Because the villages themselves determined the frequency of payments, it is reasonable to assume the schedule was the most appropriate for the community. Resulting level of access to health care among the population. The fact that partic- ipating villages had basic health care made available within the village implies significant increases in access. Near-universal membership in participating vil- lages excluded adverse selection, and the watchfulness of health workers and local communities appeared to prevent moral hazard (Eklund and Stavem 1990). Abota: Resource Mobilization Relationship between premiums and benefits. In the first year of operation, the vil- lages' drug supply was depleted within three months because the revenue from contributions was exhausted. The scheme appears not to have held reserve capi- tal to ensure the quality of the insurance policy. Nkoranza Community Financing Health Insurance Scheme, Ghana Started in 1992, the scheme, an example of a provider insurance model in which management and health care are provided by hospital personnel, is adminis- trated by a three-tier structure: · Insurance management team (IMT) in the hospital, consisting of the medical officer in charge of public health and the hospital management team. This is the scheme's decisionmaking body. · Insurance advisory board made up of traditional, political, religious, and administrative leaders in the community and district health leaders (Ministry of Health and nongovernmental organizations). · Zonal coordinators and field workers--each of the 11 health zones is managed by a team of three zonal coordinators. They supervise voluntary fieldworkers who register families into the scheme. The premiums are calculated per person and thus vary with the size of the fam- ily. The scheme ensures access to care at St. Theresa's Hospital in Nkoranza Dis- trict (Ghana) by providing free admission in the medical, surgical, and maternity wards. However, admissions for normal deliveries are excluded. In addition, insured persons who are referred to other health institutions may claim refunds Experience of Community Health Financing in the African Region 179 FIGURE 4.2 Percentage of Community Enrolled, by Distance Percent of community enrolled 50 Nkoranza 40 Ayerede 30 20 10 Kranka 0 0 14 26 Distance from hospital (km) equal to the cost of an average admission at St. Theresa's. The scheme is available only to families living in the Nkoranza District, and every family member must register. Members have to travel various distances to the hospital, and a relation- ship has been observed between the distance of a community from the hospital and the enrollment of members in that community (see figure 4.2). Registration is renewable annually, during the last two months of the year only. Nkoranza: Risk Protection Appropriateness of payment schedules. In the first year of operation, 72 percent of fieldworkers and 65 percent of household heads stated that "people find it diffi- cult to register because they are short of money between October and Decem- ber" (Dept. of Insurance 1993). Sixty-five percent of fieldworkers and 75 percent of household heads were of the opinion that "more people will register if regis- tration is between January and March." Later evaluations of the scheme, includ- ing one undertaken in mid-1999, have continued to report a mismatch between the registration period and the time of highest cash availability among district households (Atim and Sock 2000). Resulting level of access to health care among the population. The findings sug- gested that the scheme had removed a barrier to admission for people in the dis- trict. Utilization, in terms of admissions per 100 person-years, was greater than the uninsured by a factor of 2.4. Among the insured population, an estimated 9 percent of admissions was due to factors such as earlier reporting of patients and more liberal admission policy of staff; without the insurance policy this would not have occurred. 180 Health Financing for Poor People: Resource Mobilization and Risk Sharing Nkoranza: Resource Mobilization Revenue from premiums and expenditure for benefits. During the first year of opera- tion, admission cost data for the insured population indicated that income obtained from premiums was 55 percent of the expenditure for insured inpa- tients. At that time, the premium was set at c450 per person (approximately US$1 at 1992 exchange rates). Premium increases of between 40 percent and 70 percent relative to the preceding year occurred annually. In 1998, the premium had risen to c5,000 per person (approximately US$1.90 at 1998 exchange rates). Since the mid-1990s, the ratio of revenue to expenditure has been approxi- mately one, implying significant resource mobilization. In addition, the ratio of the average cost per benefit (inpatient costs) to premium was 17.5 in 1998, also suggesting that resource mobilization among those enrolled was substantial. Bwamanda Insurance Scheme, Democratic Republic of Congo Bwamanda Health Zone is linked to a development project, Center for Integrated Development (CDI), started in 1969 and supported by Belgian volunteer workers. Impetus for the insurance scheme came from the zone's medical staff, concerned about the barrier to access caused by user fees and the low recovery of hospital costs from fees. Although the zonal hospital (a 156-bed reference hospital) had recovered 48 percent of its operating costs in 1985, this was a lower rate than that of eight comparable hospitals in the country (Bitrab and others 1987). The para- meters of the insurance plan, set by hospital staff, were explained to the commu- nity, and preferences between two options of premium and copayment levels were obtained. Enrollment of members was carried out by health center staff, mainly nurses, who received 3 percent of the premiums they collected as commission. Zone administrative staff made frequent supervisory visits to health centers during the enrollment period to monitor payment records, distribute membership stamps, and transfer premiums to the hospital. They also managed the plan at the hospital by verifying the membership status of admitted patients and keeping administrative and accounting records. Members of the Bwamanda insurance plan are covered for hospitalizations, including deliveries, dental extractions, and outpatient surgery. The cost of illness treatment at health centers is also covered. A 20 percent copayment is charged, but some groups are exempt. Information in a World Bank report was used to assess the social and financial functioning of the Bwamanda health insurance plan in Democratic Republic of Congo (Shaw and Ainsworth 1995). The assessment results pertaining to the affordability criteria, payment schedules, revenue and expenditure, and associ- ated access to health care are presented below. Bwamanda: Risk Protection Affordability of premiums and copayments. Absolute unaffordability did not appear to be a problem. It was reported that, in setting the premiums, the health staff Experience of Community Health Financing in the African Region 181 used the price of two kilograms of soybeans as a measure of affordability. In addi- tion, only 16 percent of 21 nonmembers stated that they had not joined the plan because of inadequate cash. However, the combined costs of the travel and premium may have been unaffordable for some households: when in 1988 the copayment rate was lowered for those living more than 25 kilometers from the hospital, the enrollment declined less with distance than in other years. In 1987­95, the percentage of the target population enrolled ranged from 41 per- cent to 66 percent (Criel and Kegels 1997). Appropriateness of payment schedules. Communities elected to pay premiums during the months that followed the second harvest period (March to April). The explanation provided by the authors was that after the first harvest, cash was needed for school expenses and therefore appropriateness of time of collect- ing was ensured by consulting the community. Bwamanda: Resource Mobilization Revenue from premiums and expenditure for benefits. The financial data for the insurance plan showed that in 1987 and 1988, premiums and interest were equivalent to 145 percent of the expenditure on care for beneficiaries. As a con- sequence, the share of the hospital's operating costs recovered rose from 48 per- cent in 1985 to 79 percent in 1988. Resulting level of access to health care among the population. It was reported in the World Bank study that insured persons were 6.7 times more likely than the uninsured to be hospitalized, implying that access to care was greater for the for- mer group. The fact that more than 60 percent of the zone's population was insured implies a significant increase in access to care as a result of the insurance plan. Even so, households living close to the hospital were also more likely to be insured, and therefore their high access and utilization may have been due to good physical access rather than the insurance plan. The combined effects of adverse selection and moral hazard were also felt to account for the high utiliza- tion. Even so, the prevalence of adverse selection and moral hazard was likely reduced by the inclusion of copayments and the requirement that all household members join the plan. However, other researchers working in Democratic Republic of Congo reported the rate of hospitalization among the insured to be only slightly higher than that of the uninsured (Moens 1990). Dangme West Health Insurance Scheme (Dangme Hewami Nami Kpee), Ghana Empirical research initiated by the Ministry of Health4 and undertaken from 1993 to 1995 in a rural district in southern Ghana concluded that household preferences and WTP were compatible with high membership in, and satisfac- tory performance of, a proposed health insurance scheme. The research findings led a health ministry team to work with households in Dangme West District to design and implement the Dangme West Health Insurance Scheme (Dangme 182 Health Financing for Poor People: Resource Mobilization and Risk Sharing Hewami Nami Kpee). The scheme is a collaboration between a mutual society and government health providers at the district level, and therefore it is a mutual-provider partnership model. The scheme was part of research carried out to improve the quality of health care provided by Ministry of Health facilities and increase access of households, and to evaluate the outcome of the interven- tion and draw policy-relevant lessons. Quality-related objectives were to be addressed in three ways. The first was by providing in-service training of health workers to improve technical competence in patient diagnosis and treatment, drug supply and management procedures, and interpersonal skills. The second was supportive supervision to ensure the use of skills acquired during training. The third was the refurbishment of health ministry facilities to ensure that basic physical and laboratory investigations could be carried out. The health insur- ance scheme was the main strategy to increase access by eliminating user fees for participating households. As a result of consultations with the district assembly (the local government for the district) and a series of durburs (community meetings) attended by approximately 4,250 adults (7 percent of the district's adult population), a mutual-provider partnership model for the scheme emerged. The model consists of: (a) a solidarity association named Dangme West Hewami Nami Kpee (literally translated, "a good health group"; registered members of the scheme become association members); (b) the Dangme West District Ministry of Health responsi- ble for providing health care in the district through the public health facilities; and (c) the district insurance management teams (DIMTs), ultimately consisting of staff of the district health ministry and elected members of the association. The DIMT is responsible for financial and administrative matters, including allo- cating insurance funds to health centers and hospitals using an approved formula and monitoring performance to ensure that paid-up members of the association have access to good quality health care at hospitals and clinics without paying fees. Prior to the election of members of the association to the DIMT, implemen- tation and management is the responsibility of an interim team consisting of the district assembly members and the district health administration members. The first registration period for the scheme was October 2000 to January 2001. Enrollment was by household and the contribution for a household was calcu- lated using a flat rate per adult (12,000 cedis [approximately US$1.80]) and per child (6,000 cedis). As adjustments were not made for increased earnings and cost of living, in real terms the contribution rate was lower than households' WTP in 1998. Members are entitled to outpatient care at any of the district health centers and clinics and to inpatient care at secondary hospitals in the regions and nearby hospitals. Although the implementation team (with the assistance of a local information technology [IT] group) developed software to integrate registration and identification of members with an information man- agement system, plans to install appropriate inputs were abandoned due to resource constraints. Two local donors declined to support a proposal endorsed by Ministry of Health headquarters for US$28,560 to fund the required physical Experience of Community Health Financing in the African Region 183 inputs and training for four years of operations (Arhin-Tenkorang 1999). The goal of the proposal was to provide effective management and IT solutions with respect to compatibility with existing processes, hardware affordability, and technical skill requirements. The inputs included hardware and specialized ID software. Consequently, registration with manual production of registration booklets was carried out using a Polaroid camera. Immediately following the first registration period, it became evident that a manual system of registration, iden- tification of members at health facility centers, and collection of claim informa- tion would be inefficient and not cost-effective. Dangme West: Risk Protection Affordability of premiums. The data obtained from the target population in 1998, 18 months before the first premium collection, indicated that 35 percent of households would be willing and able to pay the contribution rate. The first col- lection of premiums was undertaken between October and January--a time when cash is tight in farming households. The resulting enrollment rate of 5 per- cent of the target population is evidence that financial protection was subnor- mal, due in part to the low affordability of premiums. Appropriateness of payment schedules. The period for collecting the first premi- ums, determined by scheme officials, appeared to be incompatible with cash availability at that time. However, the officials' decision to have contributions paid annually was in line with the preferences found in a household survey con- ducted prior to implementing the scheme. Dangme West: Resource Mobilization Revenue from premiums and expenditure for benefits. As of July 2001, six months after the first premium payments by enrolled households, officials of the scheme had yet to compile and analyze the financial data. However, in view of the mod- est enrollment rate, it is reasonable to assume that resource mobilization will be appreciable but suboptimal. Performance Overview of Selected "Potentially Large Population" Schemes Table 4.4 provides information relating to the second-best indicators of financial risk protection (as proposed in the conceptual section) for the selected schemes. The final column contains the author's assessment of financial risk protection (FRP) pro- vided by the scheme, as inferred from the information. The assessments are based on these indicators because data were unavailable on illness incidences, incomes, and the prices of health care and other goods faced by targeted populations concep- tualized as necessary for direct assessment of FRP. One scheme (the Community Health Fund) was assessed as providing modest financial risk protection. The risk- protection performance of two schemes (the Abota Village Insurance Scheme and (FRP) Financial of provided provided provided provided provided provided Protection FRP FRP FRP FRP FRP FRP Risk Assessment Significant Modest Significant Suboptimal Suboptimal Suboptimal for care patients health access of uninsured responsible of are admissions 2.4 villages in 6.7 = = Level care in episodes increase Increased--insured/uninsured illness Increased--75% receiving Increased--primary available Increased--insured/uninsured admissions 9% Increased--insured/uninsured admissions when when when the care care care on (FRP) package inpatient inpatient inpatient of facility Completeness and and at care care and Protection clinic clinic clinic Full referred ariable--dependentV services Full referred Inpatient Inpatient Full referred Risk Performance and Financial by by of cash November first schedules chosen cash of time Risk-Protection Indicators between at appropriate Appropriateness payment of Generally Appropriate--determined villagers Inappropriate--low availability December Appropriate--time community Inappropriate--low availability enrollment Financial in of (first of province to % 1 workers but from Schemes' enrolled 77% (in studied as implies enrolled far 2000) and two-thirds health about majority--includes majority--level villagers households by 1998) approximately to living of enrolled 1992) to mechanism districts to by and assessed premiums unaffordable (1987­1995) (in to period in 2 27% households 5% population ever (in data of Population" Affordability target Affordable 54% studied Affordable exemption 5.3% 1999) Affordable determined Premiums unaffordable households approx. Affordable copayment households hospital 41%­66% WTP affordable 35% Approx. enrollment · · · · · · · · · · · Large (CAM) Ghana by Congo( by Nami (CHF), Scheme, benefit of Financing Insurance government) Insurance "Potentially Maladie Fund model, model, (provider by Scheme, Hewami Insurance (mutual Republic Health (mutual-provider Health model) Hospital model) 4.4 Burundi model, (mutual-provider Community est insurance Dem. d'Assurance illageV model) insurance W (Dangme Insurance Ghana ABLET Carte program, insurance Community anzaniaT partnership Abota Guinea-Bissau society Nkoranza Health (provider hospital) Bwamanda Scheme, provider hospital) Dangme Scheme Kpee), partnership 184 Experience of Community Health Financing in the African Region 185 the Carte d'Assurance Maladie) was assessed as being superior to that of the CHF as they provided significant protection. Common to these three schemes are benefit packages that include outpatient and inpatient care, although the range of inpatient provision in the CHF depended on the facility with which an insured member was registered. These schemes also set premium or contribution levels that appeared to be affordable to a high percentage of their target populations and can be interpreted as being compatible with the populations' WTP. Two of the three schemes also involved significant decisionmaking by the target population. In the literature on the selected schemes, increases in the resources for health care that resulted from the insurance schemes (postulated to be the appropriate direct measure of resource mobilization in the framework presented earlier) were unavailable for the selected schemes. With the exception of information permit- ting estimation of the ratio of expenditure to revenue, information about the second-best indicators of resource mobilization (notably, combined changes in utilization rates of insured and uninsured and the ratio of "average expenditure on an individual" to "average individual contribution") was incomplete or absent. Therefore, assessments of resource mobilization of these schemes were not attempted. DISCUSSION Institutional and Technical Influences on Scheme Design The schemes selected for review in this chapter increased the consumption of health care by insured persons relative to the uninsured. The levels of increased consumption, when combined with affordable premiums, suggest that for three of the schemes, the financial risk protection provided is modest to significant. The low participation rate of the target populations in all the schemes strongly suggests that the level of resource mobilization achieved is below the potentially attainable. Some of the evidence supports the conclusion that the low perfor- mance in resource mobilization may be traced to institutional factors that influ- enced the design of these schemes and their modalities of operation. Secondary data on household preference and WTP were not available, and with the excep- tion of the Dangme Scheme, primary collection of such data was not under- taken. Contribution calculations were based largely on assumptions, and consequently, in almost all the schemes, low affordability among the target pop- ulation was a major hindrance to participation. In general, WTP data are rarely collected or used as part of the process of designing health insurance schemes in developing countries. Yet contingent val- uation theory and empirical evidence suggest studies could be undertaken in developing countries to obtain valid and reliable health-related WTP data (Rus- sell, Fox-Rushby, and Arhin 1995). Measures that may be important in increas- ing the validity and reliability of WTP studies of the demand for health insurance in developing countries include the following items (Arhin 1999). 186 Health Financing for Poor People: Resource Mobilization and Risk Sharing First, a careful questionnaire design to limit the problems of hypothetical and strategic bias. In addition to the respondents' contingent demands, two other areas of inquiry must be included in a WTP questionnaire. These are past prac- tices and expenditure for similar and substitute goods and respondent socioeco- nomic characteristics. A plausible description of the hypothetical market must precede contingent demand questions. In this regard, guidelines emerge on the appropriate structure and content of open-ended and closed questions that would create realistic scenarios. Second, the undertaking of an initial qualitative study to ensure that the WTP questions use words and phrases that are consistent with the discourse on health, health care spending, and financial transactions used by the study subjects. Third, the careful training of interviewers to administer the questionnaire accu- rately. Differences between the interviewers and the respondents in their level of exposure to, and the discourse about, insurance may compromise validity and reliability. The problem is most acute if the interviewer must translate questions from an official language in which the questionnaire is written into a local dialect. Respondents should be given adequate time to consider the hypothetical good. Where possible, questions should be asked and repeated in an interview subsequent to that in which the hypothetical good was first described to the respondent. Whereas the concept of achieving compatibility between WTP and contribu- tion levels has not been a driving force in scheme design, achieving financial sus- tainability through cost-recovery goals appears to have been a major influence among designers and implementers when estimating contributions and design- ing benefits. Greater emphasis on the social functions of risk protection provided by the insurance would have dictated lower contribution rates, more comprehen- sive benefit packages, and measures to increase physical access to benefits. Low participation among the target population of the Nkoranza hospital scheme in Ghana suggests that hospital-based schemes may have the problem of being attractive mainly to people living close to the hospital and, as a result, achieve low enrollment. The subscription to the scheme appears not to depend on income disparity but on the differences in the direct and indirect costs of travel- ing to the one district hospital operating the scheme (Noterman and others 1993). As predicted by theory, achievement of financial and social goals in these reported schemes appears to mirror enrollment rates. Enrollment rates, in turn, reflect the target population's WTP for insurance plans offered by the schemes. Setting priorities for risk-protection objectives is not incompatible with goals that relate to increased resources for the supply of services. Detailed costing of health care provided by government and not-for-profit private facilities, and subsequent comparison of costs with government funding and user fee revenue, may be undertaken to provide information on the resource gap to be filled by contributions. In planning the Dangme Scheme and in evaluating the CAM scheme, analyses of this type were undertaken. The findings suggested that mobilization of resources through contributions would exceed, or exceeded Experience of Community Health Financing in the African Region 187 FIGURE 4.3 Premiums, Participation, and Revenues: Predictions for Option C Revenue (million cedis) Percent participation 250 100 participation rate 200 80 150 60 100 40 50 revenue 20 0 0 1 5 9 13 17 21 25 29 33 37 41 45 49 Household premium (x 1,000 cedis) respectively, that realized by user fees, even when the majority could afford the premiums. Given recent accelerated rises in health care costs as a result of the HIV/AIDS crisis and the sharp decline in real earnings of households in low- income countries, such predictions of revenues and costs will become critical in the design of schemes. The demand curve and revenue curve estimated by initial research that guided the Dangme Scheme illustrate the relationship between affordability of premium enrollment rates and revenues. Figure 4.3 shows the expected partici- pation rates for a benefit package that combines inpatient and outpatient cover- age (referred to as option C; options A and B are, respectively, inpatient and outpatient packages). The figure provides the predicted revenues (assuming a district population of 150,000) at different levels of household contributions based on the demand curve from WTP data. The highest revenue is predicted at a premium level of approximately c25,000 and corresponds to a low participa- tion rate of 35 percent of the population. Above this premium level, both total revenue and the participation rate start to decline. If both revenue from the scheme and utilization of health care by individuals (hence participation) enter into the utility function of decisionmakers, setting the household contribution level above c11,000 would not be rational. The fall in participation would not be compensated by the rise in revenue, and total utility would start to fall. How- ever, below this level, a particular utility may be obtained with different premi- ums and therefore different participation rates. The optimum premium will then depend on the relative values of revenue and utilization and hence the relative revenue and participation. Introducing a concern for consumers' utility and tak- ing into account their income constraints will further lower premiums. 188 Health Financing for Poor People: Resource Mobilization and Risk Sharing FIGURE 4.4 Willingness to Pay for Adult Insurance Annual contributions (cedis) 45,000 40,000 35,000 30,000 25,000 20,000 15,000 10,000 5,000 0 0 20 40 60 80 100 Percent of households Institutional factors also have a bearing on the technical methods used to design and simulate the scheme performance. The access to, and the tradition of consulting, reports, and documents on technical issues relating to insurance are examples of factors. The disciplines from which the implementers originate will also influence the rigor with which health, epidemiological, social, and eco- nomic data will be sought and analyzed to influence design and management of the scheme. Simulation of revenue outcomes of a proposed scheme is an exam- ple of an activity that will be more readily undertaken if appropriate disciplines are used in the design and management process. Figure 4.4 illustrates an exam- ple of simulation activity that relies on planner and economic disciplines. For a proposed scheme in a district in Ghana, contingent evaluation techniques pro- vided estimates of the maximum amounts that household heads would be will- ing to pay as an annual health insurance premium per adult and per child. Taking into account consumers' tendency to understate the maximum, in the hope that their contribution will be low, estimates were based on a contribution rate of c10,000 per adult per year. Deducting administration costs of c2,500 gave an estimated net annual scheme revenue of c250,200,000 (55.9 percent of 1998 district expenditure). In 1998, government health services in the study district were funded from two sources: the government of Ghana and user fees obtained from patients, known as internally generated funds (IGF). IGF accounted for 3.21 percent of total funds utilized. The net revenue estimate from the moderate contribution rate would be equivalent to the IGF that would be obtained from 64,584.41 episodes of outpatient care (in 1998 the actual number of outpatient episodes treated was only 11,124). Experience of Community Health Financing in the African Region 189 Institutional Factors and Policy Environment Institutional factors influencing health care providers' behavior, national policy formulation, and donor reactions to local initiatives appear to be important determinants of scheme performance. In several schemes, the initiator was external to the community to which the target individuals or households belonged, leading often to increased external financial support but also to increased pressure to implement the scheme quickly to conform to the project time frame. In these circumstances, primary data collection about illness experi- ence and health care costs required to simulate the operations of the scheme were not carried out. The Dangme Scheme, however, is an example of local initi- ation and conceptualization complemented by substantial primary data collec- tion and analysis. Work was undertaken by the implementation team with the support of a full-time economist and health planner and illustrates the time- consuming nature of activities required to support innovative schemes. The familiarity of donors with the subject of health insurance is decisive in determining the outcome of schemes. Unfamiliarity may lead a local donor to underestimate the complexity of inputs required to support an emerging scheme, particularly technical advice and data management systems, as in the case of the Dangme Scheme. As available evidence for best practice is scarce, donor funding procedures and regulation need to be flexible to support experimentation by communities, local governments, and local NGOs. These actors will require support to implement, monitor, and evaluate schemes. An external agency's inability to support both intervention (implementation) and research (monitoring and evaluation) of schemes adds significant transaction costs, as local actors must secure additional support from another donor. Although the existence of user fees gives people a strong incentive to enroll in insurance schemes, the institutionalization of user fees within health care providers' practices presents an obstacle to the effective implementation of insurance schemes. Introduction of a scheme is likely to alter the power relationship between the provider and the individual consumer and requires changes in organizational culture and behavior. The provider may have to negotiate and contract with patients who interact with health care institutions as a collective group and as direct sponsors. How an implementing team manages such changes will determine the level of trust between actors in the scheme. Orientation periods will be required to introduce health workers to their new or additional roles arising from an insurance scheme. Institutional factors in the policy formulation arena may hamper systematic progress in obtaining consensus about national or local priorities and objectives for health insurance. In particular, the absence of a culture of consumer dialogue and marketing is a major disadvantage for policy formulation in this area, as con- sumer reactions govern the success or failure of a policy. A mismatch between the expectations of policymakers and of the general public will result from institu- tional emphasis on resource mobilization at the expense of financial protection. 190 Health Financing for Poor People: Resource Mobilization and Risk Sharing Organizations and institutions formed to provide health insurance to house- holds in the informal sector ideally should be accountable not only to scheme members, through officials of relevant mutual groups, but also to civil society in general. Civil society in this case refers to groups that already provide social ser- vices, including health care, internationally or nationally--for example, church- related health providers. In many African countries, mission health facilities are major health care providers, especially to low-income households and rural areas. As a result of their devotion and convictions, these faith-based health providers have proven to be accountable to the people, and their reputations are untarnished by corruption. Policy Implications By the early 1990s, despite considerable involvement in health-financing policy formulation in Africa, many of the international agencies had failed to encour- age appropriate insurance-based alternatives to fee payment at the point of use. In particular, the 1993 World Development Report did not make recommendations for low-income countries that would change the situation in the short to medium term. By the end of the decade, the situation had changed significantly, leading to several international meetings focusing on health insurance for low- and middle-income countries. Following published data in the late 1990s, there is now increasing recognition of the vital role of insurance mechanisms in health systems (WHO 2000). Many national and international departments and agencies now accept that the principles of health insurance are applicable to low-income populations and are willing to study examples of insurance initia- tives for poor and informal households. The priority placed on health insurance within national health policy will partly determine the stage of financial protection attained, as illustrated in figure 4.5. In the figure, the triangular area represents "total population" health-related financial protection. The area below each of the levels approximates the magni- tude of the population effectively protected from the financial risks of ill health as a consequence of the level. For example, at the level of "predominance of out- of-pocket payments" for personal health services, common in low-income Sub- Saharan countries, a very small percentage of the population has protection. Conversely, true universal insurance coverage, as in Scandinavian countries, cor- responds to virtually "total population" protection. Policy thrusts placed below the level are those that will facilitate attainment of the level and those placed above will support efficiency and equity within the level. The appropriateness of launching a given policy thrust to establish health insurance schemes will be determined by the socioeconomic context. The role of national policymakers and donor agencies includes establishing the principle of disassociation between utilization or access and financial contributions. This will pave the way for strategic policies aimed at providing financial protection through insurance schemes. Policies supporting the substitution of health insur- Experience of Community Health Financing in the African Region 191 FIGURE 4.5 Stages of Financial Protection and Supporting Policies NATIONAL POLICY DONOR POLICY Stages Renewed policy commitment Universal insurance coverage Optimized subsidy of low- income households by high- income households Insurance pool consolidation Interpool subsidies and Advocacy and consolidation policies consumer protection Established funding and reinsurance insurance Framework for pool pools management and interactions Set-up funding and reinsurance Community health insurance Capacity building and Commitment to dissociation of technical support contribution from utilization Evidence-based advocacy Dominance of out-of-pocket payments ance for out-of-pocket fees are therefore an initial requirement. Given the socio- economic diversity among low-income populations, in the short term, contribu- tions into subnational insurance pools offer the greatest possibility of financial and administrative feasibility (Arhin 1995a). HMO-type arrangements should be the goal, rather than third-party schemes, since the former has the potential to stabilize health care costs and thus maintain low premiums. Nevertheless, in many low-income countries, voluntary contribu- tory schemes will not have the revenue potential to fund all the health service costs for their members. Significant central government funding (fiscal transfers and budget allocations) must therefore be reinstituted in some cases, for example in China, or maintained in others, as in Ghana. Being key stakeholders in many schemes, donors can be instrumental in establishing subnational schemes by pro- viding start-up funding and reinsurance guarantees, where appropriate. Subnational insurance schemes, in which solidarity organizations form partner- ships with providers, and that meet cost containment and quality requirements, 192 Health Financing for Poor People: Resource Mobilization and Risk Sharing will represent important collaboration with civil society. Policy to support such collaboration must provide the regulatory framework (legal, financial, and infor- mational) for scheme management and interactions with other parts of the health system. This includes enabling acquisition of appropriate human resources to counteract "capture" tendencies by health workers and profit-driven private investors and to monitor financial and social outcomes of the individual schemes. Donors have significant roles to play in capacity building to support these goals. Where logistical inadequacies hinder active promotion of progressive contri- butions, as in informal sector schemes, the overriding objective should be maxi- mization of enrollment rather than revenue, and sanctioning community-rated premiums that most of the targeted households can afford. These considerations and the inability to consolidate pools into a few large pools will necessitate rein- surance for some subnational schemes. Donors as external stakeholders have crucial roles in assisting low-income countries to meet the need for reinsurance through sectorwide approaches. CONCLUSION In the foreseeable future, it is unlikely that either centralized government or large commercial schemes, as found in rich countries, can provide near-universal health insurance cover for people in Africa, most of whom live in rural areas. Yet despite the endorsement of community health financing by many African coun- tries with the support of international agencies, few large-scale community- based health insurance schemes have been attempted. Therefore there are only a small number of experiences of "potentially large population" health insurance schemes in Africa upon which to base conclusions about their performance. Nevertheless, the performance of the schemes studied suggests that insurance mechanisms in some situations, have, or may eventually have, a significant membership rate among the population and therefore are capable of increasing access to health care and, to a limited extent, mobilizing resources. Constraints that have hindered attempts to design appropriate community insurance schemes include the policy environment, inadequate administrative infrastructures, and a shortage of trained staff to manage schemes. This under- scores the need for information on the feasibility of schemes designed for rural populations. Data on WTP will be a critical part of such information since WTP will determine demand and the relationship between revenue and expenditure, hence social and financial performance. Affordability of premiums and appro- priateness of payment schedules of existing rural health schemes suggest that WTP is substantial for such schemes. Prospects for achieving resource mobilization effects that are significant relative to total health care costs are declining because of decreasing purchasing power in developing countries and the rising incidence of HIV/AIDS, TB, and other infectious diseases. Recognizing that the low income of households in the informal sector will Experience of Community Health Financing in the African Region 193 BOX 4.1 GHANA'S POLICY THRUSTS TO ENABLE EVOLUTION OF COMMUNITY HEALTH INSURANCE ". . . expediting the establishment of health insurance or prepayment mechanisms will be critical to providing financial access." ". . . the health policy goals are best served by a multischeme health insurance system." Ghana's Programme of Work 2000­02 Ministry of Health, 1999 After two decades of intent to include insurance in its health financing strat- egy, during which several policy-oriented research projects and feasibility stud- ies were commissioned, a Ministry of Health policy framework has emerged promoting a "multischeme health insurance system." The goal is a system that embraces both the formal and informal sectors, providing affordable health insurance to the majority of Ghanaians and exemptions for indigents. Insur- ance schemes for the rural population present the greatest challenge. The framework envisages that "community-based health insurance schemes and mutual health organizations for the informal rural sector" will be a component of the multischeme health insurance system that will meet this challenge. The Planning, Monitoring, and Evaluation Department of the MOH is charged with facilitating and supporting districts and regions that wish to estab- lish insurance and prepayment schemes for the populations they serve. In this new policy framework, prospects have magnified for a district initiative in Dangme West, for example, to provide policy lessons that will inform the process of developing and implementing health insurance in Ghana. In addi- tion, the availability of technical expertise to achieve this aim has permitted the initiative to undertake baseline studies that will be used to evaluate the scheme. These include household surveys of willingness to pay (WTP) for health insur- ance, studies of district health accounts, and studies of drug prescribing and stock levels of district health insurance. Future health policy on insurance will have the benefit of evidence from this and other subnational schemes. result in WTP insufficient to fund all the health care costs for scheme members, cen- tral government support of these schemes in the form of fiscal transfers and budget allocations will be required. Furthermore, schemes in which solidarity organizations form partnerships with providers create incentives for increased efficiency and accountability and should be supported by national government policies through appropriate legal, financial, and informational mechanisms. In the absence of estab- lished best practices in the design of schemes, donor-funding procedures and regu- lations need to be flexible to assist experimentation by communities, local governments, and local nongovernmental organizations. Schemes operated by NGOs (especially church-related providers or civil society in partnership with local authorities, such as district health authorities) may be effective and efficient recipi- ents of donor funds (for example, matching funds) in the short term and may pro- vide the basic structure for future national insurance schemes in the medium term. 194 Health Financing for Poor People: Resource Mobilization and Risk Sharing Acknowledgments: The author is grateful to the World Health Organization for having pro- vided an opportunity to contribute to the work of the Commission on Macroeconomics and Health and to the World Bank for publishing the material in this chapter as an HNP Discussion Paper. NOTES 1. The primary participation rate is that of the individual reference scheme; the secondary participation rate refers to the total membership of a reference group of linked schemes as in the case of insurance funds of several small schemes. 2. In third-party insurance, scheme officials are neither pool members nor providers of health care benefits, as in the case of a company whose sole function is to collect con- tributions and pay claims to patients or directly to hospitals and clinics. 3. The commune is the country's lowest unit of local administration. 4. The principal investigator was a health economist from the Ministry of Health, sec- onded to an academic health economics group in the United Kingdom. The head of the ministry was one of the research advisers. REFERENCES Abel-Smith, B., and P. Rawal. 1994. "Employer's Willingness to Pay: The Case for Compul- sory Health Insurance in Tanzania." Health Policy and Planning 9(4): 409­18. Arhin, D. C. 1991. "Community Health Insurance in Developing Countries and Its Feasi- bility in Ghana." Research proposal submitted to UNICEF and the International Devel- opment Research Centre (IDRC). ------. 1994. "The Health Card Insurance Scheme in Burundi: A Social Asset or a Non- Viable Venture?" Social Science and Medicine 39(6): 861­70. ------. 1995a. "Health Insurance in Rural Africa." The Lancet 345: 44­45. ------. 1995b. "Health Insurance in Sub-Saharan Africa: What Are the Options?" Paper presented at the Symposium on Health Care Financing at the European Conference on Tropical Medicine, October 22­26. Hamburg, Germany. ------. 1995c. Rural Health Insurance: A Viable Alternative to User Fees? Public Health and Policy (PHP) Departmental Publication 19. London School of Hygiene and Tropical Medicine (LSHTM). ------. 1996a. Book review of "Community Participation in Primary Health Care," by E. Alihonou, S. Inoussa, L. Res, M. Sagbohan, and C. M. Varkevisser. Social Science and Medicine 42(4): 629. ------. 1996b. "Health Insurance Demand in Ghana: A Contingent Valuation." Paper pre- sented at the International Health Economics Association (IHEA) Conference, May 19­23. Vancouver, Canada. ------. 1997. "Are People in Ghana Willing to Pay for Health Care?" Paper presented at the Fifteenth Annual Conference of the German Association of Tropical Paediatricians, January 23­25. Kiel, Germany. Experience of Community Health Financing in the African Region 195 ------. 1999. "Using WTP Instruments for Health Insurance in Low Income Countries." Unpublished paper. Health Economics and Financing Programme at LSHTM. Arhin, D. C., and S. Adjai. 1997. A National Health Insurance System for Ghana: A Concept Paper and Proposal. Health Research Unit, Ministry of Health, Accra, Ghana, and LSHTM, London. Arhin-Tenkorang, D.C. 1999. "A Proposal to Set Up an Integrated Registration and Infor- mation Management System and Train Health Facility Staff and MOH District Admin- istration Staff in Its Application." Submitted to the Ministry of Health (MOH), Ghana. ------. 2000. "Mobilizing Resources for Health: The Case for User Fees Re-Visited." Paper prepared for the Third Meeting of the Commission on Macroeconomics and Health (CMH), November 8­10. Paris. Arrow, K. J. 1963. "Uncertainty and the Welfare Economics of Medical Care." American Economic Review 53(5): 941­73. Asenso-Okyere, W., I. Osei-Akoto, A. Amun, and E. N. Appiah. 1997. "Willingness to Pay for Health Insurance in a Developing Economy: A Pilot Study of the Informal Sector of Ghana Using Contingent Valuation." Health Policy 42(3): 223­37. Atim, C. 1998. Contribution of Mutual Health Organizations to Financing, Delivery, and Access to Health Care: Synthesis of Research in Nine West and Central African Countries. Technical Report 18. Partnerships for Health Reform Project, Abt Associates Inc., Bethesda, Md. Atim, C., and M. Sock. 2000. An External Evaluation of the Nkoranza Community Financing Health Insurance Scheme, Ghana. Technical Report 50. Partnerships for Health Reform Project, Abt Associates Inc., Bethesda, Md. Bennett S., A. Creese, and R. Monasch. 1998. Health Insurance Schemes for People Outside Formal Sector Employment. ARA Paper 16. WHO, Geneva. Bethune, X. de, S. Alfani, and J. P. Lahaye. 1989. "The Influence of an Abrupt Price Increase on Health Services Utilization: Evidence from Zaire." Health Policy and Plan- ning 4(1): 76­81. Bitrab, R., and others. 1987. "Health Zones Financing Study Zaïre." Report prepared for the Resources for Child Health Project, USAID, Arlington, Va. Booth, D., J. Milimo, G. Bond, and others. 1995. Coping with Cost Recovery. Report to the Swedish International Development Authority, Development Studies Unit, Dept. of Anthropology, Stockholm University, Stockholm. Brown, W., and C. Churchill. 2000. Insurance Provision in Low-Income Communities: Part II--Initial Lessons from Micro-Insurance Experiments for the Poor. Calmeadow, Microenter- prise Best Practices (MBP) Project, Development Alternatives Inc. Burgess, R., and N. Stern. 1991. "Social Security in Developing Countries: What, Why, Who, and How?" In E. Ahmad, J. Drèze, J. Hills, and A. Sen, eds., Social Security in Devel- oping Countries. Oxford: Clarendon Press. Chabot, J., and F. Savage. 1984. A Community Health Project in Africa. St. Albans, U.K.: Teaching Aids at Low Cost (TALC). Criel, B. 1998. District-Based Health Insurance in Sub-Saharan Africa. Antwerp: IGT Press. Criel, B., and G. Kegels. 1997. "A Health Insurance Scheme for Hospital Care in Bwa- manda District, Zaire: Lessons and Questions after 10 Years of Functioning." Tropical Medicine and International Health 2(7): 654­72. 196 Health Financing for Poor People: Resource Mobilization and Risk Sharing Department of Insurance at St. Theresa's Hospital. 1993. "Evaluation of First Year of Health Insurance for Nkoranza District." Unpublished report. Ghana. Eklund, P., and K. Stavem. 1990. "Prepaid Financing of Primary Health Care in Guinea- Bissau: An Assessment of 18 Village Health Posts." Working Paper Series 488. World Bank, Africa Technical Department, Washington, D.C. Ghai, D. 1988. "Participatory Development: Some Perspectives from Grass-Root Experi- ences." Discussion Paper 5. United Nations Research Institute for Social Development (UNRISD), Geneva. Hatch, M. J. 1997. Organizational Theory: Modern Symbolic and Postmodern Perspective. New York: Oxford University Press. Hill, C. W. L., and G. R. Jones. 1998. Strategic Management: An Integrated Approach. Boston: Houghton Mifflin. Knippeneberg, R., and others. 1991. "Joint Report on Progress Assessment of BI in Guinea Bissau." Unpublished draft. UNICEF and WHO. Lohr, K. N., R. H. Brook, C. J. Kamberg, G. A. Goldberg, A. Leibowitz, J. W. Keesey, D. M. Reboussin, and J. P. Newhouse. 1986. "Use of Medical Care in the Rand Health Insur- ance Experiment: Diagnosis and Service-Specific Analysis of Randomized Controlled Trial." Medical Care 24 (supplement): S1-87. Manning, W. G. 1987. "Health Insurance and the Demand for Medical Care: Evidence from a Randomized Experiment." American Economic Review 77: 251­77. Mathiyazhagan, K. 1998. "Willingness to Pay for Rural Health Insurance through Com- munity Participation in India." International Journal of Health Planning and Management 13(1): 47­67. McPake, B., and others. 1992. Experiences to Date of Implementing the Bamako Initiative in Africa. London: Health Policy Unit of the LSHTM. Moens, F. 1990. "Design, Implementation, and Evaluation of a Community Financing Scheme for Hospital Care in Developing Countries: A Pre-Paid Health Plan in Bwa- manda Health Zone, Zaïre." Social Science and Medicine 130(12): 1319­27. Musau, S. N. 1999. Community-Based Health Insurance: Experiences and Lessons Learned from East and Southern Africa. Technical Report 34. Partnerships for Health Reform, Abt Asso- ciates Inc., Bethesda, Md. Noterman, J., B. Criel, G. Kegels, and K. Isu. 1993. "A Prepayment Scheme for Hospital Care in the Masisi District in Zaïre: A Critical Evaluation." Social Science and Medicine 40(7): 919­30. Nyonator, F., and J. Kutzin. 1999. "Health for Some? The Effects of User Fees in the Volta Region of Ghana." Health Policy and Planning 14(4): 329­41. Pauly, M. V. 1963. "Taxation, Health Insurance, and the Market Failure in the Medical Economy." Journal of Economic Literature 24: 629­75. Russell, S., J. Fox-Rushby, and D. C. Arhin. 1995. "Willingness and Ability to Pay for Health Care: A Selection of Methods and Issues." Health Policy and Planning 10(1): 94­101. Experience of Community Health Financing in the African Region 197 Shaw, R. P., and M. Ainsworth, eds. 1995. Financing Health Services through User Fees and Insurance: Lessons from Sub-Saharan Africa. World Bank Discussion Paper 294. Washing- ton, D.C. Tanner, K. 1990. Impact of Structural Adjustment and Policy Choices for Promoting the Welfare of Children and Women in Guinea. Bissau: UNICEF. Vogel, R. J. "Health Insurance in Sub-Saharan Africa: A Survey and Analysis." 1990. Africa Technical Department of the World Bank (Working Paper Series 476), Washington, D.C. World Bank. 1987. Health Sector Memorandum. May 21. ------. 1999. Community Health Fund in Tanzania. Washington D.C. Video. World Health Organization. 2000. World Health Report 2000: Health Systems--Improving Performance. Geneva. PART 2 Country Case Studies Using Household Survey Analysis 5. Analysis of Community Financing Using Household Surveys Melitta Jakab, Alexander S. Preker, Chitra Krishnan, Pia Schneider, François Diop, Johannes Paul Jütting, Anil Gumber, M. Kent Ranson, and Siripen Supakankunti 6. Financial Protection and Access to Health Care in Rural Areas of Senegal Johannes Paul Jütting 7. Community-Based Health Insurance in Rwanda Pia Schneider and François Diop 8. The SEWA Medical Insurance Fund in India M. Kent Ranson 9. The Potential Role of Community Financing in India Anil Gumber 10. Impact of the Thailand Health Card Siripen Supakankunti CHAPTER 5 Analysis of Community Financing Using Household Surveys Melitta Jakab, Alexander S. Preker, Chitra Krishnan, Pia Schneider, François Diop, Johannes Paul Jütting, Anil Gumber, M. Kent Ranson, and Siripen Supakankunti Abstract: Objective: to provide empirical evidence regarding the performance of community-based health care financing in terms of social inclusion and financial protec- tion. Methods: five nonstandardized household surveys were analyzed from India (two samples), Rwanda, Senegal, and Thailand. Common methodology was applied to the five data sets. Logistic regression was used to estimate the determinants of enrolling in a community-financing scheme. A two-part model was used to assess the determinants of financial protection: part one used logistic regression to estimate the determinants of the likelihood of visiting a health care provider; part two used ordinary least-squares regres- sion to estimate the determinants of out-of-pocket payments. Findings: Social inclusion-- our findings suggest that community financing can be inclusive of the poorest even in the most economically deprived context. Nevertheless, this targeting outcome is not automatically attributable to the involvement of the community; rather it depends on key design and implementation characteristics of the schemes. Financial protection-- community financing reduces financial barriers to health care as demonstrated by higher utilization and simultaneously lower out-of-pocket expenditure of scheme members con- trolling for a range of socioeconomic variables. Conclusions: Social inclusion--design and implementation characteristics of community-financing schemes matter in achieving good targeting outcome; community involvement alone does not guarantee social inclu- sion. Further research is needed to delineate which design and implementation charac- teristics allow better inclusion of the poor. Financial protection--prepayment and risk sharing, even on a small scale, reduce financial access barriers. C ommunity-based health financing (CF) has been attracting increasing attention as a potential instrument for protecting low-income populations from the impoverishing effects of health care expenditures. Proponents argue that communities have incentives as well as instruments to reach the poor and the socially excluded. In contrast, general tax- and social insurance-funded health care structures often lack instruments to achieve close targeting of the poor and ensure their financial protection at the time of illness. Market-based organizations, however, lack incentives to promote their insurance products to rural populations as high transaction costs would translate into high and unaf- fordable premiums for the poor (Preker and Jakab 2001; Wiesmann and Jütting 2001; Dror and Jacquier 1999; Jütting 2002). 202 Health Financing for Poor People: Resource Mobilization and Risk Sharing However, many observers point out that community structures are not inher- ently inclusive of the poor either. Community structures may not necessarily reflect the views of the wider population, critical decisions may not take into account the interests of the poorest, and the poorest may not be involved in the decisionmaking (Gilson and others 2000). As recent thinking related to social capital suggests, communities can be as exclusive creating a gap between the "in- community" and "out-community" groups as they can be inclusive and provide a bridge for the disadvantaged (Narayan1999). The literature on community financing provides some insights into this debate, although the evidence is far from conclusive. A recent review of 45 arti- cles on community financing found that fewer than a dozen provided some indication of whether the reviewed schemes were inclusive of the poor and whether they were effective in protecting them from the impoverishing effects of illness (Jakab and Krishnan 2001). These studies suggest that CF is effective in reaching a large number of low- income populations who would otherwise have no financial protection against the cost of illness (Desmet, Chowdhury, and Islam 1999; Diop, Yazbeck, and Bitran 1995; Arhin 1994; Liu and others 1996; DeRoeck and others 1996; CLAISS 1999; Hsiao 2001). At the same time, several studies demonstrate that the very poorest are excluded from the financial protection benefits of CF schemes. The main reason appears to be that the schemes are not able to reduce the financial access barrier for the lowest income groups (McPake, Hanson, and Mills 1993; Criel, van der Stuyft, and van Lerberghe 1999; Atim and Sock 2000; Arhin 1994; Supakankunti 1998). Nonetheless, it is difficult to make far-reaching systematic conclusions about the impact of CF on preventing impoverishment based on these studies. Most of the reviewed studies did not have access to household data to assess the impact of scheme membership on beneficiaries. In the few cases in which household data were used, the studies faced methodological difficulties. The most important dif- ficulty in a cross-sectional setting is that the variation in membership status is endogeneous. This is due to the fact that enrollment in voluntary health insur- ance is driven by both observable and unobservable characteristics, and the latter are likely to be correlated with the observable explanatory variables. In most of the available studies, these issues are not properly addressed. In addition, the large variety of indicators reported prevents comparability. Furthermore, many studies report indirect measures of social inclusion and impoverishment due to health expenditures. Because of these limitations, there continues to be a need to provide empirical evidence explaining who is covered by CF schemes, why those are covered and not other groups, how effective CF is in terms of financial pro- tection, and what structural characteristics make certain schemes more inclusive and more effective than others. This study is an attempt to address some of these shortcomings. The study analyzes household data from four countries (India, Rwanda, Senegal, and Thai- land) with a common methodology. The study addresses two principal ques- Analysis of Community Financing Using Household Surveys 203 tions: Are the surveyed CF schemes in these four countries inclusive of the poor? Are the surveyed CF schemes in the four countries effective in providing finan- cial protection from the impoverishing effects of illness? This opening section of the chapter summarizes the methodology, synthe- sizes the results, and discusses the reasons behind commonalities and differ- ences in findings. It is followed by sections presenting the country reports. The second section provides background about the four countries, their health sys- tems, and the surveyed CF schemes. The third section describes the methodol- ogy used in the analysis. The fourth section presents the results. The fifth section discusses the findings. BACKGROUND The chapter synthesizes findings from five household surveys conducted in four countries: India, Rwanda, Senegal, and Thailand. Despite their varied socioeco- nomic and cultural contexts, all four countries have sizeable poor populations-- many in the informal sector--who lack access to government and social insurance-funded health care services. Community-financing schemes are pre- sent in the four countries and aim to fill the gap in financial protection for the poor against the impoverishing effects of illness. This section reviews the socio- economic background of the four countries, their health systems, and character- istics of the community-financing schemes surveyed. Socioeconomic Background Three of the four countries surveyed are classified as low-income countries with a per capita gross national product (GNP) ranging from US$250 in Rwanda to US$510 in Senegal. Thailand is a lower middle-income country, with a per capita GNP of US$1,960. The four countries have varied sociodemographic characteris- tics (see table 5.1). All four countries are characterized by sizeable poor populations. Rwanda, one of the world's poorest countries, counts 70 percent of its population as falling below the national poverty line. In Senegal, 26 percent of the population (2.4 million people) lives in absolute poverty on less than a dollar a day. In India, 44 percent of the population (439 million people) lives in absolute poverty. Even in Thailand, the only middle-income country in the sample, 13 percent of the population lives below the national poverty line, and about 1 million people still live in absolute poverty Health System There are similarities in the health system of the four countries: a large proportion of the population is not covered by prepayment schemes, and access problems are 204 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 5.1 Socioeconomic Characteristics of Rwanda, Senegal, India, and Thailand, 1999 Rwanda Senegal India Thailand Population (millions) 8.3 9.3 997.5 61.7 Urban population (percent of total) 6 47 28 21 Labor force (percent of total population) 48 43 44 60 Adult illiteracy rate 29 (m) 55 (m) 33 (m) 3 (m) (percent of people 15+) 43 (f) 74 (f) 57 (f) 7 (f) GNP per capita (US$) 250 510 450 1,960 Population living below US$1 per day (percent) -- 26.3 44.2 < 2 (1997) (1998) Population below national poverty line (percent) 70 -- 35 13.1 (1997) (1995) (1994) (1992) Source: World Bank (2000). reported. With the exception of Thailand, the countries are classified as high child and high adult mortality rate stratum by the World Health Organization (WHO). Health spending in the three low-income countries is US$13 to US$23 per capita. Despite the much higher level of per capita spending, Thailand channels compa- rable amounts of resources through out-of-pocket payments (see table 5.2). Rwanda The Rwandan health sector has a three-tier structure, consisting of the Ministry of Health (MOH) at the central level, 11 health regions, and 38 health districts. Medical care is provided in two public referral hospitals and one private hospital, 28 district hospitals, and 283 health centers. On average district hospitals cover an area of 217,428 inhabitants, and health centers serve a population of 23,030 individuals. The Rwandan government remains the major provider of health ser- vices, with religious organizations as partners, especially in rural areas. The role of private providers is limited. There are only two health insurance companies. They insure about 1 percent of the Rwandan population, including 6 percent of the formal sector employees. Most employers contract with providers directly to ensure care for their employees. Financial barriers restrict access to medical care for the poor, who are excluded from formal sector employment. The Rwandan health sector is financed by foreign assistance (50 percent) and by private out-of- pocket spending (40 percent); the government contributes only 10 percent of total funds. The genocide in 1994 was followed by a period of humanitarian assistance. During this time, public health care services were provided free to patients, financed by donors and the government. In 1996, the Ministry of Health rein- troduced prewar-level user fees in health facilities. By 1999, utilization of pri- mary health care services had dropped from 0.3 in 1997 to a national average of Analysis of Community Financing Using Household Surveys 205 0.25 annual consultations per capita. This sharp drop in demand for health ser- vices combined with growing concerns about rising poverty, poor health out- come indicators, and a worrisome HIV prevalence among all population groups motivated the Rwandan government to develop community-based health insur- ance to assure access for the poor to the modern health system. Senegal In Senegal, as in most other African countries, large proportions of people are not covered by formal health insurance, and access problems are reported in terms of financing and geographic outreach. Social health insurance, introduced in 1975, extended coverage to private sector employees and their families. The added coverage of social insurance and the partial coverage of civil servants and their dependants, however, still leave most of the population in the urban infor- mal sector and rural sector underprotected against the financial risks associated with illness. The Senegalese health care system has three different levels: health district, regional, and central level. The health district has a health care center as well as health posts. Senegal has 50 health districts run by a chief health doctor (1998). The regional level is attached to the administrative division of a region, and the central level is attached directly to the Ministry of Health. The Senegalese health sector is financed by the central government (about 50 percent), user fees (about 10 percent), local government (about 6 percent), and donors (about 30 percent). In Senegal, the private sector plays an important part in the health care sys- tem due to both its size and its geographic distribution. Private providers are a mix of for-profits, serving urban high- and middle-income groups and charging relatively high fees, and nonprofits, mostly church-run facilities, serving rural and poor populations and charging only modest fees. Company clinics are also important. In Senegal there are about 40 private clinics (1994), about three- quarters of them located in Dakar; half of these private facilities are mainly maternity clinics. There are also about 14 diagnostic labs, 11 of them in Dakar. In the nonprofit sector, the Catholic Church health posts (about 70, mainly in rural areas) and the Catholic hospital (St. Jean de Dieu in Thiès) play an important role. The church deliberately put most of its nonprofit services in the rural areas to reach the otherwise excluded populations and the poor. The church network developed mostly in the 1950s and 1970s. Church-based providers are especially important in reaching rural areas with preventive services. India Health care in India is provided through general tax-funded public providers and insurance for the formally employed, and increasingly through nongovern- mental organizations (NGOs) and charitable institutions. 206 Health Financing for Poor People: Resource Mobilization and Risk Sharing Public system. There are concerns about access and use of subsidized public health facilities. Most of the poor households, especially rural, reside in remote regions where neither government facilities nor private medical practitioners are available. These people must depend heavily on services provided by local, often unqualified, practitioners and faith healers. In addition, wherever accessibility is not a constraint, the primary health centers are generally either dysfunctional or provide low-quality services. The government's claim to provide free secondary and tertiary care does not stand; in reality there are charges for various services (Gumber and Kulkarni 2000). Voluntary insurance. Only 9 percent of the Indian workforce is covered by some form of health insurance--the majority of those covered belong to the for- mally employed sector. Public insurance companies so far have paid very little attention to voluntary medical insurance because of low profitability and high risk as well as lack of demand. From the consumer point of view, the insurance coverage is low because information about the private health insurance plans is lacking, and the mechanisms used by the health insurance providers are not suitable to consumers. Nongovernmental organizations. NGOs and charitable institutions (not-for- profit) have played an important role in delivering affordable health services to the poor, but their coverage has remained small. The issue is how to reach the socially excluded and, more recently, how to ensure that the uninsured get min- imum affordable quality services. Thailand Health insurance schemes in Thailand can be classified into three types: welfare and fringe benefit, compulsory, and voluntary health insurance. Surveyed Community-Financing Schemes In all four countries, community financing plays a role in raising resources for health care and providing financial protection. In India, Senegal, and Thailand, community financing has had a relatively long tradition, while in Rwanda, it is a new phenomenon. The community-financing arrangements surveyed in the four countries are similar in that they are based on prepayment and risk sharing. Using the categorization developed by Jakab and Krishnan (2001), India, Rwanda, and Senegal fall under modality 2--community health funds and mutual health organizations--while Thailand belongs to modality 4--community financing with substantial government or social insurance involvement. Rwanda In early 1999, the Rwandan MOH in collaboration with local communities, and the technical and financial support of Partnerships for Health Reform (PHR), started the process to pilot test prepayment schemes in three health districts. At Analysis of Community Financing Using Household Surveys 207 the end of the first operational year, the 54 schemes had enrolled more than 88,000 members. Under Rwandan law, the schemes are deemed mutual health associations, headed by an executive bureau with four volunteers, elected by, and among, the scheme members during a community-based health insurance (CBHI) general assembly. On a district level, the schemes have formed a federation. Six members have been elected by, and among all, prepayment schemes' (PPS) executive bureau representatives in their general assembly to constitute the bureau of the district federation of prepayment schemes. The federation is the partner to the district hospital as well as to the health district and other authorities. Each pre- payment bureau has signed a contract with the affiliated health center, and each federation with the district hospital, defining in 17 articles the rules of collabo- ration between the insurer and provider. According to the schemes' bylaws, members are invited, at least once a year, to attend the prepayment scheme's general assembly. The CBHI schemes promote group membership. Households that would like to be insured pay, at the time of enrollment, an annual premium of 2,500 francs1 per family of up to seven persons to the CBHI affiliated with their "preferred" health center.2 An individual pays 2,000 francs upon enrollment. PPS member- ship entitles members--after a one-month waiting period--to a basic health care package covering all services and drugs provided in their preferred health center, including ambulance transfer to the district public or church-owned hospital, where a limited package is covered.3 Group enrollment and the one-month wait- ing period are designed in the voluntary CBHI schemes of Rwanda to minimize adverse selection. In case of sickness, members contact first their preferred health center, which is usually the nearest public or church-owned facility. Members pay a 100 francs4 copayment for each health center visit. Health cen- ters play a gatekeeper function, and hospital services are covered for members only if they are referred by their preferred health centers. This is done to dis- courage members and providers from frivolous use of more expensive hospital services. Senegal Senegal has had a long tradition of mutual health insurance schemes. The first experience began in the village of Fandène in the Thiès region in 1990. From the beginning, the movement in Senegal has been supported by a local health care provider, the nonprofit St. Jean de Dieu Hospital. Sixteen mutual health insur- ance schemes operate in the area of Thiès. The schemes' main features are: · They are community based. · Ninety percent of the schemes operate in rural areas. · With the exception of one mutual--Ngaye Ngaye--the schemes cover only hospitalization. 208 Health Financing for Poor People: Resource Mobilization and Risk Sharing · The mutuals have a contract with St. Jean de Dieu Hospital, where they get a reduction of up to 50 percent for treatment. · In general, the household is a member of a mutual and participates in the decisions. The insured has a membership card on which he or she can list all or selected family members (beneficiaries). The membership fee is per person insured. India The Self-Employed Women's Association's (SEWA's) Integrated Social Security Scheme was initiated in 1992. SEWA is a trade union of more than 2 million women, all workers in the informal sector. This integrated income-protection scheme provides life insurance, medical insurance, and asset insurance. Women who pay the annual Social Security Scheme premium of 72.5 rupees--30 rupees of which is earmarked for the Medical Insurance Fund (herein referred to as the Fund)--are covered to a maximum of 1,200 rupees yearly in case of hospitaliza- tion in any registered (private or public) facility. Women between the ages of 18 and 58 are eligible for membership in the Fund. Women also have the option of becoming lifetime members of the Social Security Scheme by making a fixed deposit of 700 rupees. Interest on the deposit is used to pay the annual premium, and the deposit is refunded when the woman reaches 58 years of age. Upon discharge from the hospital, members must first pay for the hospitalization out-of-pocket. They submit receipts and doctors' certificates to SEWA, and if the insurance claim is approved, they are reimbursed by check. Excluded from Fund coverage are certain chronic diseases (for example, chronic tuberculosis, certain cancers, diabetes, hypertension, piles) and "disease caused by addiction" (SEWA brochures, 2000). Throughout the 10 districts of Gujarat where it operates, the Fund had 23,000 members in 1999­2000 (this compares with coverage of roughly 150,000 women under the broader SEWA trade union, statewide). Thailand The Health Insurance Card Scheme was introduced in 1983. Its three main objectives are to promote community development under the primary health care program, foster the rational use of health services via a referral system, and increase health resources based on a community-financing concept. The target population is the near-poor and middle-income classes in rural areas or people who can afford the premium. The health insurance card costs bath 1,000 (US$40) per year per household of up to five members. A household contributes half of the price, and the other half is subsidized by general tax rev- enue through the Ministry of Public Health. The benefits include outpatient care for illness and injuries, inpatient care, and mother and child health services. There is no limit on utilization of the services. However, the beneficiaries can go only to health care provider units under the Ministry of Public Health. The first Analysis of Community Financing Using Household Surveys 209 TABLE 5.2 Health Outcomes and Expenditures in Rwanda, Senegal, India, and Thailand, 1999 Rwanda Senegal India Thailand Health outcomes U5MR 189 (m) 134 (m) 97 (m) 40 (m) 163 (f) 126 (f) 104 (f) 27 (f) Life expectancy 41.2 (m) 53.5 (m) 59.6 (m) 66 (m) 42.3 (f) 56.2 (f) 61.2 (f) 70.4 (f) Health expenditures Total (percent of GDP) 5 4.5 5.2 5.7 Total per capita (US$) 13 23 23 133 Out-of-pocket (as percent of total) 49.9 44.3 84.6 65.4 Note: U5MR means under-five mortality rate. Source: WHO (2000). Health expenditure figures for Rwanda: Schneider and others (2000). contact is at either the health center or the community hospital; patients must then follow a referral line for higher levels of care. There is a specific time for card selling at each health card cycle. At present the cycle is one year, and the specific time for card selling depends on the seasonal fluctuations in income. The premium is collected when cash incomes are highest (for example, when crops are harvested). In 1992, the population coverage by the health card program was 3.6 million, about 5 percent of the total population. METHODS Data Sources The use of standardized and nationally representative surveys (Living Standard Measurement Surveys [LSMS], Demographic and Health Surveys [DHS]) was explored, but they did not prove to be useful for the purpose of this study. Though preferred for their representative sampling and standardized measures of socioeco- nomic status, these large-scale surveys did not allow us to identify households with access to community-based health financing. Even where the survey included health-financing questions, coverage through community financing could not be separated from other health-financing instruments such as private insurance or social insurance. The appendix provides the complete list of survey instruments reviewed in 21 countries as well as the variables predefined as selection criteria. Eleven smaller scale nonstandardized surveys matched the requirements for the core list of variables. The five available for further analysis were included in this report. The other six were either impossible to access for further analysis (four), data collection was incomplete (one), or authors were not available to col- laborate within the short time frame of the project (one). 210 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 5.3 Characteristics of 5 Survey Instruments Year Sample size # of data of observations Organization associated Name of scheme collection (households) with the survey Rwanda 54 prepayment schemes in 3 districts 2000 11,583 individuals Partnerships for Health of Kabutare, Byumba, and Kabgayi 2,518 HH Reform (PHR) in collaboration with National Population Office Senegal 4 Mutual Health Insurance Schemes 2000 2,987 individuals Center for Development (Thiés Region) 346 HH Research (ZEF Bonn) in cooperation with the Institute for Health and Development India (1) Self-Employed Women's Association 1998­99 1,200 HH National Council of (SEWA) Applied Economic Research (NCAER) India (2) Self-Employed Women's Association 1998­99 1,200 HH London School of (SEWA) Hygiene and Tropical Medicine Thailand Voluntary Health Card Scheme (HCP) 1994­95 1,005 HH National Statistics Office The five household surveys identified and accessible for analysis for the pur- poses of this report represent nonstandardized, relatively small scale data collec- tion efforts with a sample size of 346 to 1,200 households. The surveys were not nationally representative but were rather random samples of the local popula- tion. With the exception of Thailand's, the surveys are very recent. Table 5.3 summarizes the key characteristics of these surveys. The individual country sec- tions provide more detailed information about the survey instruments. Rwanda The Rwanda household survey was carried out by PHR in collaboration with the Rwandan National Population Office (ONAPO). Data collection took place dur- ing 40 days in October­November 2000 in three pilot districts. The household survey includes 2,518 households (11,583 individuals) successfully interviewed in the three pilot districts. The sample was based on the same sampling frame as the Rwandan Demographic and Health Survey (DHS) 2000, covering 11 health regions in Rwanda.5 Households for the prepayment household survey in the three districts were sampled at random from a list of primary sample households from sample cells identified in the national DHS sample, rendering the house- hold survey sample representative to the district level. The prepayment household survey used three structured questionnaires for data collection: a socioeconomic household questionnaire, a curative care ques- tionnaire, and a preventive care questionnaire. The household questionnaire col- Analysis of Community Financing Using Household Surveys 211 lected information on households' and individuals' sociodemographic and eco- nomic characteristics, including household expenditures for consumption goods, health and education, and participation in CBHI. The curative care questionnaire was addressed to household members who were sick two weeks prior to the inter- view, and the preventive care questionnaire was used to interview women of childbearing age who had delivered a child in the past five years or who were pregnant during the year preceding the interview (Diop and Schneider 2001). Senegal The Senegal household survey was carried out by the Institute for Health and Development (ISED) in Dakar in cooperation with the Center for Development Research in Bonn between March 2000 (pretest) and May 2000 (final survey). Households were randomly selected in four villages of Fandène, Sanghé, Ngaye Ngaye, and Mont Rolland in the Thiès region of Senegal where mutual health organizations are in place. A total of 346 households were surveyed, 70 percent of which are members of the mutual health organizations (MHOs). The data set consists of 2,987 persons, 60 percent of them members (some households did not insure their whole families). The participation rate in the interviews was very high, more than 95 percent. India (1) The first household survey for India was carried out by the National Council of Applied Economic Research (NCAER). A primary survey of 1,200 households was conducted in the Ahmedabad District of Gujarat, India, between December 1998 and December 1999. The survey included households from four types of health insurance enrollment status in rural and urban areas: · 360 households belonging to a contributory plan known as Employees' State Insurance Scheme (ESIS) · 120 households subscribing to a voluntary plan (Mediclaim) · 360 households belonging to a community- and self-financing scheme run by an NGO called Self-Employed Women's Association (SEWA) · 360 uninsured households purchasing health care services directly from the market (control group). The survey sample came from eight localities (about 90 households each) domi- nated by slum populations in the city of Ahmedabad and six villages (about 60 households each) in the neighborhood. India (2) The second survey of India also surveyed the SEWA population. This was a cross- sectional cohort study; respondents were interviewed at only one point in time, 212 Health Financing for Poor People: Resource Mobilization and Risk Sharing and we fixed in advance the number of SEWA and uninsured households (the two "cohorts"). Two-stage, random cluster sampling was used. The primary sampling units (PSUs) were villages. Twenty villages were selected randomly (using random- number tables); the probability of selection was equal for all villages regardless of size. The secondary sampling units were households. Within each village, insured were randomly selected from lists compiled by SEWA and uninsured were ran- domly selected from census or voting lists. In 10 villages, 14 SEWA households and 14 uninsured households were sampled, and in 10 villages, 14 SEWA households and 28 uninsured households were sampled (20 villages × 14 SEWA households = 280 SEWA households; 10 villages × 14 controls + 10 villages × 28 controls = 420 con- trols; therefore 700 households are included in this analysis).6 Thailand The Thailand survey was conducted in 1994­95 in the Khon Kaen Province of Thailand, which had experience with the voluntary health insurance scheme or the Health Card Program dating to 1983. The pilot study was implemented in six districts, where a sample of 1,005 households was selected and categorized into four groups: · Those who did not have a health card between 1993 and 1995 (495 households) · New health cardholders in 1995 (297 households) · Continued card purchasers (132 households) · Health card dropouts (81 households). Secondary data (National Statistics Office) statistics of card usage rates and uti- lization rates, reimbursement from providers, and the number of insured and uninsured in the province before and after the implementation of the program by the type of insurance scheme was also used in the study. Empirical Methods This section describes the general methods applied to the data sets. Since each of the five data sets is different, some variations in methodology exists and are reported in the individual reports. Determinants of Inclusion in Community Financing To assess the determinants of social inclusion in community-financing schemes, we assume that the choice of whether to enroll is influenced by two main determi- nants: individual and household characteristics, and community characteristics. Individual and household characteristics such as age, gender, income, and health status shape individuals' preferences toward health risks as well as their ability to pay membership fees and thus influence their demand for insurance. Analysis of Community Financing Using Household Surveys 213 We hypothesize that individual choice about whether to enroll in a prepay- ment scheme is moderated through certain social characteristics of the commu- nity. Social, ethnic, and religious values may shape peoples' preferences and attitudes toward health, risk, and solidarity. This may alter the outcome of the rational choice process for two individuals with similar individual and house- hold characteristics. They may decide differently about joining depending on encouragement from community leaders, availability of information, ease of maneuvering unknown processes, and the like. To estimate the weight of these determinants, a binary logit model was applied to four of the data sets, and a binary probit was applied to the Senegal data set. The model can be formally written as follows. (5.1) Prob (membership > 0) = X11 + X22 + The independent variable takes on a value of 1 if the individual is a member of a community-financing scheme and 0 if he or she is not. X1 represents a set of independent variables that are characteristics of the individual and the house- hold, such as income, gender, age, marker on chronic illness, or disability. X2 rep- resents a set of independent variables that approximate the social values in the communities: religion, marker on various communities where appropriate. Other variables specific to the surveys as well as interaction terms were included where appropriate. 1 and 2 are vectors of coefficient estimates, and is the error term. The two variables of primary interest are income and marker on different communities. Income. Income is the key variable in measuring the extent of social inclusion achieved by community-financing schemes. This assumes that income is a good approximation of social inclusion. Admittedly, this assumption ignores the fact that poverty and social exclusion have many dimensions and causes other than income such as ethnicity, religion, and mental and physical disabilities, to men- tion only a few. However, we hypothesize that effective demand for insurance is strongly determined by ability to pay. This hypothesis is supported by some of the literature on community financing, which suggests that the poorest and higher income groups are not included in pooling arrangements. Community characteristics. Community-specific dummy variables are our key variables in picking up unobservable characteristics of communities, such as social values and solidarity, to see if those variables influence individual choice to enroll in a community-financing scheme. Our hypothesis is that the impact of the com- munities is a significant determinant of the probability of enrolling in a scheme. Ideally, one would control for social values, social capital, and collective atti- tudes toward solidarity, risk, and health with more sensitive and direct variables. However, only the Senegal data set included a variable that directly measured the perceived level of solidarity among survey respondents. For the other sur- veys, community and district specific dummy variables were included. Admit- tedly, this is a crude measure to assess variation in social values, collective attitudes toward risk, health, and solidarity. Therefore a statistically significant 214 Health Financing for Poor People: Resource Mobilization and Risk Sharing finding regarding the community and district dummy variables will call for fur- ther examination of which community characteristics are really measured and picked up by these variables. In addition to income and community-specific dummies, other control vari- ables are also included: gender, age, disability or chronic illness, religion, and distance to the health center under the scheme. Some of these variables are important to control for the different probability of health care use (for example, age, health status, distance from provider). These variables also allow us to test the presence and importance of adverse selection to which all voluntary prepay- ment schemes are subject. Other variables are included to control for the different individual and house- hold attitudes toward investment in health at a time when illness is not neces- sarily present (for example, gender, religion). The literature has shown that distance gradient to the hospitals and local health centers and existence of out- reach programs influence the decision to purchase membership in the scheme. Determinants of Financial Protection Provided by Community Financing To empirically assess the impact of scheme membership on financial protection, a two-part model was used.7 The first part of the model analyzes the determinants of using health care services. The second part of the model analyzes the determi- nants of health care expenditures for those who reported any health care use. There are several reasons for taking this approach. First, using health expendi- ture alone as a predictor of financial protection does not allow us to capture the lack of financial protection for those who choose not to seek health care because they cannot afford it. As the first part of the model assesses the determinants of utilization, this approach allows us to see whether membership in community financing reduces barriers to accessing health care services. Second, the distribution of health expenditures is typically not a normal dis- tribution. There are a large number of nonspenders who do not use health care in the recall period. The distribution also has a long tail due to the small number of very high spenders. To address the first cause of non-normality, we restrict the analysis of health expenditures to those who report any health care use. As the first part of the model assesses determinants of use, we will still be able to look into whether scheme membership removes barriers to care. To address the sec- ond part of non-normality, a log-linear model specification is used. Part one of the model is a binary logit model for the India, Rwanda, and Thai- land data sets and a probit model in the Senegal model. The model estimates the probability of an individual's visiting a health care provider. Formally, part one of the model can be written as follows: (5.2) Prob (visit > 0) = X + Part two is a log-linear model that estimates the incurred level of out-of-pocket expenditures, conditioned on positive use of health care services. Formally, part two of the model can be written as follows: Analysis of Community Financing Using Household Surveys 215 Log (out-of-pocket expenditure visit > 0) = X + µ where X represents a set of individual and household characteristics that are hypothesized to affect individual patterns of utilization and expenditures. and are vectors of coefficient estimates of the respective models; and µ are error terms. The two variables of primary interest are scheme membership status and income. Scheme membership status. The key independent variable of interest is mem- bership status in community-financing schemes. The key question is that, con- trolling for a number of individual and household characteristics, do members of community-financing schemes have better access and a lower financial bur- den of seeking health care? Our hypothesis is that well-functioning prepayment and pooling schemes remove financial barriers to health care access demon- strated by increased utilization and reduced out-of-pocket spending of scheme members relative to nonmembers. Interaction terms between insurance mem- bership status and income are also explored. Income. Without the financial protection afforded by insurance, demand for health care is heavily determined by ability to pay, and for those who use health care, out-of-pocket expenditures are likely to mean a heavy financial burden. Through prepayment and pooling, we expect that financial barriers to care are reduced and income becomes a less significant predictor of health care utiliza- tion and out-of-pocket expenditures. Other control variables were also included in the estimation model to control for the differences in need for health care (for example, age, gender), differences in preferences toward seeking health care (for example, gender, religion), and differences in the cost (direct and indirect) of seeking health care (for example, distance). Limitations of the Methodology The applied methodology has several limitations. First, the estimated coeffi- cients might reflect bias due to adverse selection in the model trying to assess the inclusiveness of community-financing schemes. Adverse selection occurs because participation in the schemes is voluntary, and therefore those with greater than average health risk are more likely to enroll than those with lower than average risk. While multiple regression techniques can adjust for the observable characteristics that affect adverse selection, they cannot adjust for unobservable characteristics. This leads to biased coefficient estimates and thus undermines the internal validity of the results. We try to control for the adverse selection by including variables associated with health risk such as age, gender, and perceived health status. Second, bias may also be present in part two of the model because of potential endogenity between the choice of whether to enroll in health insurance and health care use. Individuals who self-select into the insurance program have 216 Health Financing for Poor People: Resource Mobilization and Risk Sharing unobservable characteristics--related to preference or health status (adverse selection)--that make them more likely than others to join the program and which also influence their decision to use health care services. An observed asso- ciation between health insurance affiliation and health care use and expenditure may therefore be due not to insurance but to the underlying unobservable char- acteristics. In the Senegal study, the Hausman test has been performed to control for the effects of the unobservable characteristics. Third, the variables we use to approximate social inclusion and community characteristics are indirect. Social inclusion-exclusion is measured here only in terms of income, and other determinants, such as ethnicity or religion, are not taken into account. Social values, social capital, collective attitudes toward risk, health, and solidarity are not measured through direct variables--only through indirect community or district­specific dummy variables that measure all unob- served characteristics that vary across the surveyed communities. Both of these weaknesses can be addressed in the future by finding more appropriate and more sociology-driven measures for social exclusion. Fourth, we do not have a direct measure of financial protection. Ideally, we would like to measure the impact of community financing on impoverishment directly. Such data were not, however, available within the time frame of this study. Our measures, utilization plus out-of-pocket spending, are indirect approximations of financial protection. RESULTS This section presents the key findings from the individual country analyses. Not all variables and findings that are important for the specific countries are included here. This section aims to present the findings with regard to the key variables of interest described under the methods section. The individual coun- try chapters provide more comprehensive descriptions of regression results as well as discussion of those findings. Determinants of Social Inclusion in Community Financing The results in terms of the determinants of social inclusion through community financing are varied. Table 5.4 presents the determinants that were found statis- tically significant in the five household surveys. Income and Other Socioeconomic Determinants We had hypothesized that household income is a significant determinant of mem- bership in a prepayment scheme as ability to pay influences demand for prepayment. The results of the five studies neither confirmed nor disproved this hypothesis. Household income was a significant determinant of membership in a prepayment scheme in Senegal and Thailand but not significant in India and Rwanda. Analysis of Community Financing Using Household Surveys 217 TABLE 5.4 Statistically Significant Determinants of Inclusion in Community Financing Rwanda Senegal India (1) India (2) Thailand Model Logit Probit Logit Logit Logit Dependent variable Dependent Proportion of Proportion of Proportion of Proportion of Proportion of variable population population population population population that enrolled in 54 enrolled in 1 enrolled in enrolled in purchased new health schemes in of 4 schemes SEWA insurance SEWA insurance card, continued, 3 districts dropped out, never purchased Independent variables: individual and household characteristics Income/assets No Yes No For the poorest No fifth Age No No Yes Yes No Education Yes No No No Yes Gender No No -- -- No Health status No No Yes Yes Yes Household size Yes -- Yes No -- Marital status -- --- Yes No No Religion -- Yes -- No -- Ethnic group -- Yes -- -- -- Membership in other organization -- Yes -- -- -- Distance of household from scheme provider Yes -- -- -- -- Independent variables: community characteristics Community marker for unobservable characteristics Yes Yes -- -- -- Solidarity N/A No N/A N/A N/A Note: Yes: variable is significant at least at the 10 percent level. No: variable is not significant. (--) : not included in the particular model. In the Senegal data set, the income variable was significant. Income was included in the model in three forms: as a continuous variable (significant at the 1 percent level) divided into terziles (significant at the 10 percent level for the lower terzile and at the 5 percent level for the upper terzile), and as a self- reported measure of being poor or nonpoor (poor significant at 1 percent). In all cases, income had a positive impact on the probability of being a scheme mem- ber. This indicates that ability to pay does make a difference in the decision to 218 Health Financing for Poor People: Resource Mobilization and Risk Sharing join: lower income households and the self-reported poor were less likely to join a scheme than higher income households. In Rwanda, the income variable, divided into quartiles, was not a significant determinant of scheme membership. Of the two included asset variables, cattle and radio ownership, only radio ownership was significant and had a positive impact on the probability of membership. As further explored in the discussion section, the strong impact of radio ownership is more likely to measure the suc- cess of the information campaign than the impact of assets on the decision. These results are robust in alternative models: in particular, the results do not change as variables colinear with income were excluded. Both household surveys in India had similar findings. SEWA membership was not strongly influenced by income or by household assets. In the first survey, household income was included as quintiles. Only being in the highest income quintile had any impact on membership status. Being in the top quintile increased the probability of membership in SEWA by 1.87 percentage points as compared with the lowest quintile. There is some indication, however, that the income variable is measured with error, and thus it does not pick up the true welfare characteristics of households. For example, large household size (six or more members) is a significant predictor of the probability of being a member. Households with more than six members are 2 to 7 percentage points less likely to join than households with fewer than four members. Similarly, activity status is also a significant predictor of the probability of membership. Members of any trade are 2 to 4 percentage points more likely to be members of SEWA than non- workers. The insignificance of the income variable remains when activity status is excluded from the model for colinearity reasons. To the extent that household size and employment status is a proxy for income and welfare, there is reason- able doubt as to whether the income variable is measured with error. In the second survey, an asset index was developed to measure household wealth. The asset index was not found to significantly influence the probability of being a SEWA member. Other Individual and Household Characteristics Health status was included in the analysis of all five surveys. The hypothesis explored was that adverse selection was present: people with worse health status are more likely to join the prepayment scheme as their expected value from insur- ance is higher than those with better health status. The hypothesis was disproved in Senegal and Rwanda. In the other three studies, the hypothesis was confirmed. In the Rwanda survey, households with a pregnancy over the past year were marked as well as households with children below the age of 5. Neither variable was significant. In Senegal, two variables were introduced to capture health status: illness ratio, measuring reported illness in the previous six months, and hospitalization ratio, measuring the frequency of hospitalization in the previous two years. Both variables were found to have no influence on participation in a mutual scheme. Analysis of Community Financing Using Household Surveys 219 In the India (1) survey, three variables aimed to control for health status: whether the respondent had a chronic ailment, whether the respondent had been hospitalized over the recall period, and whether the respondent had given birth during the previous two years. The significance of the variables varied in the alternative three models. Childbirth was significant in two of the models, and reported hospitalization was significant in one. When significant, both childbirth and reported hospitalization increased the likelihood of SEWA mem- bership threefold. In the India (2) survey, the number of acute illness episodes reported over the previous 30 days was included to control for general level of health. The variable was significant at the 5 percent level. Those who reported illness over the previous 30 days were 68 percent more likely to be part of SEWA than those who did not. In Thailand, there were several control variables for health status: presence of illness in the household in the previous three months, the number of sick mem- bers with chronic illness in the family, and economic problems during sickness of family members. The presence of illness was a significant determinant of pur- chasing a health card while the other two health status variables were not (co- linearity). Those reporting illness in the previous three months are 57 percentage points more likely to have ever purchased a health card compared with those who reported no illness over the previous three months. Community Characteristics To control for community characteristics and test the hypothesis of whether community characteristics modified individual decisionmaking, two models included a dummy variable to control for all the unobservable differences at the community level that may influence individual decisionmaking. In both cases, the dummy variables were significant predictors of the probability of enrolling in the prepayment scheme. In Rwanda, three communities were surveyed, and the community dummies were found to be highly significant at the 1 percent level. Households from Kab- gayi were 3.5 times as likely to purchase a prepayment plan and households from Byumba were 15.8 times as likely to purchase a prepayment plan as house- holds from Kabutare. In Senegal, different model variations show that the inhabitants of the vil- lages of Sanghé and Mont Rolland have significantly lower probability of mem- bership than people from Ngaye Ngaye and Fandène. These results indicate that the different types of health insurance provided--primary health care in Ngaye Ngaye and inpatient care in the other three mutuals--had no significant influ- ence on the decision to participate. Instead, specific village factors such as the management of the mutual seem to play a role. The mutual of Sanghé has faced several financing and managerial difficulties that have led to a suspension of operations for some time. As a consequence, several people left the mutual. Efforts to reestablish the mutual have been successful; the mutual is functioning again today, but with a lower participation rate. 220 Health Financing for Poor People: Resource Mobilization and Risk Sharing Determinants of Financial Protection in Community Financing The results in terms of the determinants of financial protection through com- munity financing are varied. Table 5.5 presents the determinants that were found statistically significant in four of the household surveys. The household survey conducted in Thailand does not permit analysis of the determinants of out-of-pocket payments and was therefore excluded from the analysis. Insurance Effect In three of the four household surveys, scheme membership was a significant determinant of the probability of using health care and the level of out-of- pocket payments. This confirms our original hypothesis that even small-scale prepayment and risk pooling reduce financial barriers to health care. In Rwanda, scheme members are six times more likely to enter the modern health care system when sick than nonmembers. Scheme members who report any visit to a professional provider have lower out-of-pocket payments per ill- ness episode than nonmembers. In Senegal, scheme members are 2 percentage points more likely to use hospi- tal care than nonmembers (marginal effect). Their out-of-pocket payment for hospital care decreased by 50 percent in comparison with nonmembers, with all other factors constant. In India, the picture is mixed. Model one reports significant impact of SEWA membership on the probability of using health care but finds no impact on the total annual cost of health care utilization. In contrast, model two finds that SEWA membership has no impact on the likelihood of being admitted for hospi- tal care but finds that membership reduces the total annual out-of-pocket pay- ments for hospitalization. Socioeconomic Determinants Our second key hypothesis was that insurance coverage makes income a less sig- nificant determinant of health care utilization and out-of-pocket payments. This hypothesis is neither confirmed nor disproved by our findings, which are quite varied. In Rwanda, income continues to be a determinant of the likelihood of using health care as well of the average out-of-pocket payment for the poorest quintile, with all other variables constant. In Senegal, income is a determinant of using hospital care only for the richest third of the sample. Income is a significant determinant of the level of out-of- pocket spending on hospital care. In India, model 1 reports that income is not a significant determinant of use and is a significant determinant of out-of-pocket payments only for the richest quintile. Model 2 confirms the finding that income is not a significant determi- nant of use and thereby confirms our original hypothesis. It also finds that income is a significant determinant of out-of-pocket expenditures for the richest quintile. for care out-of- conditional 0) OOPs > payment hospital annual (use of s s s s s (2) Log-linear on otalT pocket use Ye Ye No Ye -- -- Ye No Ye -- India sample quintile of one group Patterns Utilization least at richest oldest nditure Logit Proportion w/ hospitalization No For For No -- -- No No No -- payments. Expe of size direct cost use conditional 0) OOPs > care annual household out-of-pocket indirect (use s s (1) Log-linear on otalT and health No Ye -- -- -- Ye Small -- -- -- OOPs: Out-of-Pocket India table. and this sample health in of any severe them Utilization use very Utilization s of Logit Proportion reporting care Ye No No No No For No No -- -- include not did we, conditional 0) on group OOPs > Determinants 26 (use < chapter s s Log-linear on Out-of-pocket spending hospitalization Ye Ye For No No No -- -- No -- this in percent) Senegal 10 at sample of one terzile discussed not least Utilization least at richest are s s s (at Logit Proportion w/ hospitalization Ye For Ye No Ye No -- -- No -- they as but full episode the quartile Significant conditional 0) OOPs per for studies > illness-related the (use poorest illness s characteristics s of Log-linear on otalT out-of-pocket payment of episode Ye For No No Ye No No -- -- No some in Statistically Rwanda visit health household sample of one and included effect Findings: Utilization least provider were at professional s s s s s Logit Proportion w/ to care Ye individual Ye Ye No No Ye No -- -- Ye insurance variables Summary variable: variables: variable control 5.5 variable size from membership of status/severity status provider Other ABLET illness Model Dependent Dependent Independent Scheme Independent Income/assets Age Education Gender Health of Household Marital Religion Distance household scheme Note: 221 222 Health Financing for Poor People: Resource Mobilization and Risk Sharing Other Determinants Where included, distance from the scheme provider was a significant determi- nant of the likelihood of using health care. In Rwanda, people who live close to the health facility are significantly more likely to seek care (61 percent) than those who live farther away. Patients in the lowest income quartile are far less likely to seek care than those in the highest income quartile. This means that while the prepayment scheme has significantly increased access to health care for members, including those who are poor, the impact at the district level in increased access to health care for the poor remains an issue. The solution is to find mechanisms to increase enrollment of poor households in the prepayment schemes. DISCUSSION Determinants of Social Inclusion in Community Financing Socioeconomic Determinants We had hypothesized that household income is a significant determinant of membership in a prepayment scheme. Ability to pay would influence the demand for prepayment. Review of the literature suggested that this was the case for many schemes and that, for the very poorest, financial barriers to care remained even with the introduction of community financing. The findings of this research suggest that financial barriers can be overcome by community-financing schemes even in a very poor context (India, Rwanda). In the case of two schemes, income was found not to be a significant determi- nant of membership status, suggesting that the poor were just as likely to be included in the schemes as the better-off community members. This finding is no doubt due to the fact that, because all clients of community- financing schemes are poor, there is no large variation in the income variable. But this is true in all four countries and does not explain why the schemes in India and Rwanda have achieved inclusion of the poor while those in Senegal and Thailand have not. In our interpretation, assuming that this finding is not due to methodological error, it indicates that certain design and implementation features allow poor communities to overcome the inability of their poorest residents to pay. In other words, how schemes are designed and implemented makes a difference in terms of their success in targeting the very poor. Further analysis is required to com- pare the structural, managerial, organizational, and institutional characteristics of the surveyed schemes to determine precisely which features contribute to bet- ter targeting outcomes. Three design and implementation features of the Rwanda scheme stand out as potential explanatory factors that have allowed for the inclusion of the poorest: Analysis of Community Financing Using Household Surveys 223 participatory process, subsidies and facilitation of contribution payments, and information campaigns. First, participatory design and democratic management of the scheme led to a sense of ownership and trust toward the health scheme. This has shaped the preferences and attitudes of households toward investment in their health care. Participation was achieved through community-level meetings and allowing community members to vote. This finding is consistent with the social capital literature that suggests that voice leads to empowerment, which in turn con- tributes to better sustainability. Second, the poorest's ability to pay was given special consideration. Those who could not afford to pay a one-time enrollment fee were allowed to pay in installments. Furthermore, the community and the churches collected money to contribute for the enrollment fee of the indigent, the disabled, orphans, and other disadvantaged people. This finding is consistent with the literature that suggests that poorly designed contribution schemes often stand in the way of expanding enrollment. Flexibility in scheme design makes a difference, for example, allowing cash contributions or timing collection time to coincide with cash-endowed periods. Third, information campaigns were conducted through 30 workshops in the three pilot districts. The information campaign in the districts of Byumba and Kabgayi was more intensive than in the control region. This may explain why the coefficient of the two variables marking these districts was a strong determi- nant of the likelihood of participation. Households in these two regions were 3.5 and 15.8 percentage points more likely to be members of the prepayment scheme than the inhabitants of the control Kabutare region. An interesting question is whether the Rwanda scheme can maintain this high level of inclusion as the scheme ages and matures. A significant difference between the Senegal and Rwanda schemes is that the surveyed schemes in Sene- gal have been in existence for 10 years. In contrast, the Rwanda scheme is recent. As the years go by, will the Rwanda scheme become subject to the often-reported issues of adverse selection? In the case of SEWA in India, successful targeting of the poor can potentially be attributed to the linkages that the prepayment scheme has to other social pro- tection mechanisms SEWA has in place. The fact that SEWA's Social Security Scheme is nested within a larger development organization has undoubtedly been an important factor in ensuring inclusion of the poor. Other factors that have facilitated inclusion of the poor include: affordable premiums, village-level representatives who are themselves poor, self-employed women, and efforts to serve geographically isolated villages. An interesting question requiring further exploration is: To what extent is bet- ter social inclusion due to explicit subsidies for the poor (through churches, gov- ernment, or donors) versus participatory social structures? In other words: To what extent can income deprivation be overcome through giving voice to the poor? Participatory structures have their weaknesses as well. Because the rich 224 Health Financing for Poor People: Resource Mobilization and Risk Sharing always have the financial incentive to opt out of income-pooling arrangements if they can, achieving a high level of participation may be costly, especially for the poor. In sum, further assessment is required to identify the factors that contribute to better social inclusion and which of them can be influenced at the household, community, and government levels. Community Determinants We found significant results for community variables. Individuals living in dif- ferent villages or districts have a different likelihood of joining a prepayment scheme, holding all other facts constant. Our original intention was to attribute some of the variation in of the community dummies to the variation in social values and collective attitudes toward risk, health, and solidarity. Upon further reflection, however, many other factors differ across these communities that other variables do not control for. Thus the community variables may measure some other aspects of the design and implementation of community financing. For example, the strength of the advertising campaign as in the case of Rwanda. In the Rwanda case, there was a considerable difference in terms of the advertisement campaign and the involve- ment of the local leaders in the three districts. It is likely that this is what is picked up by the variables. In the case of Senegal, people in Fandène and Ngaye-Ngaye tend to enroll pro- portionately more than people living in the villages of Mont Rolland and Sanghé. For Fandène this is not surprising: its mutual is the oldest, it functions quite well, and it is closest to the hospital. For Ngaye-Ngaye, the interpretation is more difficult, as people have stated that they were not satisfied with the mutual's functioning. The scheme in Ngaye-Ngaye covers primary care and not hospitalization. Hence this result could also be interpreted as showing that there is also a demand for ensuring high-frequency, low-cost risks. This suggests that community variables as they are crudely constructed pick up variations in the design and implementation of community-financing schemes that directly influence the value people get from being enrolled. This suggests that while community appears to be a significant determinant of enroll- ment, better measures are needed to assess which community characteristics encourage social inclusion and which characteristics tend to be more exclusive. Variables that better capture values, attitudes toward health, social cohesion, and solidarity would enable delineation of the community characteristics that create a fostering environment for community financing and those that do not. From a policy perspective, this would contribute to our understanding about which characteristics can be influenced and which ones cannot and therefore need to be taken into consideration as a constraint when designing community- financing schemes. For example, attitudes toward health can be shifted through health education campaigns and information while social cohesion, or the lack thereof, is hard to foster if it is not present at the onset. Analysis of Community Financing Using Household Surveys 225 Determinants of Financial Protection in Community Financing The results confirm our initial hypothesis that community financing through prepayment and risk sharing reduces financial barriers to health care, as demon- strated by higher utilization but lower out-of-pocket expenditure of scheme members. At the same time, income still influences use and expenditure, although its influence is more pronounced for higher income groups than for lower income groups. This suggests that community financing has been particularly successful in reducing the financial barrier to access for the lower income groups in the sur- veyed population. These findings confirm that risk pooling and prepayment, no matter how small scale, improve financial protection for the populations they serve. The pol- icy implication of this finding is that it is critical to move away from resource- mobilization instruments that are based on point-of-service payments. If prepayment and risk sharing can be encouraged, it is likely to have an immedi- ate poverty impact--directly, by preventing impoverishment due to catastrophic health expenditures, or indirectly, by ensuring access to health care, thereby improving health and allowing the individual to take advantage of economic and social opportunities. In this, the critical question becomes: What form of community financing is better able to provide financial protection for its members? Those that include hospital care? Those that include primary care? Those that are based on some common professional characteristics? Those that have strong government sup- port? Provider-based ones? Assessing what scheme characteristics will encourage financial protection is the next important research step. CONCLUDING REMARKS This study aimed to present initial findings from five household surveys regard- ing the social inclusion and financial protection impact of community-financing schemes. While the findings are preliminary, a number of common performance patterns have emerged. First, successful inclusion of the poorest is not an automatic outcome of com- munity structures. Community involvement can be exclusionary as well as inclusionary. This suggests that certain community characteristics as well as scheme design and implementation features are important determinants to achieve pro-poor targeting outcomes. These determinants and the direct causal- ity are not well explored with regard to health financing, and further investiga- tion is warranted. In particular, the role of external financial support (such as government subsi- dies, donor funding, reinsurance) in encouraging social inclusion needs further exploration. So, too, does the role of participation, by providing the poor with a voice. 226 Health Financing for Poor People: Resource Mobilization and Risk Sharing Second, community financing reduces financial barriers to health care, as demonstrated by higher utilization but lower out-of-pocket expenditure of scheme members relative to nonmembers. This suggests that prepayment and risk sharing--even on a small scale--lower financial barriers to health care. APPENDIX 5A. LIST OF REVIEWED SURVEY INSTRUMENTS An extensive search of available research instruments was undertaken during January­March 2001. The objective of the search was to identify household sur- veys that would allow us to test the impact of community-based health financ- ing on social inclusion and household level financial protection. Twenty-one countries were identified as likely candidates (those where we had prior knowledge of some kind of community-financing initiatives). In these countries, four survey instruments were reviewed: Living Standards Measurement Surveys (LSMS) Demographic and Health Surveys (DHS) Household budget surveys Other nonstandard surveys The following set of variables was defined as minimum criteria for the survey to be useful for our project. Key independent variable. Identifier for the type of prepayment scheme that covers the household. Of these, we looked for those in which we could separate households covered by community-based health financing from those either not covered or covered by formal (general tax, social insurance) mechanisms. Control (independent) variables. Socioeconomic status, religion, age, gender, income level, and chronic illness or disability. Outcome (dependent) variables of interest. Health outcomes, financial protection (that is, some sense of out-of-pocket spending and household income level), and consumer satisfaction. Having initially reviewed four LSMS, nine DHS surveys, and six household bud- get surveys, we concluded that they did not allow us to identify households with access to community-based health financing. Even when the survey included health-financing questions, coverage through community financing could not be separated from other health-financing instruments such as private insurance or social insurance. As a result, not all available LSMS and DHS survey question- naires were reviewed for the selected countries, and instead we focused on iden- tifying small-scale nonstandardized surveys. analysis analysis analysis analysis analysis further further further further further Data progress for for for for for of in still is Status accessible accessible accessible accessible accessible contact not not not not not not collection set set set set set Could Data Data Data Data Data Data Development for Bonn data of Thailand, Research, Administration Surveys Holder Economics University Other Belgian Health Development Jütting India HSPH for for CREDESA CREDESA Gumber Desmet, Supakankunti NCAER, Anil M. Harvard Center Chulalonkron S. Center Johannes PHR PHR PHR PHR UNICEF UNICEF PHR seY seY No seY No seY seY seY seY seY No No seY seY seY Useful Search No No No No No No Survey Useful Budget HH Search No No No No No No No No No Development. Countries Useful and 21 DHS for Health Search for Centre Surveys No No No No No Useful Regional LSMS = Reviewed Search CREDESA 5.6 America UNICEF ABLET Country Asia India Bangladesh Philippines China Cambodia Thailand Nepal Africa Senegal Niger Rwanda Ghana anzaniaT Mali Uganda Zambia Nigeria Benin Guinea Latin Bolivia Ecuador Guatemala Note: 227 228 Health Financing for Poor People: Resource Mobilization and Risk Sharing Acknowledgments: The authors are grateful to the World Health Organization (WHO) for having provided an opportunity to contribute to the work of the Commission on Macroeco- nomics and Health and to the World Bank for having published the material in this chapter as an HNP Discussion Paper. Valuable guidance on methodological issues was provided by Adam Wagstaff. NOTES 1. July 1999: RWF 2,500 = US$7.50. 2. Premium rates were set by taking into account existing user fees and by assuming that uti- lization rates would increase by 25 percent over baseline levels. See TR 45: chapter 3.1.1. 3. The Kabgayi PPS covers full stays at the hospital for caesarian sections (C-sections), malaria, and nonsurgical pediatrics, whereas the Kabutare PPS and Byumba PPS cover C-sections, physician consultation, and an overnight stay at the district hospitals. 4. July 1999: RWF 100 = US$0.3. 5. The DHS was conducted by the ONAPO in collaboration with Macro International and USAID in 2000­01. Households for the DHS were selected as primary sample units from sample cells identified for the Living Conditions Monitoring Survey (LCMS), con- ducted by the Ministry of Finance in collaboration with UNDP in 2000­01. 6. Note this study was actually designed to look at two community-based insurance schemes, SEWA and the Tribhuvandas Foundation, and a total of 1,120 households were actually interviewed. However, TF households and their controls are not included in this analysis. 7. This model is similar to the two-part demand model developed as part of the Rand Health Insurance experiment to estimate demand for health care services (see Duan and others 1982; Manning and others 1987). For a recent application of the model that ana- lyzes the access impact of school health insurance in Egypt, see Yip and Berman (2001). REFERENCES Arhin, D. 1994. "The Health Card Insurance Scheme in Burundi: A Social Asset or a Non- Viable Venture?" Social Science and Medicine 39(6): 861­70. Atim, C. 1998. Contribution of Mutual Health Organizations to Financing, Delivery, and Access to Health Care: Synthesis of Research in Nine West and Central African Countries. Technical Report 18. Partnerships for Health Reform Project, Abt Associates Inc., Bethesda, Md. Atim, C., and M. Sock. 2000. An External Evaluation of the Nkoranza Community Financing Health Insurance Scheme, Ghana. Technical Report 50. Partnerships for Health Reform Project, Abt Associates Inc., Bethesda, Md. Carrin, G., R. Aviva, H. Yang, H. Wang, T. Zhang, L. Zhang, S. Zhang, Y. Ye, J. Chen, Q. Jiang, Z. Zhang, J. Yu, and L. Xi. 1999. "The Reform of the Rural Cooperative Med- ical System in the People's Republic of China: Interim Experience in 14 Pilot Counties." Social Science and Medicine 48: 961­72. Analysis of Community Financing Using Household Surveys 229 CLAISS (Latin American Center for Health Research). 1999. "Synthesis of Micro-Insurance and Other Forms of Extending Social Protection in Health in Latin America and the Caribbean." Under the supervision and guidance of the ILO and PAHO counterparts, for the ILO-PAHO initiative of extending social protection in health in Latin America. Presented to the Mexico City tripartite meeting of the International Labour Organiza- tion (ILO) with the collaboration of the Pan American Health Organzation (PAHO), Mexico City. Criel, B., P. Van der Stuyft, and W. Van Lerberghe. 1999. "The Bwamanda Hospital Insur- ance Scheme: Effective for Whom? A Study of Its Impact on Hospitalization Utilization Patterns." Social Science and Medicine 48: 897­911 Desmet, A., A. Q. Chowdhury, and K. Islam. 1999. "The Potential for Social Mobilization in Bangladesh: The Organization and Functioning of Two Health Insurance Schemes." Social Science and Medicine 48: 925­38. DeRoeck, D., J. Knowles, T. Wittenberg, L. Raney, and P. Cordova. 1996. Rural Health Ser- vices at Seguridad Social Campesino Facilities: Analyses of Facility and Household Surveys. Technical Report 13. Health Financing and Sustainability Project, Abt Associates Inc., Bethesda, Md. Diop, F., and P. Schneider. 2001. Household Survey in 5 Rwandan Districts. Technical Report. Partnerships for Health Reform Project, Abt Associates Inc., Bethesda, Md. Diop, F., A. Yazbeck, and R. Bitran. 1995. "The Impact of Alternative Cost Recovery Schemes on Access and Equity in Niger." Health Policy and Planning 10(3): 223­40 Dror, D., and C. Jacquier. 1999. "Micro-Insurance: Extending Health Insurance to the Excluded." International Social Security Review 52(1): 71­97. Duan, N., W. G. Manning Jr., C. M. Morris, and J. P. Newhouse. 1982. A Comparison of Alternative Models for the Demand for Medical Care. Report R-2754-HHS. The Rand Cor- poration, Santa Monica, Calif. Gilson, L., D. Kalyalya, F. Kuchler, S. Lake, H. Oranga, and M. Ouendo. 2000. "The Equity Impacts of Community Financing Activities in Three African Countries." International Journal of Health Planning and Management 15: 291­317. Gumber, A., and V. Kulkarni. 2000. "Health Insurance for the Informal Sector: Case Study of Gujarat." Economic and Political Weekly, September 30, 3607­13. Hsiao, W. C. 2001. "Unmet Health Needs of Two Billion: Is Community Financing a Solu- tion?" Draft paper for the WHO Commission on Macroeconomics and Health. World Bank, Washington, D.C. Jakab, M., and C. Krishnan. 2001. "Community Involvement in Health Care Financing: Impact, Strengths and Weaknesses: A Survey of the Literature" Draft Paper for the WHO Commission on Macroeconomics and Health. World Bank, Washington, D.C. Jütting, J. 2002. The Impact of Community-Based Health Financing on Financial Protection: Case Study Senegal. HNP Discussion Paper, World Bank. Washington, D.C. Liu, Y., S. Hu, W. Fu, and W. C. Hsiao. 1996. "Is Community Financing Necessary and Fea- sible for Rural China?" Health Policy 31: 155­71. Manning, W. G., J. Newhouse, N. Duan, E. Keeler, A. Leibowitz, and M. Marquis. 1987. "Health Insurance and the Demand for Medical Care: Evidence from a Randomized Experiment." American Economic Review 77(3): 251­77. 230 Health Financing for Poor People: Resource Mobilization and Risk Sharing McPake, B., K. Hanson, and A. Mills. 1993. "Community Financing of Health Care in Africa: An Evaluation of the Bamako Initiative." Social Science and Medicine 3(11): 1383­95. Narayan, D. 1999. "Bonds and Bridges: Social Capital and Poverty." Policy Research Work- ing Paper 2167. World Bank, Poverty Reduction and Economic Management Network, Washington, D.C. Preker, A., and M. Jakab. 2001. "The Role of Communities in Providing Financial Protec- tion against Illness." In D. Dror and A. Preker, eds., Social Reinsurance: Improved Risk Sharing for Low-Income and Excluded Populations. Forthcoming. Schneider, P., A. K. Nandakumar, D. Porignon, M. Bhawalkar, D. Butera, and C. Barnett. 2000. Rwandan National Health Accounts 1998. Technical Report 53. Partnerships for Health Reform Project, Abt Associates Inc., Bethesda, Md. SEWA brochures. 2000. Supakankunti, S. 1998. Comparative Analysis of Various Community Cost Sharing Imple- mented in Myanmar. Paper presented to the Workshop of Community Cost Sharing in Myanmar, November 26­28. Toonen, J. 1995. Community Financing for Health Care: A Case Study from Bolivia. Bulletin 337. Amsterdam: Royal Tropical Institute. Wiesmann, D., and J. Jütting. 2001. "Determinants of Viable Health Insurance Schemes in Rural Sub-Sahara Africa." Quarterly Journal of International Agriculture 50(4): 361­78. World Bank. 2000. World Development Report 2000: Attacking Poverty. Washington, D.C. WHO (World Health Organization). 2000. World Health Report 2000: Health Systems-- Improving Performance. Geneva. Yip, W., and P. Berman. 2001. "Targeted Health Insurance in a Low-Income Country and Its Impact on Access and Equity in Access: Egypt's School Health Insurance." Health Economics 10(2): 207­20. Ziemek, S., and J. Jütting. 2000. Mutual Insurance Schemes and Social Protection. Bonn: Cen- ter for Development Research (ZEF). CHAPTER 6 Financial Protection and Access to Health Care in Rural Areas of Senegal Johannes Paul Jütting Abstract: Community-based health insurance schemes are becoming increasingly rec- ognized as an instrument to finance health care in developing countries. Taking the example of les mutuelles de santés (mutual health organization) in rural Senegal, this chap- ter analyzes whether members in a mutual health insurance scheme have better access to health care than nonmembers. A binary probit model is estimated for the determinants of participation in a mutual and a logit/log linear model is used to measure the impact on health care utilization and financial protection. The results show that, while the health insurance schemes reach otherwise excluded people, the very poorest in the communi- ties are not covered. Regarding the impact on the access to health care, members have a higher probability of using hospitalization services than nonmembers and pay substan- tially less when they need care. Given the results of this study, community-financing schemes have the potential to improve the risk-management capacity of rural house- holds. To reduce identified limitations of the schemes, an enlargement of the risk pool and a scaling up or linking of the schemes is, however, a prerequisite. Appropriate instru- ments to be further tested should include reinsurance policies, subsidies for the poorest, and developing linkages to the private sector via the promotion of group insurance poli- cies. All these instruments call for a stronger role for public health policy. H ealth insurance schemes are an increasingly recognized tool for financing health care provision in low-income countries (WHO 2000). Given the high latent demand for good quality health care services and the extreme under- utilization of health services in several countries, it has been argued that social health insurance may improve access to acceptable quality health care. Many authors have criticized other forms of health care financing and cost-recovery strategies, such as user fees (see, for example, Gilson 1998); insurance seems to be a promising alternative as it offers the opportunity to pool risks, thereby transfer- ring unforeseeable health care costs to fixed premiums (Griffin 1992). However, there is some evidence that neither purely statutory social health insurance nor commercial insurance schemes alone can significantly contribute to an increase in coverage rates and thereby broaden access to health care. Especially in rural and remote areas, unit-transaction cost of contracts are too high, which often leads to a state and market failure (Jütting 2000). Recently, mainly in Sub-Saharan Africa but also in a variety of other countries, nonprofit mutual, community-based health insurance schemes1 have emerged (Bennett, Creese, and Monasch 1998; Wies- mann and Jütting 2001; Jakab and Krishnan 2001).2 These schemes are character- ized by an ethic of mutual aid, solidarity, and collective pooling of health risks 232 Health Financing for Poor People: Resource Mobilization and Risk Sharing (Atim 1998). In several countries, these schemes operate in conjunction with health care providers, mainly hospitals in the area. Proponents argue that these schemes have the potential to increase access to health care (see, for example, Dror and Jacquier 1999). The authors of the few stud- ies so far available, however, are less optimistic (for example, Bennett, Creese, and Monasch 1998; Criel 1998; Atim 1998). They argue that often the risk pool is too small, adverse selection problems arise, the schemes are heavily dependent on sub- sidies, financial and managerial difficulties arise, and overall sustainability does not seem to be ensured. However, while these studies are important contributions to our knowledge about the schemes' general strengths and weaknesses, not enough attention has been given to the context in which these schemes have been intro- duced and the objectives of the schemes themselves. Furthermore, the potential social benefit of the schemes, that is, their impact on health care access, labor pro- ductivity, and households' risk-management capacity, has been largely ignored. Against this background, this chapter analyzes whether mutual health insur- ance schemes improve access to health care in rural Senegal. We tackle two prin- cipal questions: What are the important socioeconomic determinants that explain membership in a voluntary health insurance scheme? (We thereby iden- tify important factors influencing the demand for health insurance.) What is the schemes' impact on the utilization of health care by, and the level of financial protection for, members as opposed to nonmembers? To answer these questions, we use a binary probit model for estimating marginal coefficients for the determinants of participation and a logit/log-linear model to analyze the impact on health care utilization by, and financial protection for, mem- bers as opposed to nonmembers. By applying this methodology, we go beyond most of the available studies on the impact of community-financing schemes, which have either relied on secondary literature (for example, Bennett, Creese, and Monasch 1998) or restricted their data analysis to qualitative interpretations (for example, Atim 1998). We have chosen the case of Senegal because we find here a relatively long (10 years) experience with mutual health insurance schemes and an innovative institutional setting (Tine 2000). There is a contract between a nonprofit health care provider, a Catholic-run hospital, and the mutuals, which allows the latter to receive health care at a lower rate. The outline of the chapter is as follows. First the chapter gives a quick overview of health insurance schemes in rural Sub-Saharan Africa and presents the specific situ- ation in Senegal. Next it describes the research design and the methodology used. The results of the estimations are then discussed and conclusions presented. HEALTH INSURANCE IN RURAL SUB-SAHARAN AFRICA Wiesmann and Jütting (2001) present a detailed overview of health insurance schemes outside formal sector employment in Sub-Saharan Africa, which is Financial Protection and Access to Health Care in Rural Areas of Senegal 233 FIGURE 6.1 Urban and Rural Health Insurance Schemes in Sub-Saharan Africa, Year of Inception and Size Number of Beneficiaries Urban Rural < 1,000 1,000 - 10,000 10,001 - 100,000 > 100,000 Unknown Year of Inception before 1979 1980 - 1989 1990 - 1998 Note: For some schemes, the location is only approximately indicated due to lack of exact data or space problems. Source: Wiesmann and Jütting (2001). based on extensive research done in the past few years (Bennett, Creese, and Monasch 1998; Atim 1998; Musau 1999). Most of the schemes were set up in the 1990s. The reasons that promote and foster the development of mutual health insurance schemes have not been analyzed in depth so far, but some trends are obvious (Wiesmann and Jütting 2001). First, people have been forced to think about alternative solutions as health care is no longer offered for free at the pub- lic facilities and the introduction of user fees has had negative effects, especially for the poor. Second, in the context of decentralization, more power has been delegated to the communities, which allows them to also assume more responsi- bilities in the provision of local public goods. Third, the quite positive experi- ence with credit and financing institutions is leading to a discussion about whether the mutuals should enlarge their portfolios to include insurance prod- ucts. Finally, the debate in the literature over the cost of illness has shown that health shocks often force households into high-cost risk-coping strategies. Access to insurance could reduce these costs substantially (Weinberger and Jüt- ting 2000; Asfaw and others 2001). The map in figure 6.1 gives a view of health insurance schemes outside the for- mal employment sector in Sub-Saharan Africa (the following section draws on 234 Health Financing for Poor People: Resource Mobilization and Risk Sharing Wiesmann and Jütting 2001). The map clearly shows that, thus far, community- based health insurance is more common in West Africa than in Central or East Africa. In some countries, these new schemes are mainly an urban phenomenon-- such as those in Côte d'Ivoire and Tanzania--whereas in such states as Uganda, Ghana, and Benin, they predominantly cover people in rural areas. Some of the schemes are confined to a local cooperative of craftsmen or traders, and therefore the schemes are often very small, perhaps fewer than 100 beneficiaries (Kiwara 1997). Other insurance schemes extend over the whole country and many communities and include up to 1 million or more beneficia- ries (Bennett, Creese, and Monasch 1998). The number of beneficiaries can change rapidly and neither reveals the financial balance of the schemes nor says much about the scheme's sustainability. Indeed, a few schemes had to be termi- nated after some years (Criel 1998; Bennett, Creese, and Monasch 1998), while others have been in operation for decades. Senegal has a relatively long tradition of mutual health insurance. The first experience began in the village of Fandène in the Thiès region in 1990. From the beginning, the movement in Senegal has been supported by a local health care provider, the nonprofit St. Jean de Dieu Hospital. Today, 16 mutual health insurance schemes operate in the area of Thiès. The main features of the schemes: · They are community based. · Ninety percent of them operate in rural areas. · With the exception of one mutual--Ngaye Ngaye--the schemes cover only hospitalization. · The mutuals have a contract with St. Jean de Dieu Hospital, where they get a reduction of up to 50 percent for treatment. · In general, the household is a member of a mutual, which participates in deci- sions. The member has a membership card on which he or she can list all or selected family members (beneficiaries). The membership fee is per person insured. Table 6.1 shows that a member has to pay a minimum amount of 3,000 F CFA for a treatment. For surgery, he pays 50 percent of the total costs for the opera- tion. The daily cost of hospitalization, including laboratory analysis, consulta- tion, and some radiography, is paid by the mutual, which receives a reduction of 50 percent. A mutual pays 3,750 F CFA per day for each member hospitalized; nonmembers pay 7,500 F CFA per day. In case of hospitalization, the member has to bring with him a letter of guarantee from the mutual's manager, which is issued if the member has paid the insurance premium regularly. A hospital stay of between 10 and 15 days is entirely paid by the mutual. If the hospitalization exceeds this limit, the mutual pays the hospital the entire invoiced amount because it guaranteed that it would do so. Afterward the member reimburses the Financial Protection and Access to Health Care in Rural Areas of Senegal 235 TABLE 6.1 Hospitalization Fees for Members and Nonmembers at St. Jean de Dieu Hospital Hospitalization Ticket for consultation Daily cost Operation (surgery) Members 3,000 F CFA 3,750 F CFA mutual 750 F CFA/unit Nonmembers 6,000 F CFA 7,500 F CFA 1,500 F CFA/unit Source: ZEF-ISED survey (2000). mutual in installments. To receive the described benefits, a mutual member has to pay a monthly premium of between 100 and 200 F CFA, and the head of household has to buy a membership card for 1,000 F CFA, a one-time fee. RESEARCH DESIGN AND METHODOLOGY Research Design A household survey was carried out by the Institute for Health and Development (ISED) in Dakar in cooperation with the Center for Development Research in Bonn. The survey began with a pretest in March 2000; the final stage of the sur- vey took place in May 2000. The participation rate in the interviews was very high--more than 95 percent. For the survey, we chose a two-stage sampling procedure. First, we selected 4 vil- lages out of the 16 villages in which mutuals operate. Each of the selected villages-- Fandène, Sanghé, Ngaye Ngaye, and Mont Rolland--has only one mutual, which has the same name as the village. Table 6.2 summarizes the major differences between the analyzed schemes: In a second step, we randomly selected households for the interviews. In all four villages, members and nonmembers were interviewed. To get a random sample from the four villages, we used household lists of all inhabitants (mem- bers and nonmembers) of the four villages to calculate the percentage distribu- tion between members and nonmembers and their respective weight in the TABLE 6.2 Selection Criteria for Mutual to Be Included in the Survey Name of mutual/ Years of Distance Percent of member village operation from hospital households in villages Services Fandène 10 6 km 90.3 Hospitalization Mont Rolland 4 15 km 62.6 Hospitalization Ngaye Ngaye 6 30 km 81.5 Primary health care Sanghé 3 8 km 37.4 Hospitalization Source: ZEF-ISED survey (2000). 236 Health Financing for Poor People: Resource Mobilization and Risk Sharing sample. We interviewed 346 households, 70 percent members and 30 percent nonmembers. The data set contains information on roughly 2,900 persons, 60 percent members and 40 percent nonmembers. This means that some house- hold heads have not insured all of their family. The data were entered immediately after completing the survey using SPSS Windows. In addition to the household survey, we interviewed key persons (leaders of the mutuals) to get complementary information about the mutuals' functioning, problems, and success. Methodology The modeling of mutual health insurance schemes' impact on health care use and expenditure faces the important challenge of dealing with the problem of "endo- genity" and "self-selection." This problem is currently receiving a great deal of attention in different areas of development economics, including measuring the impact of microfinance institutions (see Coleman 1999; Nada 1999), estimating the returns of education (see Bedi and Gaston 1999), and analyzing the impact of health insurance on various outcomes, such as health demand and financial pro- tection (see Waters 1999; Yip and Berman 2001). In each of these cases, the evalu- ation of a policy intervention or institutional innovation involves the problem of assigning individuals randomly to nonprogram control groups and others to pro- gram treatment groups. Thus the identification of an adequate control group is the first, and even the most important, step in trying to control for self-selection. With respect to the impact of health insurance on health care use, Waters (1999) names the potential endogenity of the choice of insurance as the main problem, leading to potential selection bias. Individuals who self-select into the insurance program have unobservable characteristics--related to preference or health status (adverse selection)--that might make them more likely than others to join the program and also might influence their decision to use health care ser- vices. An observed association between health insurance affiliation and health care use and expenditure may therefore be due not to the insurance but to the underlying unobservable characteristics. To control for this effect, in the Senegal study an omitted variable version of the Hausman test (Hausman 1978) is applied. This test is based on two steps: estimating the reduced form of the par- ticipation equation, and adding the fitted values to the health care demand equa- tion as a regressor. A significantly nonzero coefficient for the predicted value term is an indication that the suspected endogenous variable is in fact endogenous (Waters 1999).3 To specifically control for self-selection into the program, proxies for the health status and health risks have been included in all of the studies. Finally, village or district dummies are included to control for unobservable char- acteristics of communities, such as social values and solidarity, to see if they influ- ence individual choice to enroll in a community-financing scheme. To control for a sample selection bias in the demand equation for health care, the total sample is included, that is, those sick and those not sick as well as Financial Protection and Access to Health Care in Rural Areas of Senegal 237 those who are members and nonmembers. Finally, the models are checked for stability and robustness by adding and subtracting key variables and by apply- ing the F-test. To estimate the determinants of participation in a mutual health organization, we follow an approach applied by Weinberger and Jütting (2000). In that approach, participation in a local organization is dependent on the rational choice of an individual weighting costs and benefits of membership. It is assumed that participation of a household (p) in a mutual depends on the current income of the household (y), characteristics of the household head (H), who decides if the household joins or not, household characteristics (Z), community character- istics (C), and the error term u, who is uncovariant with the other regressors. The following equation describes our model: (6.1) pi = f (yi, Zi, Hi, C). To estimate the probability of participation we use a binary probit model: (6.2) pi* = yi + Zi + Hi + C + ui, pi = 1 if p* > 0, meaning the household is a member of the insurance scheme, i pi = 0 otherwise. To assess the impact of mutual health organization on financial protection of members, two aspects have to be taken into account: the probability of visiting a health care provider, and the out-of-pocket expenditure borne by the individu- als. The strong disadvantage of using health care expenditure alone as a predic- tor of financial protection is that this would allow to capture the lack of financial protection for those who choose not to seek health care because they cannot afford it. The first part of the model assesses the determinants of utilization, and we can thereby analyze whether membership in a mutual reduces barriers to assess health care services. We use a two-part model developed as part of the Rand Health Insurance Experiment in the United States (Manning and others 1987)4: a logit model that assesses the probability of visiting a health care provider (6.3) Prob (visit > 0) = X + M + u, where X stands as a vector for individual, household, and community characteristics (including membership); and a log-linear model that estimates the incurred level of out-of-pocket expendi- tures, conditioned on positive use of health care services: Log(out-of pocket expenditure / visit > 0) = Xy + M + e. X again represents a set of independent variables that are hypothesized to affect individual patterns of utilization, M represents a dummy variable for member- ship in a mutual health organization, and u and e represent terms of interfer- ence. The independent variables determining the demand for health care and expenditure in the case of illness are--among others--age, gender, education, health status, and income. 238 Health Financing for Poor People: Resource Mobilization and Risk Sharing As noted above, the modeling of a mutual health insurance scheme's impact faces the important challenge of dealing with the problem of endogenity and self-selection.5 To control for self-selection, the potential sources have to be identified. With respect to the impact of health insurance on health care use, Waters (1999) names the potential endogenity of the choice of insurance for health care use as the main problem, leading to potential selection bias. To con- trol for this source of self-selection, we estimate the demand for health insurance taking the entire sample instead of only those insured. This also helps us control for potential selection bias due to the potential endogenity of illness and health care use in a reduced sample of sick persons. To further reduce potential sources of self-selection by bias, each of the four selected villages, as outlined above, has a treatment group (members in a mutual) and a randomly assigned control group (nonmembers). In addition, in the models we have explicitly included a proxy for the health status of individu- als as an exogenous variable for health care use and expenditure. Selection bias due to village effects and health status is therefore controlled for. Furthermore, within a single household there are members and nonmembers, which reduces the potential bias related to systematically different preferences between mem- ber and nonmember households for health care use and expenditure. RESULTS Determinants of Membership in a Health Insurance Scheme Table 6.3 gives an overview of the variables included in the analysis of the deter- minants of participation. As outlined above, the decision of a household to par- ticipate in a mutual health organization is supposed to be influenced by individual, household, and community characteristics. The variables represent- ing individual characteristics of the household head involve age, education, gender, and membership in another organization. With respect to age, we hypothesize that younger household heads are more open to innovations (age group 1: positive coefficient) and that with increasing age people tend to partic- ipate less (age group 3: negative coefficient). Furthermore, we expect that better educated people and male-headed households tend to join a mutual more often than people with less education and female-headed households. The following characteristics of the household are supposed to influence membership in a mutual: income, ethnic group, religion, and the illness ratio (see table 6.3). The most important variable to be looked at in the context of our research question is income and its effect on the decision to participate or not. In our study, we have measured "income" as calculated by household expenditure per year and member.6 We assume that income has a positive influence on the deci- sion to participate and that the poorer strata of the population will not partici- pate because of difficulties in paying the premium. In addition, it will be interesting to analyze whether the richer part of the population participates, as Financial Protection and Access to Health Care in Rural Areas of Senegal 239 TABLE 6.3 Overview of Variables Used Exp. sign for Variable Description participation decision Individual and household characteristics Age group 1 Age between 21 and 40 years + Age group 3 Age between 61 and 90 years ­ Literacy (dummy) Ability to read/read and write (1 = yes) + Sex Male (1 = yes) + Other organization (dummy) Membership in another group (1 = yes) + Relationship (dummy)a Relation to household head (1 = self, spouse, parents, children, and 0 otherwise) + Income Log income/household member in F CFA + Income terziles Lower terzile ­ Middle terzile +/­ Upper terzile ­ Self-wealth Self-classification of household (poor, average, rich) ­; +/­; + Wolof (dummy) Household belonging to ethnic group of Wolof (1 = yes) + Religion (dummy) Christian household (1 = yes) + Illness ratio Number of cases of illness per household in the last six months divided by number of household members + Frequency of illnessa Number of cases of illness of an individual in the last six months Community characteristics Fandène (dummy) Household belonging to Fandène community (1 = yes) + Sanghé (dummy) Household belonging to Sanghé community (1 = yes) ? Ngaye Ngaye (dummy) Household belonging to Ngaye Ngaye community (1 = yes) ? Mont Rolland (dummy) Household belonging to Mont Rolland community (1 = yes) ? Solidarity (dummy) Perceived solidarity in the village (1 = yes) + a. These variables are only used in the equation of determinants of participation on the individual level (see table 6.5). this is important for risk-pooling reasons. Hence we include in the regression analysis income terziles, that is, we divided our sample into three subgroups-- rich, average, and poor. Added to the quantitative measures of wealth was rela- tive wealth. Households were asked to classify themselves according to relative 240 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 6.4 Marginal Coefficients for Determinants of Participation in Mutual Health Insurance (household level). Dependent Variable: Membership in a Mutual (1 if the household is a member and 0 otherwise). Variable Model 1 Model 2 Model 3 Constant ­2.048a ­0.223 0.064 (0.541) (0.155) (0.147) Individual characteristics of household head Sex (1 = male) 0.054 0.071 ­0.001 (0.083) (0.083) (0.083) Age group 1 (age 21­40) 0.088 0.085 0.079 (0.092) (0.092) (0.091) Age group 3 (age > 60) 0.087 0.079) 0.101 (0.061) (0.061) (0.062) Literacy (can read/read and write, 1 = yes) 0.059 0.062 0.043 (0.063) (0.063) (0.063) Other organization (membership in other group, 1 = yes) 0.180a 0.183a 0.120c (0.066) (0.066) (0.065) Household characteristics Wolof (household belonging to ethnic group of Wolof, 1 = yes) 0.249c 0.284b 0.229c (0.135) (0.137) (0.133) Religion (1 = Christian) 0.370a 0.369a 0.347a (0.085) (0.085) (0.083) Income (expenditures per household member log) 0.167a (0.046) Income terzile: lower ­0.110c (0.063) Income terzile: upper 0.165b (0.073) Self-wealth (self-classification of household): ­0.254a Poor (0.058) Self-wealth: rich 0.018 (0.113) Illness ratio (number of cases of illness per household divided by 0.002 0.007 0.037 number of household members) (0.088) (0.088) (0.086) (continued) wealth within the community on a rank from one (poorer than the average) to three (wealthier than the average). We expect the same findings in tendency for the relative measures as for the quantitative measures. We have included a dummy variable "Wolof" to measure the influence of belonging to a specific ethnic group. The Wolof are known for their openness to institutional innovations in the Senegalese context (Diallo 2000). The variable "religion" is included to take into account that the mutuals have an exclusive contract with the Catholic-owned St. Jean de Dieu Hospital. Moreover, the mutu- Financial Protection and Access to Health Care in Rural Areas of Senegal 241 TABLE 6.4 Continued Variable Model 1 Model 2 Model 3 Community characteristics Fandène (household belonging to Fandène community, 1 = yes) ­0.029 ­0.011 ­0.119 (0.151) (0.152) (0.150) Sanghé (household belonging to Sanghé community, 1 = yes) ­0.277b ­0.261c ­0.383a (0.132) (0.134) (0.130) Mont Rolland (household belonging to Mont Rolland community, 1 = yes) ­0.225 ­0.202 ­0.308b (0.139) (0.141) (0.137) Solidarity (perceived solidarity in the village, 1 = yes) 0.103 0.100 0.104c (0.066) (0.067) (0.065) Number of observations 338 338 341 Pseudo R 0.567 0.569 0.568 Chi 120.32 121.39 127.96 Prob > Chi 0.000 0.000 0.000 Frequencies of actual/predicted outcomes 80% 80% 80% a. Significant at 0.01 level. b. Significant at 0.05 level. c. Significant at 0.1 level. Source: Author's estimation based on ZEF-ISED survey data. als get active support from the diocese of Thiès. Hence we expect that more Christians will enroll than Muslims. We also assume a positive relationship between membership in a mutual and membership in other organizations. Peo- ple who already have experience participating in local organizations are more willing to join a mutual insurance group than people who have no experience participating. To control for adverse selection, we integrate the illness ratio of the household as a proxy for the health status. The variable describes the num- ber of cases of illness of household members in relation to the overall household size. It is assumed that less healthy households tend to join mutuals more than healthier ones, leading to adverse selection problems. Finally, we include dummy variables capturing village characteristics: acknowledgement of solidarity in the village (solidarity) and village factors. We assume that people acknowledging a high value of solidarity in their village tend to participate more. With respect to the village effects, we want to control for the type of insurance--whether it covers hospitalization or only primary health care (Ngaye Ngaye)--as well as for the specific local setting--the cultural environ- ment in the specific village and specific characteristics of the mutual, such as dis- tance to the hospital and the functioning of the mutual. The results presented in table 6.4 show the marginal effects of the probit analysis. Three different models were evaluated, differing in their definition of the income variable. In the first model, income is defined as a metric variable so 242 Health Financing for Poor People: Resource Mobilization and Risk Sharing as to analyze whether income has an influence on membership in a mutual. In the second model, income groups are established to determine effects between differ- ent income groups. In the third model, income groups were also formed, but in contrast to model 2 they were not based on expenditure but on self-assessment by the people surveyed.7 Table 6.4 shows that all three methods used are highly significant. Income has the anticipated positive influence on membership. Models 2 and 3 show further- more that the lower income groups in the villages are significantly less repre- sented in the mutuals. That means that the wealthy people in the communities are more likely to (be able to) participate in the insurance schemes. At the house- hold level, religion and ethnic identity also play an important role. The higher participation by Christians--the probability increases by nearly 40 percentage points compared with that for non-Christians--was to be expected because of the Catholic Church's intensive promotion of the mutuals. While household characteristics do have an influence on the membership decision, this is obviously not the case for the individual characteristics of the head of the household, such as education, gender, and age. None of these three characteristics is significant. Membership in other organizations, however, is a positive factor. People who have already experienced the advantages and disad- vantages of being associated with local groups are obviously more disposed toward membership in a health insurance scheme. The village effects that were discovered are also interesting. Different model variations show, for example, that the inhabitants of the villages of Sanghé and Mont Rolland have a significantly lower probability of membership than people from the villages of Ngaye Ngaye and Fandène. These results indicate that the different type of health insurance provided--primary health care in Ngaye Ngaye and inpatient care in the other three mutuals--had no significant influ- ence on the decision to participate. Instead, specific village factors, such as the management of the mutual, seemed to play a role. The mutual of Sanghé has faced several financial and managerial difficulties that led to a suspension of operations for some time. As a consequence, several people left the mutual. Efforts to reestablish the mutual have been successful, and today it is function- ing again, but with a lower participation rate than before. Thus far the results have shown that the main factors influencing the demand for health insurance in rural Senegal are religion, income, belonging to a certain ethnic group, access to a social network, and village effects. These results are largely confirmed by looking at the determinants of participation at the individual level. Regarding the individual level, it is interesting to analyze which type of household members are insured. From a theoretical perspective, one would assume that individuals more prone to the risk of illness are insured. As table 6.5 shows, this is largely confirmed as the probability for women and older people is higher than for men and younger persons in the household. It is reasonable to assume that women of child-bearing age and older people do need hospitalization care more often than other household members. TABLE 6.5 Marginal Coefficients for Determinants of Participation in Mutual Health Insurance (individual level). Dependent Variable: Membership in a Mutual Variable Model 2 Constant ­0.100c (0.056) Individual and household characteristics Sex (1 = male) ­0.042b (0.021) Age group 1 (age < 26) 0.000 (0.027) Age group 3 (age > 50) 0.077b (0.035) Literacy (can read/read and write, 1 = yes) 0.109a (0.022) Other organization (membership in other group, 1 = yes) 0.070b (0.028) Relationship (self, spouse, parents, children, 1 = yes) 0.115a (0.022) Health status (number of cases illness in last 6 months) ­0.011 (0.020) Wolof (household belonging to ethnic group of Wolof, 1 = yes) 0.182a (0.049) Religion (1 = Christian) 0.386a (0.033) Income terzile: lower ­0.047b (0.024) Income terzile: upper 0.219a (0.028) Community characteristics Fandène (household belonging to Fandène community, 1 = yes) ­0.058 (0.058) Sanghé (household belonging to Sanghé community, 1 = yes) ­0.358a (0.050) Mont Rolland (household belonging to Mont Rolland community, 1 = yes) ­0.332a (0.055) Number of observations 2.855 Pseudo R 0.549 Chi 989.02 Prob > Chi 0.000 Frequencies of actual/predicted outcomes 77% a. Significant at 0.01 level. b. Significant at 0.05 level. c. Significant at 0.1 level. Source: Author's estimation based on ZEF-ISED survey data. 243 244 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 6.6 Probability of Hospitalization and Determinants of Expenditure in Case of Hospitalization Variable Model 1a Model 1b Model 2a Model 2b (hospital) (hospital) (expend.) (expend.) Constant ­0.301a ­0.137a 4.611a 9.445a (0.065) (0.021) (2.016) (0.642) Individual and household characteristics Sex (1 = male) ­0.014b ­0.014b 0.370 0.401 (0.007) (0.006) (0.214) (0.21) Age group 1 (age < 26) ­0.016b ­0.016b ­0.495a ­0.520a (0.008) (0.008) (0.258) (0.210) Age group 3 (age > 50) 0.022b 0.022b ­0.008 ­0.141 (0.009) (0.009) (0.323) (0.327) Literacy (can read/read and write, 1 = yes) ­0.107 ­0.010 0.07 0.035 (0.007) (0.007) (0.243) (0.239) Membership (in health insurance without 0.020b 0.020b ­0.452b ­0.514b Ngaye Ngaye, 1 = yes) (0.009) (0.009) (0.287) (0.291) Frequency of illness 0.009 0.008 ­0.02 ­0.03 (0.006) (0.006) (0.16) (0.157) Type of illness (complications during 1.273b 1.125b pregnancy/childbirth, 1 = yes) (0.303) (0.299) Severity of illness (number of days hospitalized) 0.015a (0.005) Wolof (household belonging to ethnic group ­0.007 ­0.005 ­0.002 ­0.033 of Wolof, 1 = yes) (0.020) (0.019) (0.576) (0.582) Religion (1 = Christian household) ­0.005 ­0.004 0.089 0.142 (0.012) (0.012) (0.324) (0.323) Income (expenditures per household member log) 0.015a 0.441b (0.005) (0.174) Income terzile: lower ­0.008 ­0.120 (0.008) (0.273) Income terzile: upper 0.016b 0.67a (0.008) (0.238) (continued) Whereas the coefficient for both variables is significant, the marginal effect with less than 0.1 percentage points is rather low, which makes it too difficult to diag- nose severe adverse selection problems. Impact of Membership on Access to Modern Health Care Services In this section, we test the hypothesis that members of a mutual have better access to modern health care facilities than nonmembers. We measure access in two respects: the probability of frequentation of a health care facility, in this case a hospital, and the out-of-pocket expenditure at point of use. Our primary vari- Financial Protection and Access to Health Care in Rural Areas of Senegal 245 TABLE 6.6 Continued Variable Model 1a Model 1b Model 2a Model 2b (hospital) (hospital) (expend.) (expend.) Community characteristics Fandène (household belonging to Fandène 0.046b 0.046b 0.550 0.568 community, 1 = yes) (0.022) (0.022) (0.67) (0.676) Sanghé (household belonging to Sanghé 0.017 0.018 1.573 1.588 community, 1 = yes) (0.020) (0.020) (0.643) (0.643) Mont Rolland (household belonging to Mont Rolland 0.027 0.027 1.986c 1.779 community, 1 = yes) (0.022) (0.021) (0.636) (0.629) Number of observations 2,855 2,855 118 118 Chi/F value 103.00 103.96 3.990 4.176 Corrected r squared 0.264 0.289 Prob > Chi/F value 0.000 0.000 0.000 0.000 Frequencies of actual/predicted outcomes 94.7 % 94.7 % a. Significant at 0.01 level. b. Significant at 0.05 level. c. Significant at 0.1 level. Source: Author's estimation based on ZEF-ISED survey data. able of interest is membership in a mutual. We hypothesize that the probability of members' frequenting a hospital is higher, while at the same time they pay less for their treatment in comparison to nonmembers after controlling for indi- vidual, household, and community characteristics. This would mean that mem- bership has a positive coefficient for health care demand and a negative one for the effect on expenditure. Besides membership, the other key variable is income as we want to see how much of demand health care utilization and out-of- pocket expenditure is due to the income level and the ability to pay. As control variables we include age, gender, education, and frequency of ill- ness, which capture the need for health care and the health status of an individ- ual. Household characteristics are included to control for health preferences due to factors such as religion and belonging to an ethnic group. Finally, village effects are taken into account for differences in the cost of seeking health care as well as in the specific design of the mutuals. One assumption here is that inhab- itants from the village in Fandène have better access to health care because of their relatively short distance from the hospital as well as the fact that the mutual reportedly functions well. The results of the estimates for the determi- nants of demand for health care services and costs in the case of illness are pre- sented in table 6.6. Both models are highly significant. Of the 2,856 people, 151 have been in the hospital within the past two years.8 The findings of the estimates for both models suggest that the members of a mutual have better access to health care services 246 Health Financing for Poor People: Resource Mobilization and Risk Sharing than nonmembers. The probability of making use of hospitalization increases by 2 percentage points with membership, and expenditure in case of need is reduced by about 50 percent compared with nonmembers. Regarding individual characteristics, in addition to membership, age and gender play a role. More- over, the results suggest that younger people make less use of the hospital than the elderly, and they pay less on average if they do fall ill. Furthermore, women use the hospital more than men. Women go to the hospital especially when they have problems during pregnancy or childbirth. As far as the variables at the household level are concerned, it turns out that income has an impact on the demand for health care services and expenditure. The relatively better-off people in a community make more use of services and spend more money in the event of hospitalization. This is in line with findings on demand for health care in other developing countries (Gertler and van der Gaag 1990). With respect to village effects, people living in Fandène have a higher effec- tive demand for hospitalization than people in the other three communities.9 A possible explanation is the fact that Fandène is the oldest mutual, and according to our interview partners, it is well-organized and functions well. It is also the closest mutual to St. Jean de Dieu Hospital. To sum up, it can be said that members are (can be) hospitalized more often and pay considerably less for treatment than nonmembers. Other important fac- tors are age, type of illness, gender, income, and village effects. The case study on the community-based health insurance schemes in Senegal shows that the formation of a health insurance scheme for households in rural areas is possible and can result in better access to health care for otherwise excluded people. Especially in places where local institutions have already devel- oped forms of mutual help, possibilities seem to exist for developing the schemes into more formalized approaches. From the Senegalese case study, in addition to an existing local network, the existence of a viable health care provider is of tremendous importance. Without the financial support of the hos- pital as well as a perception that the care provided is of good quality--the hospi- tal is well known for its good quality in service provision--it is difficult to imagine that the mutuals would still exist. Hence subsidies seem to be necessary if one wants to set up an insurance scheme for poor people. Finally, individual and household characteristics also play a role in the viabil- ity of rural health insurance schemes. In areas with widespread poverty and a scattered population, setting up a health insurance scheme is much more diffi- cult than in richer and more densely populated areas. As the analysis of the determinants of participation in microinsurance schemes has revealed against the expectations of most donors and policymakers, they do not necessarily reach all population groups in a village. In fact, for the lowest income group the pre- mium to insure the whole family reaches nearly 8 percent of the household's annual income.10 Support for this group should therefore be secured by the state. This could be done, for example, in the form of subsidized premiums. Financial Protection and Access to Health Care in Rural Areas of Senegal 247 CONCLUSIONS The results of experience with mutual health organizations in Senegal suggest that rural health insurance for the poor is feasible under certain conditions. More important, it could be shown that access to health insurance can have a positive impact on the members' economic and social situation. Further investi- gation should be devoted to discovering the extent to which health insurance, or its lack, affects labor productivity and investors' willingness to undertake risky, but potentially profitable, investments. To enlarge access to health care for the poor and the rural population, community- based health insurance schemes can be an important element and a first step. They allow some limited pooling of risks and thereby lead to an improvement in the health care system, since most people otherwise have to pay their health expenditure out-of-pocket. However, the study also points to the persistent problem of social exclusion--that the community's poorest members have no opportunity to participate and not enough resources to pay the required pre- mium. To overcome these limitations of community-based health insurance, broader risk pools are required. In particular, the role of external financial sup- port, such as government subsidies, donor funding, and reinsurance in encour- aging social inclusion needs to be further explored. More research is needed on how these schemes can be scaled up, replicated, and linked to other social risk- management instruments such as social funds. Acknowledgments: The author is grateful to the World Health Organization (WHO) for hav- ing provided an opportunity to contribute to the work of the Commission on Macroeconomics and Health and to the World Bank for having published the material in this chapter as an HNP Discussion Paper. The author would like to thank Hana Ohly for her valuable research assistance. Financing from the ILO-STEP project is gratefully acknowledged. The findings, interpretations, and con- clusions expressed in the chapter are entirely those of the author and do not necessarily repre- sent the views of the World Bank, its Executive Directors, or the countries they represent. NOTES 1. The terms community-based health insurance and mutual health insurance are used inter- changeably throughout this chapter. 2. For a more detailed typology see Jakab and Krishnan (2001). 3. The test of endogenity of the membership variable in health care use and expenditure had to be rejected, that is, we supposed that membership is exogenous. 4. For a recent application see Yip and Berman (2001). 5. The problem of self-selection is not relevant for our first research question on determi- nants of participation. But it is relevant for the second research question, which looks at the demand for health care and the amount of expenditure, with "health insurance membership" as an exogenous variable. 248 Health Financing for Poor People: Resource Mobilization and Risk Sharing 6. Alternatively, we have also measured income as calculated by the revenue from on- farm and off-farm activities as well as remittances. It turned out, however, that there was some estimation bias in the data because some of the people interviewed were unwilling to report their true income. 7. Estimating the income of households in developing countries is difficult. Since many of the people surveyed are reluctant to reveal their real income, income is generally measured by using expenditure. This method of measuring can be supplemented by asking the protagonists to do a self-assessment, comparing themselves with other households in the neighbourhood. 8. A certain percentage of the hospitalized persons had to be excluded from the "expen- diture" analysis as they were not aware of the costs they had to pay because other family members made the payments. 9. This effect clearly pops up when the Fandène mutual is left outside and the remain- ing mutuals get a significant negative coefficient. 10. An individual household has to weigh these costs against the probability of being hospitalized and the average cost for treatment. The direct average financial costs for one hospitalization of a household member is already above 20 percent of the annual income of the household. REFERENCES Asfaw, A., A. Admassie, J. von Braun, and J. Jütting. 2001. "New Dimensions in Measuring Economic Costs of Illness: The Case of Rural Ethiopia." Submitted to Social Science and Medicine. Atim, C. 1998. Contribution of Mutual Health to Financing, Delivery, and Access to Health Care: Synthesis of Research in Nine West and Central African Countries. Technical Report 18. Partnerships for Health Reform Project. Abt Associates Inc., Bethesda, Md. Bedi, A. and N. Gaston. 1999. "Using Variation in Schooling Availability to Estimate Edu- cational Returns for Honduras." Economics of Educational Review 18: 107­16. Bennett S., A. Creese, and R. Monasch. 1998. Health Insurance Schemes for People Outside Formal Sector Employment. ARA Paper 16. WHO, Geneva. Coleman, B. 1999. "The Impact of Group Lending in Northeast Thailand." Journal of Development Economics 60: 105­41. Criel, B. 1998. "District-Based Health Insurance in Sub-Saharan Africa. Part II: Case- Studies." Studies in Health Services Organisation and Policy (Antwerp) 10. Diallo, I. 2000. "Impact des mutuelles de santé sur l'accessibilite des populations aux soins de santé modernes dans la région de Thiès au Sénègal." Institute for Health and Devel- opment, Dakar. Dror, D., and C. Jacquier. 1999. "Microinsurance: Extending Health Insurance to the Excluded." International Social Security Review 52(1): 71­97. Gertler, P., and J. van der Gaag. 1990. The Willingness to Pay for Medical Care. Baltimore: World Bank, Johns Hopkins University Press. Gilson, L. 1998. "The Lessons of User Fee Experience in Africa." In A. Beattie, J. Doherty, L. Gilson, E. Lambo, and P. Shaw, eds., Sustainable Health Care Financing in Southern Financial Protection and Access to Health Care in Rural Areas of Senegal 249 Africa: Papers from an EDI Health Policy Seminar Held in Johannesburg, South Africa, June 1996. Washington, D.C.: EDI Learning Resources Series. Griffin, C. 1992. Health Care in Asia: A Comparative Study of Cost and Financing. World Bank Regional and Sectoral Studies. Washington, D.C. Heckman, J. 1979. "Sample Bias as a Specification Error." Econometrica 47(1): 153­61 Jakab, M., and C. Krishnan. 2001. "Community Involvement in Health Care Financing: Impact, Strengths and Weaknesses: A Synthesis of the Literature." Background paper prepared for Working Group 3 of the Commission on Macroeconomics and Health of the WHO. World Bank, Washington, D.C. Jütting, J. 2000. "Social Security Systems in Low Income Countries: Concepts, Constraints, and the Need for Cooperation." International Social Security Review 53(4): 3­25. ------. 2001. "Health Insurance for the Poor?" In Development and Cooperation 6: 4­5. Kiwara, A. 1997. "Mutual Society for Health Care in the Informal Sector (UMASIDA) Backup Report--January to August 1997." Dar es Salaam. Manning, W., J. Newhouse, N. Duan, E. Keeler, A. Leibowitz, and M. Marquis. 1987. "Health Insurance and the Demand for Medical Care: Evidence from a Randomized Experiment." American Economic Review 77: 251­77. Musau, S. 1999. Community-Based Health Insurance: Experiences and Lessons Learned from East and Southern Africa. Technical Report 34. Partnerships for Health Reform Project, Abt Associates, Inc., Bethesda, Md. Nada, P. 1999. "Women's Participation in Rural Credit Programmes in Bangladesh and Their Demand for Formal Health Care: Is There a Positive Impact?" Health Economics and Econometrics 8: 415­28. Tine, J. 2000. Les mutuelles de santé rurales de la région de Thiès au Sénegal: Des initiatives communautaires pour améliorer l'accès aux soins de santé. ILO/ZEF Project Report 4. Cen- ter for Development Research, Bonn. Waters, H. 1999. "Measuring the Impact of Health Insurance with a Correction for Selec- tion Bias: A Case Study of Ecuador." Health Economics and Econometrics 8: 473­83. Weinberger, K., and J. Jütting. 2000. "The Role of Local Organizations in Risk Manage- ment: Some Evidence from Rural Chad." Quarterly Journal of International Agriculture 39(3): 281­99. ------. 2001. "Women's Participation in Local Organizations: Conditions and Con- straints." World Development 29(8): 1391­404. WHO (World Health Organization). 2000: World Health Report 2000: Health Systems-- Measuring Performance. Geneva. Wiesmann, D., and J. Jütting. 2001. "Determinants of Viable Health Insurance Schemes in Rural Sub-Saharan Africa." Quarterly Journal of International Agriculture 50(4): 361­78. Yip, W., and P. Berman. 2001. "Targeted Health Insurance in a Low-Income Country and Its Impact on Access and Equity in Access: Egypt's School Health Insurance." Health Economics 10: 207­20. ZEF-ISED. 2000. Survey info. CHAPTER 7 Community-Based Health Insurance in Rwanda Pia Schneider and François Diop Abstract: This chapter evaluates the impact of prepayment schemes on access to health care for poor households, based on household survey data. Rwanda is one of the poorest countries in the world. After the genocide in 1994, public health care services were pro- vided free to patients, financed by donors and the government. In 1996, the Ministry of Health reintroduced prewar level user charges. By 1999, utilization of primary health care services had dropped from 0.3 in 1997 to a national average of 0.2 annual consultations per capita. This sharp drop in health service use, combined with growing concerns about rising poverty, poor health outcome indicators, and a worrisome HIV prevalence among all population groups, motivated the Rwandan government to develop community- based health insurance to assure access to the modern health system for the poor. In early 1999, the Rwandan Ministry of Health, in collaboration with the local communities and with the technical support of the USAID-funded Partnerships for Health Reform (PHR) project, began the process to pilot test 54 prepayment schemes in three districts. At the end of their first operational year, the 54 schemes comprised more than 88,000 members. The findings presented in this chapter reveal that insurance enrollment is determined by household characteristics such as the health district of household residence, education level of household head, family size, distance to the health facility, and radio ownership; health and economic indicators did not influence enrollment. Insurance members report up to five times higher health service use than nonmembers. The analysis confirms find- ings reported by PHR based on provider data: health insurance has significantly improved equity in health service use for members while out-of-pocket spending per episode of illness has decreased. I n response to declining health service utilization after the reintroduction of user fees for services and drugs in public health facilities in 1996, the Rwandan Ministry of Health (MOH) decided to pilot test alternative health care financing and provider payment methods. During the first six months of 1999, the MOH developed an ongo- ing collaboration with the local communities involving 54 community-based health insurance (CBHI) plans--each of them partnering with a health center. The MOH selected three pilot districts, Kabutare, Byumba, and Kabgayi, to pilot test these prepayment schemes (PPS). The three districts were chosen based on the extent of their health infrastructure, the repeated demand for technical assistance from the population in developing and implementing CBHI, and the districts' political will to participate in the health insurance pilot experience. By July 1, 1999, all 54 health centers, which had collaborated in the set up of their CBHI in the three pilot districts, signed a contract with their partnering health insurance scheme. Thereafter, the district population began to enroll in the schemes, which are 252 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 7.1 Community-Based Health Insurance in Rwanda, First Year Performance (July 1999­June 2000) Pilot districts with CBHI Indicators Byumba Kabgayi Kabutare All 3 districts All prepayment schemes (status on 6/30/2000) Total number of PPS 21 17 16 54 Total target population in districts 459,329 368,020 288,160 1,115,509 Total population enrolled 48,837 21,903 17,563 88,303 Average number of members per PPS 2,326 1,288 1,098 1,635 First year average PPS enrollment rate 10.6% 6.0% 6.1% 7.9% Source: Schneider and others (2001). democratically managed by their members as mutual health associations. At the end of the first year, membership in the 54 health insurance plans rose to 88,303 individuals, corresponding to 8 percent of the total population of the three dis- tricts (see table 7.1). Technical and financial assistance for the entire design, development, implementation, and evaluation phase was provided to the Rwan- dan government by Partnerships for Health Reform (PHR), a project funded by USAID and administered by Abt Associates. (For detailed information on the development and implementation of prepayment schemes in Rwanda, see Schneider, Diop, and Bucyana 2000.) The objective of this chapter is to respond to two questions about the pre- payment schemes' impact: What are the population groups that enroll in com- munity-based health insurance schemes? Does health insurance membership improve financial accessibility to care without increasing the burden of out-of- pocket health expenditures? These questions are addressed by presenting a syn- opsis of the findings from the household survey conducted by PHR in the three health districts. The impact of prepayment schemes on insurance and providers' utilization, cost, and finances has been analyzed from monthly routine data col- lected from providers and health insurance schemes over a two-year period in the three districts. Findings of this detailed analysis on the financial sustainabil- ity of CBHI plans and their impact on health care providers are presented in the PHR Technical Report No. 61 (Schneider and others 2001). This chapter will therefore respond to the question about the insurance impact on households' financial accessibility to the modern health care system by focusing on informa- tion collected in the household survey. The next section provides a brief summary of the design, development, and implementation of prepayment schemes in Rwanda, which took place from Jan- uary 1999 until September 2000. The third section introduces the method used to address the research questions and presents the household survey data and variables used. Results are presented in the fourth section; and the fifth section has a discussion and conclusion. Community-Based Health Insurance in Rwanda 253 BACKGROUND In January 1999, the Rwandan MOH initiated the design and development phase of prepayment plans by creating a strategic steering committee, which was headed by the director of health care and included representatives and stake- holders from the central and regional levels. The design and implementation of health insurance modalities and management features were discussed and agreed upon during 28 district-level workshops attended by community and health care representatives and in a series of community gatherings with the local populations. Proposals stemming from these district and community meet- ings were shared with the central steering committee, which provided feedback and advice to the communities. As a result of this ongoing discussion between the central and local levels, the scheme features were designed, the legal, con- tractual, and financial tools were developed, and workshop participants were trained and prepared to manage the 54 prepayment schemes, each entering into partnership with a health center on July 1, 1999. Under Rwandan law, the schemes are deemed mutual health associations, headed by an executive bureau with four volunteers, elected by and among the scheme members during a CBHI general assembly. At the district level, the schemes have formed a federation. Six members have been elected by and among all PPS executive bureau representatives in their general assembly to con- stitute the district federation of prepayment schemes. The federation is the part- ner to the district hospital as well as to the health district and other authorities. Each prepayment bureau has signed a contract with the affiliated health center, and each federation with the district hospital, defining in 17 articles the rules of collaboration between the insurer and the provider. According to the schemes' bylaws, members are invited at least once a year to attend the prepayment scheme general assembly. Individuals and households who would like to be insured pay, at the time of enrollment, an annual premium of 2,500 francs per family, for up to seven per- sons (July 1999: RWF 2,500 = US$7.50). The premium is paid to the CBHI affili- ated with their "preferred" health center.1 In case of sickness, members first contact their preferred health center, which is usually the nearest public or church-owned facility. Health centers play a gatekeeper function; hospital ser- vices are covered for members only if the preferred health center has referred them. This is done to dissuade members and providers from frivolous use of more expensive hospital services. PPS membership entitles members--after a one-month waiting period--to a basic health care package covering all services and drugs provided in their preferred health center, including ambulance trans- fer to the district public or church-owned hospital, where a limited package is covered.2 Members pay a 100 francs copayment for each health center visit (July 1999: RWF 100 = US$0.30). The MOH was concerned that the availability of health insurance to a popu- lation group with an accumulated demand for health services could lead to 254 Health Financing for Poor People: Resource Mobilization and Risk Sharing adverse selection and moral hazard, causing health care costs to rise. Therefore, the MOH recommended to the district workshop participants that they incorporate into the design of the prepayment schemes a provider-payment mechanism that would set the necessary incentives for providers to improve their productivity while controlling for unnecessary use of health services. After several discussions between providers and future scheme managers, the workshop participants selected capitation provider payment to the health center as a measure for control- ling cost escalations caused by supply-side induced increases in demand for health care. Each prepayment bureau disburses monthly one-twelfth of its accumulated premium fund, 5 percent of which is withheld to cover the scheme's administrative costs, 10 percent is paid to the district's prepayment federation, and the rest is paid as capitation payment to the partnering health center. The federation reimburses the district hospital for covered services provided to members, paying per episode of illness (for caesarean sections, malaria, and pediatrics), and by service for overnight stays and physician consultations for all other illnesses. Thus insurance members share their hospital costs on a district level and health center costs on a health center catchment-area level. (For a detailed analysis of PPS impact on uti- lization, cost, and finances in health centers, see Schneider and others 2001.) DATA SOURCES AND METHODOLOGY The analysis presented in this chapter is based on data collected in the prepayment scheme household survey, conducted by PHR in collaboration with the Rwandan National Population Office (ONAPO). Data collection took place during 40 days in October­November 2000. The household survey includes 2,518 households that were successfully interviewed in the three pilot districts, and number 11,583 indi- viduals. The sample was designed to provide information on the impact of prepay- ment schemes on households' enrollment and health care seeking behavior, as well as on the related financial implications. The sample was based on the same sampling frame as the Rwandan Demographic and Health Survey (DHS) 2000, covering 11 health regions in Rwanda.3 Households for the prepayment house- hold survey in the three districts were sampled at random from a list of primary households from sample cells identified in the national DHS sample, rendering the household survey sample representative to the district level. The prepayment household survey used three structured questionnaires for data collection: a socioeconomic household questionnaire, a curative question- naire, and a preventive care questionnaire. The household questionnaire col- lected information on households' and individuals' sociodemographic and economic characteristics, including household expenditures for consumer goods, health, and education, and participation in CBHI. The curative care ques- tionnaire was addressed to household members who were sick two weeks prior to the interview, and the preventive care questionnaire was used to interview women of childbearing age who had delivered a child in the past five years or Community-Based Health Insurance in Rwanda 255 who were pregnant during the year preceding the interview (Diop, Schneider, and Butera 2001). The three pilot districts (Byumba, Kabgayi, Kabutare) are similar in their socioeconomic situation. There is little urban activity, and the populations are mainly active in agriculture and animal husbandry. Households are assumed to be equally poor, with few of them owning cattle, a sign of wealth. For the analy- sis, the sample population is divided into two groups: CBHI members in pilot districts, and CBHI nonmembers in pilot districts. Three models are used in this chapter to estimate first, the probability of buying health insurance for specific population groups in the three pilot districts; second, the probability of access to basic health care services for the insured and uninsured population groups in the pilot districts; and third, the estimated out-of-pocket health expenditures per episode of illness for all sick individuals and for those who sought professional care, based on a set of explanatory variables. For each categor- ical variable used in the three models, one category has been selected as a refer- ence category. Odds ratios are estimated in the logit regression models for each category to estimate the factor that measures the magnitude of the difference in relation to the reference category. Interaction effects were tested for significance. Model 1: Demand for Health Insurance The following model estimates the probability of CBHI enrollment for households in pilot districts. The objective is to determine if the poorest buy basic health insurance, and if the poor benefit from a redistribution from richer members of the financial pool. The willingness to join CBHI is a discrete choice--to join or to not join. A logit regression model is used to determine households' CBHI enroll- ment probability and the extent to which this decision is influenced by specific sociodemographic and economic characteristics. The hypothesis to be tested is that the CBHI member and nonmember households do not differ in their socio- economic characteristics. In a logit regression, the dependent variable "demand for insurance" (Di) will equal 1 if individuals buy insurance or 0 otherwise. Formally, the logit model can be written as a linear function of the explanatory variables: (7.1) Li = b1 + b2 X2i + .... + bk Xki and Pi (D for CBHI membership) = 1/(1 + 1/e ). Li The second equation shows that the conditional probability of buying insurance Pi is a nonlinear function of the explanatory variables Xi, which represents a series of attributes assumed to have caused a household to buy health insurance membership in the three pilot districts. We will estimate the unknown coeffi- cients bi, which are the weights assigned to each of the households' sociodemo- graphic and economic characteristics in the probability that Di = 1 for given Xi. Insurance was an option only for households within the pilot districts. Therefore, 256 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 7.2 Summary Statistics: Independent Variables Used to Determine Probability of CBHI Independent variables Mean Std. dev. N (Households) Kabgayi District 0.41 0.492 2,518 Byumba District 0.22 0.411 2,518 Kabutare District 0.37 0.482 2,518 Male HH head 0.66 0.474 2,518 Average age HH head 44.33 15.789 2,512 HH head 40 years and older 0.83 0.377 2,518 HH head attended school 0.54 0.498 2,515 HH with child < 5 0.07 0.249 2,518 HH with pregnancy in past year 0.03 0.167 2,518 HH with cattle 0.18 0.386 2,503 HH with radio 0.34 0.475 2,505 HH with 5 or more members 0.45 0.498 2,518 HH size 4.57 2.255 2,518 Less than 30 minutes from HH cell to health facility 0.38 0.487 2,518 Income quartiles 1 0.26 0.437 2,518 2 0.25 0.431 2,518 3 0.25 0.435 2,518 4 0.24 0.430 2,518 the logit regression was performed with weighted household survey data from pilot districts, based on household heads as the unit of analysis. Description of Variables Included in Model 1 The response to enroll in CBHI is the dependent variable in this model and is made primarily by the head of household, based on a set of independent vari- ables X. These explanatory variables are classified into demographic, socioeco- nomic, and health attributes of the household. Table 7.2 presents their sample size, mean, and standard deviation. Model 2: Access to the Modern Health Care System Patients' health-seeking behavior was measured based on weighted data for individ- uals who reported sickness during the two weeks preceding the interview in the household survey and who responded to the curative care questionnaire. As in the first model, the second model applies a logit regression model to estimate the prob- ability of entering (or not entering) the modern health care system for the insured and uninsured in the pilot districts. Access probabilities are estimated based on spe- Community-Based Health Insurance in Rwanda 257 TABLE 7.3 Summary Statistics of Explanatory Variables for Probability of Professional Visit Independent variable Mean Standard deviation N (Sick individuals) Kabgayi District 0.23 0.421 3,130 Byumba District 0.37 0.482 3,130 Kabutare District 0.40 0.491 3,130 Prepayment member in pilot district 0.06 0.236 3,130 Male patient 0.42 0.494 3,130 Average age patient 24.60 21.003 3,127 Patient age 0­5 years 0.23 0.423 3,127 Patient with pregnancy in past year 0.07 0.251 3,130 Patient spent 4 or more days in bed 0.56 0.497 1,599 HH with 5 or more members 0.57 0.495 3,130 HH head attended school 0.56 0.496 3,130 HH with cattle 0.21 0.404 3,130 HH with radio 0.36 0.481 3,130 Less than 30 minutes from HH cell to health facility 0.39 0.488 3,130 Income quartiles 1 0.23 0.421 3,130 2 0.26 0.441 3,130 3 0.26 0.439 3,130 4 0.24 0.430 3,130 cific sociodemographic and economic household characteristics that determine a sick individual's care-seeking behavior. The hypothesis is tested that the sick who access health care do not significantly differ in their sociodemographic, economic, and health characteristics. Therefore, the logit regression was performed with weighted curative survey data from the pilot districts, based on sick individuals as the unit of analysis. The logit model, based on equation 7.1 presented in the first model, leads to the following definition of the probability of accessing modern health care: (7.2) Pi (access to professional care) = 1/(1 + 1/e ), Li where X represents a set of explanatory variables that are assumed to have caused a sick person to seek care with a professional provider at a health center or district hospital during the two weeks prior to the interview. Description of Variables Included in Model 2 The decision to seek professional care is influenced by households' socioeco- nomic conditions, insurance status, and the sick individuals' health status. These explanatory variables are summarized in table 7.3, showing for each attribute the sample size, mean, and standard deviation. 258 Health Financing for Poor People: Resource Mobilization and Risk Sharing Model 3: Financial Impact of Household Out-of-Pocket Health Expenditures The third model is a log-linear regression that serves to estimate first, that sick indi- viduals' total average out-of-pocket spending per episode of illness and second, that total out-of-pocket spending conditioned on the positive use of health care services (Manning and others 1987; Yip and Berman 2001). The model is a linear regression for the logarithm of total health-related spending per episode of illness of the sick and for the logarithm of total health-related spending for the sick who reported at least one visit. The logarithmic transformation of health expenditures per episode of illness eliminates skewness in the distribution of health expenses among users, yielding roughly normal error distributions. The model can be written as follows: (7.3) Log (total illness related out-of-pocket spending) = a + b X + e and Log (total illness related out-of-pocket spendingvisit > 0) = a + b X + e, where X represents a set of continuous and dummy attributes assumed to influence patients' health expenditures. Detailed health expenditures are reported by episode of illness, which includes spending before and during a professional care visit, and will show to what extent patients rely on alternative sources of care outside the for- mal health sector. The regressions were performed with weighted curative survey data from the pilot districts, based on sick individuals as the unit of analysis. It is assumed that the amount spent on nonprofessional medicine will be higher for patients whose access to professional care is limited by financial barriers. Description of Variables Included in Model 3 Model 3 uses the same variables as in the second model and adds the variable pro- fessional care visit to estimate out-of-pocket health expenditures for those indi- viduals who were sick and for those who sought professional care. Table 7.4 presents sample size, mean, and standard deviation for each explanatory variable. RESULTS Description of the Sample Group Tables 7.5 and 7.6 describe sociodemographic and economic characteristics for the sample population included in the household survey conducted in the three pilot districts. Males head about 82 percent of PPS member households. One- third of the heads of member households have five or more years of schooling, 41 percent of the household heads are in the 40­59 age group and belong to households in higher expenditure quartiles as compared with the nonmember households in the same districts. Distance to the health facility also seems to be Community-Based Health Insurance in Rwanda 259 TABLE 7.4 Summary Statistics of Explanatory Variables for Estimated Out-of-Pocket Spending per Episode of Illness Independent variable Mean Standard deviation N (Sick individuals) All sick with 1+ professional care visits 0.16 0.371 3,130 Kabgayi District 0.23 0.421 3,130 Byumba District 0.37 0.482 3,130 Kabutare District 0.40 0.491 3,130 Prepayment member in pilot district 0.06 0.236 3,130 Male patient 0.42 0.494 3,130 Average age patient 24.60 21.003 3,127 Patient age 0­5 years 0.23 0.423 3,127 Patient with pregnancy in past year 0.07 0.251 3,130 Patient spent 4 or more days in bed 0.56 0.497 1,599 HH with 5 or more members 0.57 0.495 3,130 HH head attended school 0.56 0.496 3,130 HH with cattle 0.21 0.404 3,130 HH with radio 0.36 0.481 3,130 Less than 30 minutes from HH cell to health facility 0.39 0.488 3,130 Income quartiles 1 0.23 0.421 3,130 2 0.26 0.441 3,130 3 0.26 0.439 3,130 4 0.24 0.430 3,130 an important criterion as almost 50 percent of the member households live within 15 minutes of the health facility. Membership begins to taper off as the distance to the health facility increases. Almost 61 percent of insured households interviewed in the survey said all individuals living in the household had been enrolled. In the remaining house- holds, individuals not enrolled were usually young adults above the age of 18, who are supposed to enroll either in a group category or as individuals.4 Table 7.6 presents households' monthly average per capita monetary expenditures for each expenditure quartile, as well as the average household size for insured and uninsured households. Independent means tests were performed to compare dif- ferences. Households' monetary expenditures were used as proxy to classify households in income quartiles. Insured households number on average signifi- cantly more individuals than uninsured ones in the pilot districts. The possibil- ity of signing up in a CBHI plan as a family of up to seven members for the same annual premium might have been an incentive for larger households to enroll with all their family members. Each of the two groups shows a decreasing aver- age household size with higher expenditure quartiles. 260 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 7.5 Descriptive Sample Characteristics (Column Percentages Sum to 100 Within Each Category) Pilot districts Nonmembers (N 2,337) PPS members (N 181) Total (N 2,518) Characteristics of head of household Gender Female 35.4% 17.8% 34.1% Male 64.6% 82.2% 65.9% Level of schooling Never 47.0% 28.4% 45.7% Primary < 5 22.6% 22.5% 22.6% Primary = or > 5 24.2% 33.3% 24.9% Above primary 6.2% 15.9% 6.9% Age group < 30 17.1% 17.6% 17.2% 30­39 25.2% 27.2% 25.4% 40­59 36.3% 41.2% 36.7% 60 & + 21.3% 14.0% 20.8% Characteristics of household Income quartiles 1 26.1% 19.8% 25.6% 2 24.7% 24.2% 24.6% 3 25.0% 29.9% 25.3% 4 24.3% 26.1% 24.4% Time distance (minutes) to health facility 15 37.6% 48.6% 38.4% 45 14.6% 20.8% 15.1% 75 25.6% 24.2% 25.5% 105 22.1% 6.4% 21.0% Smaller households in higher expenditure quartiles pay the same premium per household as larger families in the lowest expenditure quartiles. Depending on members' service use and financial contribution to health for uncovered ser- vices, this negative relationship between household size and income status can lead to cross-subsidies from the smaller, predominantly richer to the larger, poorer families in the prepayment health insurance pool. Table 7.7 presents the summary statistics for monthly per capita monetary expenditures for households living in the pilot districts. Figure 7.1 shows that the monetary expenditure data are very right-skewed, confirming that this is an equally poor population with very few households reporting high monetary expenditures: the 90th percentile amount is 5,975 francs compared with the maximum amount of 192,950 francs. The figure shows that Rwandan households living in these rural districts are poor. They live mostly from subsistence farming in areas with a high-density Community-Based Health Insurance in Rwanda 261 TABLE 7.6 Household Characteristics, by Income Quartile Household characteristics Nonmembers PPS members Total Monthly average per capita expenditure (RWF) Quartile 1 333 347 334 Quartile 2 1,050 1,007 1,047 Quartile 3 2,241 2,056 2,225 Quartile 4 8,154 9,367 8,247 In all 4 quartiles 2,884 3,370 2,919 Average household size, number of individuals Quartile 1 4.6 5.5 4.7 Quartile 2 4.7 5.6 4.8 Quartile 3 4.6 5.8 4.7 Quartile 4 4.1 4.8 4.1 In all 4 quartiles 4.5 5.5a 4.6 Note: t-tests were performed to compare the average values of the insured with the uninsured sample. a. Significant at 1 percent level of significance. TABLE 7.7 Summary Statistics on Distribution of Monthly Monetary Expenditures per Capita Monetary expenditures Households N 2,518 Mean RWF 2,919.1 Standard Error of Mean 126.798 Median 1,475.5 Standard Deviation 6,362.787 Minimum 0 Maximum 192,950 Percentiles 10 267.0 25 624.5 50 1,475.5 75 3,190.0 90 5,975.0 population. Rwanda is recovering from the recent civil war, and an estimated 10 percent of the male population is still missing. Muller (1997) found in a house- hold survey conducted in 1983 that the average land area farmed by Rwandan households is very small at 1.24 hectares, and households produced agricultural product worth an average of US$51 per capita per year, 90 percent of which is 262 Health Financing for Poor People: Resource Mobilization and Risk Sharing FIGURE 7.1 Monthly Monetary Expenditure per Capita used for consumption. Findings in our household survey, conducted 17 years later, estimated annual monetary expenditure is approximately US$100 per capita for these rural households. The Rwandan Ministry of Finance is conduct- ing a living standard survey in Rwanda, which will provide more insight into the socioeconomic conditions of Rwandan households. Model 1: Who Demands Health Insurance? The means comparison in tables 7.1 and 7.2 has shown that, compared with the uninsured, insured households are more likely to be headed by a male individual who has attended some schooling. In addition, proportionally more CBHI mem- ber households are likely to come from higher income quartiles and from larger households. The logit regression results presented in table 7.8 show that the household head's level of education, family size, district of residence, distance to the health facility, and radio ownership are the major determining factors in whether to join a health insurance plan. Households' health and economic indicators did not influence demand for health insurance. Radio ownership is indicative of a household's ability to access information and exposure to advertising about the CBHI. It may also be seen to a certain extent as an economic indicator for these very poor households. Households in Kabgayi are more than twice as likely, and those in Byumba about 15 times as likely, to buy health insurance as households in Kabutare. Household heads who attended school are 103 percent more likely to enroll in health insurance than the illiterate. Households with five and more members are Community-Based Health Insurance in Rwanda 263 TABLE 7.8 Logit Regression Results for Households' Probability to Demand Community-Based Health Insurance (Prepayment Schemes) Insurance membership in pilot districts Explanatory variable Reference category variable Odds ratio S.E. Sign Kabgayi District Kabutare District 3.51a 0.362 0.001 Byumba District 15.80a 0.268 0.000 Male HH head Female HH head 1.55 0.253 0.084 HH head, age 40+ HH head, younger than 40 1.13 0.239 0.598 HH head, attended school HH head, illiterate 2.03a 0.196 0.000 Large HH size, 5+ Small HH size, fewer than 5 1.60a 0.189 0.013 HH with child < 5 No child < 5 0.87 0.488 0.768 HH with pregnancy in past year No pregnancy in past year 1.23 0.674 0.761 Less than 30 minutes to health facility More than 30 minutes to 3.96a 0.187 0.000 health facility HH with cattle No cattle 1.28 0.210 0.237 HH with radio No radio 1.47a 0.184 0.038 Quartile 1 Quartile 4 1.19 0.264 0.513 Quartile 2 1.21 0.244 0.437 Quartile 3 1.15 0.228 0.535 Ancillary statistics N (households) 2,474 ­ 2 Log likelihood 1,054.901 Goodness fit (chi-squared 236.998 test) Degree of freedom 14 Nagelkerke R Square 22% Note: HH = Household. Z-tests were performed to test the probability of enrollment for each characteristic in a logit model. a. Significant at 1 percent level of significance. 60 percent more likely to buy insurance than smaller households. This is proba- bly because, irrespective of the family size (up to seven members), households pay a 2,500 francs membership fee per year.5 Therefore, larger families effectively pay less per household member. Households who live within 30 minutes of their health facility have a 296 percent higher probability of joining than those who live farther away. This latter result might have been influenced by health centers' and prepayment schemes' awareness campaigns, which could have been more intense in the vicinity of a health facility. Households who own a radio are 47 percent more likely to enroll than those without a radio, another result that might have been caused by the regular awareness campaigns transmitted by radio. Although male-headed households are 55 percent more likely to join than female-headed, and households with pregnant women are 23 percent more 264 Health Financing for Poor People: Resource Mobilization and Risk Sharing likely to join, these results are not significant. Other economic attributes, such as household cattle ownership and different income quartiles were not significant in the demand for health insurance. Households in the lowest and lower income quartiles were as likely to enroll as those in the fourth income quartile. The findings from the first model respond to the question, Who enrolls in CBHI? Households living in Byumba and Kabgayi, who number five and more individuals, whose household head attended school, who live in the vicinity of a health center, and who own a radio appear to be more likely to buy insurance. Other important factors also influence households' probability of enrolling, such as their risk aversion and their exposure to effective information campaigns on prepayment schemes, as well as their trust in the scheme management, which is related to households' willingness to see CBHI as an investment and which supports the argument that enrollment in health insurance is not neces- sarily driven by economic conditions such as household income. The following reasons were identified in different surveys (focus group, household, and patient exit interview) to be important in households' enrollment decisions: · Both Byumba and Kabgayi had intensive awareness and information campaigns on PPS during the first year, supported by the district authorities and prepay- ment federation, which resulted in steady monthly enrollment increases. · The prepayment schemes' features, including benefit package, premium level, enrollment categories, copayments, and waiting period, were designed, dis- cussed, and agreed upon (by voting) in a series of about 30 workshops in the three districts. These workshops were attended by the local populations. As a result, the health insurance schemes were "tailored" as desired by, and in response to, the needs of the local people. · The main determinant of PPS participation is trust, which might be captured by the time variable. People living near the facilities are more likely to enroll because they know the health center personnel, as well as the prepayment scheme management team, and have been exposed to regular information campaigns on prepayment. · The participatory approach and the democratic management of PPS lead to sentiments of "ownership" and increased trust among the poor, which are basic conditions for poor households to engage in any investment. · Households that do not have the 2,500 francs (about US$8) to pay the one- year PPS enrollment fee join a "tontine." Over a five-week period, each tontine household pays 500 francs per week to the "tontine-caisse" as an installment toward the 2,500 francs total fee. Households are enrolled as full members once they have contributed 2,500 francs. · Local initiatives (churches and members who attended the PPS general assem- blies) have helped to pay enrollment fees for indigents, widows, orphans, and poor high-risk patients such as HIV-positive individuals. Community-Based Health Insurance in Rwanda 265 This shows that poor households will enroll in well-designed health insurance schemes that improve access to health care. At the same time, these solidarity groups contribute to positive social capital in a society that is recovering from a civil war. Therefore, community-based health insurance becomes a form of social cohesion and provides a link between the poor and the health facilities. The following section answers the second question, Does health insurance membership improve financial accessibility to care without increasing members' financial burden? Model 2: Equity in Financial Accessibility to Professional Care6 Evaluation of the first year of prepayment schemes in Rwanda has been based on extensive data collection. The analysis of monthly health service utilization data in health centers and hospitals has revealed that the overall use of curative ser- vices for adults and children, and preventive health services for children and women, was up to five times higher for PPS members than for nonmembers (Schneider and others 2001).7 Results in table 7.9 reveal that the insured report considerably better access to the modern health care system with a visit probability of 0.45, compared with a 0.15 visit probability for the uninsured in the pilot districts. Given the relatively moderate visit probability for sick PPS members of 0.45, it can be assumed that prepayment enrollment was not driven by adverse selection, and there was no frivolous use of health care caused by members' moral hazard behavior. Probability of visit by members does not vary by patients' gender, age, and income quartile; it is determined by patients' geographic access to the health facility (time distance) and health status, with the sick and very sick being three times more likely to seek care than healthier individuals. Interestingly, CBHI members who said they were not very sick or were sick reported higher visit like- lihood (0.22 and 0.64, respectively) than the average nonmember (0.05 and 0.15), indicating that CBHI membership causes sick individuals to seek care at the onset of illness. Access to professional care is lowest for the uninsured in the lowest income quartile, who are about four times less likely to seek care than the insured in the same income group. The logit regression results presented in table 7.10 estimate the probability of a professional health care visit for members and nonmembers and controls for skewness in the data distribution that could have influenced the access. The logit regression coefficient estimates were translated into odds ratios to facilitate interpretation. Findings show that health insurance has tremendously improved the finan- cial accessibility of its members to the modern health care system, particularly for women, children, and the poor. Access to care is determined by prepayment membership, patient age, pregnancy, patients' health status, distance to the health facility, and households' income group. Most important, prepayment members are 559 percent more likely than nonmembers to enter the modern 266 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 7.9 Probability of Using a Professional Provider by Insurance Status Pilot districts Probability of visit Members (N 376) Nonmembers (N 3,459) Sick individuals (N 3,835) 0.45a 0.15 Patient gender Female 0.42 0.14 Male 0.50 0.16 Patient age 6 years and older 0.45 0.13 0­5 years 0.46 0.19 Time from household to health facility More than 30 minutes 0.33 0.12 Less than 30 minutes 0.60 0.19 Income (expenditure) groups Quartile 1 0.40 0.06 Quartile 2 0.35 0.13 Quartile 3 0.49 0.14 Quartile 4 0.54 0.26 Self-perceived health status Not very sick 0.22 0.05 Sick 0.64 0.15 Very sick 0.61 0.30 Note: Probability of sick individuals with at least one visit with a professional provider during the two weeks prior to the interview. T-tests were performed to compare the rates of the insured with the uninsured in the pilot districts. a. Significant at 1 percent level of significance. health care system when sick. Health-related indicators significantly influenced health-seeking behavior, with children under five years of age being 92 percent more likely to report a visit than older patients. In addition, pregnant women report 65 percent higher probability than nonpregnant women for seeking care, and the sick individuals who spent four or more days in bed were 96 percent more likely to go to a modern health care provider than people who were not sick in bed as long. People who live close to the health facility are significantly more likely to seek care (61 percent) than those who live farther away. Patients in the lowest income quartile are far less likely to seek care than those in the highest income quartile. This means that while the prepayment scheme has sig- nificantly increased access to health care for members, including those who are poor, the impact at the district level in increased access to health care for the poor remains an issue. The solution is to find mechanisms to increase enroll- ment of the poor households in the prepayment schemes. Community-Based Health Insurance in Rwanda 267 TABLE 7.10 Logit Regression Results for Probability of at Least One Professional Provider Visit for Members and Nonmembers Probability of visit Explanatory variable Reference category variable Odds ratio S.E. Sign Prepayment members Nonmembers 6.59a 0.263 0.000 Male patient Female patient 1.21 0.140 0.170 Patient age 0­5 years Patient age 6 years and older 1.92a 0.158 0.000 Pregnant in past year No pregnancy in past year 1.65a 0.248 0.043 Patient spent 4 or more days in bed Less than 4 days in bed 1.96a 0.139 0.000 Less than 30 minutes to health facility More than 30 minutes 1.61a 0.137 0.000 to health facility HH with 5 or more members HH with fewer than 1.17 0.142 0.277 5 members HH head attended school HH head illiterate 0.91 0.141 0.519 HH with cattle No cattle 1.26 0.162 0.162 HH with radio No radio 1.33 0.143 0.050 First income quartile Fourth income quartile 0.18a 0.230 0.000 Second income quartile 0.44a 0.174 0.000 Third income quartile 0.46a 0.172 0.000 Ancillary statistics N 1,502 ­ 2 Log likelihood 1,434.941 Goodness fit (chi-squared 211.744 test) Degree of freedom 13 R Square 19.3% Note: Z-tests were performed to test the probability of enrollment for each characteristic in a logit model. a. Significant at 1 percent level of significance. Model 3: Average Out-of-Pocket Health Expenditures per Episode of Illness8 Based on detailed provider and insurance data, the evaluation of the prepay- ment pilot phase has shown that health insurance has a substantial impact on members' financial contribution to health care as well as on the providers' cost and financial situation (Schneider and others 2001). Findings in PHR report no. 61 have shown that the PPS members' annual per capita contributions to the modern health care system are up to five times higher than those of nonmem- bers (see Schneider and others 2001, table 3.14).9 However, the out-of-pocket expenditures per episode of illness are significantly lower for members as com- pared with those for nonmembers. This means that by paying insurance, scheme 268 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 7.11 Average Health Expenditure (RWFa) per Sick Individual with or without a Visit to a Professional Provider, by Health Insurance Status and Expenditure Quartile, in Pilot Districts Out-of-pocket Income spending RWF quartile Pilot districts Nonmembers Members No prof. visit 1+ prof. visit Total No prof. visit 1+ prof. visit Total Home and other care 1 90 85 90 276 0 165 2 124 178 131 68 23 53 3 171 230 180 28 20 24 4 277 322 288 79 133 108 Total 160 245 172 93 52 74 First professional 1 11 693 49 10 112 51 provider 2 10 1,356 180 20 178 75 3 43 1,445 246 9 220 112 4 36 2,228 600 105 966 572 Total 24 1,693 269 32 418 207 Other professional 1 0 262 15 0 0 0 providers 2 0 27 3 0 0 0 3 1 42 7 0 1 1 4 9 22 13 0 91 50 Total 2 50 9 0 28 12 Total illness-related 1 101 1,041 154 286 112 216 expenditure 2 134 1,561 314 88 201 128 3 215 1,717 433 37 242 137 4 322 2,573 901 184 1,190 730 Total 186 1,987 450 126 497 294 Source: National Bank of Rwanda. a. Nominal Exchange Rate: US$1 = RWF 370 (official period average in 2000). members face lower fees at the time of illness and have greater access at times of need. Furthermore, in Rwanda, with very low per capita use rates, the higher uti- lization by members should not be interpreted as an effect of moral hazard but rather as improved access to essential basic health services. The third model in this chapter documents insured and uninsured patients' out-of-pocket health expenditures per episode of illness. This information is first shown for all sick, independent of their care-seeking behavior, and second for those who reported at least one health facility visit. Table 7.11 presents total health-related out-of-pocket expenditures paid for each of the different health-related services during an episode of illness, which includes care received before visiting a provider, out-of-pocket spending at the first professional visit, and out-of-pocket spending for other professional providers. This total health expenditure information is shown for the insured and uninsured sick in the pilot districts and is further broken down for each group by "with and without a professional provider visit." Within these cate- Community-Based Health Insurance in Rwanda 269 gories, the different health expenditures are shown by patients' income quar- tiles. It is found that per episode of illness, sick members pay on average 294 francs for all health-related expenditures. This amount is higher for nonmem- bers in the pilot districts (450 francs). The differences are even stronger when comparing only the sick insured and uninsured with a professional care visit. Sick members with a professional visit pay 497 francs for the full episode of ill- ness, whereas nonmembers' out-of-pocket health expenditures per episode of ill- ness with a professional visit amount to 1,987 francs in the pilot districts. Insurance membership has significantly decreased out-of-pocket spending for a full episode of illness for sick members with and without a visit and at the same time has substantially improved members' access to the modern health care system. In addition, health insurance has changed patients' health care­seeking behavior. A comparison of members' and nonmembers' average out-of-pocket spending for home and other care in table 7.11 shows that the uninsured spend almost two and a half times more on home care and traditional remedies than the insured, who are more likely to seek quality care in the mod- ern health system. Thus not only have prepayment schemes reduced financial barriers in accessing better quality care and thus equity in accessing care, but insurance membership has shifted the demand for health care toward more effi- cient care as well. Members who seek care pay a 100 francs copayment per episode of illness at the health center. Sick individuals from richer households spend more on home care and on professional care compared with lower income groups, and this holds for the insured and the uninsured. The fact that the richer insured pay up to 10 times more per episode of illness than the poorest CBHI members supports the assumption made in the previous model. That is, prepayment schemes favor cross-subsidies from richer to poorer members if there is a uniform premium per family, if poorer families number more individuals than richer families, and if members' care-seeking behavior is independent of their income status. Thus far this argument holds. However, the richer insured may still pay more because they are willing to pay additional amounts for care not covered by the insurance scheme, such as drugs excluded from the MOH essential drug list. The following log-linear regression (table 7.12) estimates sick individuals' average health expenditures for the insured and uninsured in the pilot districts. Findings show that prepayment schemes have significantly decreased out-of- pocket spending for the entire episode of illness for sick individuals who are members. Individuals' out-of-pocket health expenditures are positively influ- enced by the patient's gender, with men paying more than women, by house- hold size, and by use of professional care. Patients classified in the three lower income quartiles report significantly lower out-of-pocket spending for an episode of illness, with or without a visit, than those in the fourth income quar- tile. In addition, out-of-pocket spending per episode of illness is significantly influenced negatively if patients live in the health center's vicinity and if they own cattle (which can be interpreted as a sign of wealth). 270 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 7.12 Log-Linear Regression Results: Estimated Total Health-Related Expenditures per Episode of Illness for Sick Individuals with and without a Visit Out-of-pocket all sick Explanatory variable Reference category variable Coeff. S.E. P > t Sick insured members Sick uninsured in pilot districts ­0.604a 0.141 0.000 Male patient Female patient 0.056a 0.061 0.000 Patient age 0­5 years Patient age 6 years and older ­0.006 0.075 0.362 Pregnant in past year No pregnancy in past year ­0.227 0.120 0.933 Patient spent 4 or more Less than 4 days in bed 0.228 0.060 0.057 days in bed Less than 30 minutes to More than 30 minutes to health facility ­0.125a 0.062 0.000 health facility HH with 5 or more HH with fewer than 5 0.111a 0.063 0.045 members members HH head attended HH head illiterate 0.262 0.061 0.075 school HH with cattle No cattle ­0.090a 0.075 0.000 HH with radio No radio 0.258 0.067 0.234 First income quartile Fourth income quartile ­0.544a 0.091 0.000 Second income quartile ­0.290a 0.085 0.000 Third income quartile ­0.183a 0.085 0.001 All sick w/ 1+ professional All sick without visit 1.645a 0.077 0.030 care visit (Constant) 1.048a 0.101 0.000 Ancillary statistics N 1,596 F 52.686 Degree of freedom (141,582) Prob > F 0.000 R square 0.318 Note: Includes total health related out-of-pocket spending for sick with and without visit. T-tests were performed to test significant difference for each characteristic. a. Significant at 1 percent level of significance. Table 7.13 presents log-linear regression results on out-of-pocket spending per episode of illness for the 336 patients who reported a professional health care visit. As can be expected, CBHI membership has significantly decreased health care costs for the sick members with a visit. Patients classified in the first income quartile report significantly lower out-of-pocket spending per episode of illness if they had a visit than do patients in the fourth income quartile. However, those in the second and third quartiles report lower out-of-pocket spending per Community-Based Health Insurance in Rwanda 271 TABLE 7.13 Log-Linear Regression Results: Estimated Out-of-Pocket Health Expenditures per Episode of Illness for Sick Individuals Who Reported a Professional Visit in Past 2 Weeks Log of total health expenditures with a visit Out-of-pocket with a visit Explanatory variable Reference category variable Coeff. S.E. P > t Sick insured members Sick uninsured in pilot districts ­0.808a 0.145 0.000 Male patient Female patient 0.274a 0.099 0.006 Patient age 0­5 years Patient age 6 years or older 0.068 0.107 0.523 Pregnant in past year No pregnancy in past year ­0.127 0.172 0.460 Patient spent 4 or more Less than 4 days in bed 0.088 0.101 0.386 days in bed Less than 30 minutes to More than 30 minutes to health facility ­0.121 0.098 0.215 health facility HH with 5 or HH with fewer than 5 members 0.061 0.103 0.556 more members HH head HH head illiterate 0.133 0.106 0.213 attended school HH with cattle No cattle ­0.068 0.112 0.547 HH with radio No radio 0.007 0.103 0.948 First income quartile Fourth income quartile ­0.464a 0.181 0.011 Second income quartile ­0.152 0.124 0.221 Third income quartile ­0.143 0.124 0.250 (Constant) 2.850a 0.155 0.000 Ancillary statistics N 336 F 5.286 Degree of freedom (13 323) Prob > F 0.000 R square 0.175 Note: Includes total health-related out-of-pocket spending for the full episode of illness for those who were sick and went to seek professional care. T-tests were performed to test significant difference for each characteristic. a. Significant at 1 percent level of significance. episode but are not statistically significant. Male patients with visits spend sig- nificantly more per episode of illness than female patients. Combining these findings with the results presented in the first and second model show that community-based health insurance in Rwanda has been suc- cessfully used as a tool to improve financial accessibility to care for the poor who enroll in the scheme while, at the same time, their out-of-pocket health care expenditures could be reduced per episode of illness. 272 Health Financing for Poor People: Resource Mobilization and Risk Sharing DISCUSSION AND CONCLUSION Findings confirm that prepayment schemes in Rwanda successfully target the rural poor, with their members having an annual average per capita income of approximately US$100. Logit regression results have shown that the probability of enrolling in prepayment schemes is equal among all income groups and is determined by factors such as households' distance from the health center, expo- sure to radio awareness campaigns, age, gender, and education status of the household head. Other factors have influenced enrollment, such as precaution- ary behavior for one's family's health, foresight, and the possibility of seeing the insurance premium as a trustworthy investment. Survey findings also allude to the fact that social capital such as trust and sentiments of ownership are impor- tant determinants of participation in prepayment schemes. Prepayment schemes are forms of solidarity and social cohesion that help strengthen the link between the poor and the health facilities where the local people would like to seek care. Health insurance has significantly improved equity in financial accessibility to care for members by increasing their probability of a visit while at the same time possibly reducing the financial burden per episode of illness. This argument holds for all income groups in the insurance pool, although richer members still pay up to 10 times more out-of-pocket, supporting the argument of possible cross-subsidies to poorer patients. Health insurance has helped to eliminate financial barriers in access to care for the poorest among the insured members, whereas the uninsured in the lowest expenditure quartiles continue to report significantly worse access to care than the richer insureds. In addition to improved access, faster access to care of the insured patients has contributed to a shift in demand for care from the traditional to the modern health sector and has improved the efficient use of limited medical resources such as drugs and staff in district health facilities. The analysis of the financial impact of prepayment schemes on health care providers has shown that community-managed health plans, combined with provider capitation payment, have built up expertise and capacity among insur- ance members about their rights and obligations and, as a result, have empow- ered consumers in discussions for better quality care with health center managers during the schemes' general assemblies. Although data collection during the pilot phase was extensive and included patient exit interviews and focus group information as well as routine provider insurance and household data, it is too early to conclude that better access to care due to prepayment membership has caused members' health to improve. However, findings from different sources suggest that conclusion. Still, about 90 percent of the population in the three districts has not enrolled with prepayment schemes and continues to report dismal health care utilization patterns. Although the large majority of nonmembers interviewed in the house- hold survey said they would like to become members, three-fourths of them had serious doubts that they would have the 2,500 francs available to pay the annual Community-Based Health Insurance in Rwanda 273 fee for their family, raising concerns about how successful the schemes are in improving equity of access to care. During the pilot year, church groups facili- tated enrollment by financing membership for widows, orphans, indigents, and poor HIV-infected individuals. This targeted demand-side subsidy contributed to a welfare gain if they benefited the indigents' insurance enrollment without decreasing benefits for patients who did not enroll. The church groups' experi- ence with financing health insurance membership has caused prepayment to become a promising tool to subsidize targeted access to care for the vulnerable. Acknowledgments: The authors are grateful to the World Health Organization (WHO) for having provided an opportunity to contribute to the work of the Commission on Macroeco- nomics and Health and to the World Bank for having published the material in this chapter as an HNP Discussion Paper. This analysis is supported by the Partnerships for Health Reform Pro- ject, which is administered by Abt Associates Inc. and funded by the United States Agency for International Development (USAID). Helpful comments on this analysis were received from Charlotte Leighton, A. K. Nandakumar, and Manjiri Bhawalkar. NOTES 1. Premium rates were set by taking into account existing user fees and by assuming that utilization rates would increase by 25 percent compared with baseline levels. See Schneider, Diop, and Bucyana (2000), chap. 3.1.1. 2. The Kabgayi PPS covers full episodes of caesarean sections, malaria, and nonsurgical pediatrics at the hospital, whereas the Kabutare and Byumba PPSs cover full episode C-sections, as well as each physician consultation and overnight stay at the district hospitals. 3. The DHS was conducted by the ONAPO in collaboration with Macro International and USAID in 2000­01. Households for the DHS were selected as primary sample units from sample cells identified for the Living Condition Monitoring Survey (LCMS), conducted by the Ministry of Finance in collaboration with UNDP in 2000­01. 4. The enrollment category "household" includes two adults and all children up to the age of 18 living in the same household. Other household members need to enroll in a group or individual category. 5. Households with up to seven members pay 2,500 francs per household per year. Indi- vidual enrollment costs 2,000 francs per year, and enrollment in a group for eight and more individuals costs 530 francs per person per year. Premiums are slightly higher in Kabgayi due to the larger hospital coverage (household--2,600 francs, individual-- 2,200 francs, and group enrollment per person--550 francs). 6. Professional care means public- and church-owned health centers, district hospitals, and dispensaries. This excludes sick individuals who sought care at traditional healers and others (for example, drug vendors, pharmacies). 7. District averages are 1.5 curative consultations per member per year in Kabutare and Kabgayi, and 1.1 curative consultations per member in Byumba, whereas nonmembers' curative care consultation level scores around 0.2 consultation per nonmember per year. 274 Health Financing for Poor People: Resource Mobilization and Risk Sharing 8. See note 6 above. 9. Members in Byumba reported 580 francs annual per capita contributions to health cen- ters, whereas this amount is only 104 francs per capita per year for nonmembers, due to their lower health service use. REFERENCES Diop, F., P. Schneider, and D. Butera. 2001. Summary of Results: Prepayment Schemes in the Rwandan Districts of Kabgayi, Byumba, and Kabutare. Technical Report No. 59. Partner- ships for Health Reform Project, Abt Associates Inc. Bethesda, Md. Manning, W., J. Newhouse, N. Duan, E. Keeler, A. Leibowitz, and M. Marquis. 1987. "Health Insurance and the Demand for Medical Care: Evidence from a Randomized Experiment." American Economic Review 77(3): 251­77. Muller, C. 1997. "Transient Seasonal and Chronic Poverty of Peasants: Evidence from Rwanda." WPS/97­8. Centre for the Study of African Economics, Institute of Econom- ics and Statistics, University of Oxford, Oxford. Schneider, P., F. Diop, and S. Bucyana. 2000. Development and Implementation of Prepayment Schemes in Rwanda. Technical Report No. 45. Partnerships for Health Reform Project, Abt Associates Inc. Bethesda, Md. Schneider, P., F. Diop, D. Maceira, and D. Butera. 2001. Utilization, Cost and Financing of District Health Services in Rwanda. Technical Report No. 61. Partnerships for Health Reform Project, Abt Associates Inc. Bethesda, Md. Yip, W., and P. Berman. 2001. "Targeted Health Insurance in a Low-Income Country and Its Impact on Access and Equity in Access: Egypt's School Health Insurance." Health Economics 10(2): 207­20. CHAPTER 8 The SEWA Medical Insurance Fund in India M. Kent Ranson Abstract: This chapter assesses the impact of the Self-Employed Women's Association's (SEWA's) Medical Insurance Fund, Gujarat, in terms of inclusion of the poor, hospital uti- lization, and expenditure. Age-matched insured and uninsured women were compared using survey data (2000). We found that wealth was not a determinant of membership in the Fund; that is, the poor were not excluded. Of 28 hospitalizations among Fund mem- bers over one year, only 5 were reimbursed. Membership in SEWA was not significantly associated with increased frequency of hospitalization, but there was a significant associ- ation with lower costs of hospitalization, net of reimbursement. Unlike many other CBHI schemes, the Fund has overcome barriers that exclude the poorest. This is due in part to nesting of the Fund within a larger development organization. Utilization of the Fund, and thus impact on hospital utilization and expenditure was minimal. This may relate to a lack of awareness of benefits among Fund members or costs and difficulties associated with submitting an insurance claim. C ommunity-based health insurance (CBHI) schemes--also referred to as micro- insurance units and mutual health insurance--are mechanisms wherein people prepay for some component of health care and there is some pooling of revenues and risks, with the healthy cross-subsidizing health care for the sick. Policymakers generally see CBHI as a means of improving access to effective health care, particularly among the poor, and preventing indebtedness and impoverishment as a result of trying to access such care (WHO 2000). A few studies have investigated the impact of CBHI schemes in developing countries. In general, these studies suggest that it is difficult to include the poor- est individuals and households in a CBHI scheme but that a well-designed and well-managed scheme can increase demand for, and utilization of, health care while protecting members from catastrophic costs. There are many reasons why the poorest in a population might not join a CBHI scheme, including lack of information about the scheme, lack of solidarity within the population, limited participation in planning or managing the scheme, unaffordable premiums, and priorities that are more important or imme- diate than health and medical insurance (for example, food and shelter). In their review of 83 health insurance schemes for the informal sector, Bennett, Creese, and Monasch (1998) found that most of them relied on flat-rate premiums and that for several schemes unaffordable premiums were a major deterrent to partici- pation. For example, a study conducted in rural Senegal found the average income 276 Health Financing for Poor People: Resource Mobilization and Risk Sharing of members of four CBHI schemes to be three times as high as that of randomly selected nonmembers; the authors attributed this difference to premiums that were unaffordable to the poorer part of the population (Jütting 2001). Despite problems of affordability, very few schemes have adopted sliding scales or exemptions for people who could not afford to pay (Bennett, Creese, and Monasch 1998). Numerous studies have found that CBHI schemes increase utilization while (or as a result of) decreasing costs to the consumer. Schemes that cover hospital inpa- tient care have resulted in increased rates of utilization in such diverse settings as China (Bogg and others 1996), the Democratic Republic of Congo (Criel and Kegels 1997), Ghana (Atim 1999), and Kenya (Musau 1999). In Bwamanda Dis- trict, Democratic Republic of Congo, Criel and Kegels found that rates of hospital utilization by members of a voluntary insurance scheme for hospital care were twice as high as those for the uninsured (49 versus 24.9 per thousand per year). The Nkoranza Community Financing Scheme in Ghana (Atim 1999) covers 100 percent of the costs of hospitalization. Members of the scheme were consistently more likely to be admitted to the hospital (4.6 to 6.3 percent admitted per year) than nonmembers (1.5 to 2.6 percent per year). We make the assumption in this study that increasing utilization rates among the poor in developing countries is a "good thing," at least from the perspective of scheme members. However, the inefficient overutilization of services (moral hazard) and the escalation of costs borne by the insurer or provider have been problematic, particularly in schemes that cover hospital inpatient care (Bennett, Creese, and Monasch 1998). The Integrated Social Security Scheme of the Self-Employed Women's Associa- tion (SEWA) was initiated in 1992. This scheme provides life insurance, medical insurance, and asset insurance. Those who pay the annual Social Security Scheme premium of 72.5 rupees--30 rupees of which is earmarked for the Medical Insur- ance Fund (herein referred to as the Fund)--are covered to a maximum of 1,200 rupees yearly in case of hospitalization in any registered (private or public) facility. Only women between the ages of 18 and 58 are eligible for membership in the Fund. Women also have the option of becoming lifetime members of the Social Security Scheme by making a fixed deposit of 700 rupees. Interest on this is used to pay the annual premium, and the deposit is returned to the woman when she turns 58. Upon discharge from the hospital, members must first pay for the hospitaliza- tion out of pocket. They submit receipts and doctors' certificates to the Fund, and if the insurance claim is approved, they are reimbursed by check. Excluded from cov- erage under the Fund are certain chronic diseases (for example, chronic tuberculo- sis, certain cancers, diabetes, hypertension, piles) and "disease caused by addiction" (SEWA brochures 2000). Throughout the 10 districts of Gujarat where it operates, the Fund had approximately 23,000 members in 1999­2000. This compares with roughly 150,000 women covered under the broader SEWA trade union statewide. The purpose of this chapter is to assess the impact of the Fund. The data for this analysis were collected from households in the Anand and Kheda Districts using an interview-administered questionnaire. We will look at impact in terms of (a) population reach of the Fund, particularly inclusion of the poor; (b) hospi- The SEWA Medical Insurance Fund in India 277 tal utilization during the one-year period preceding the survey; and (c) annual cost of hospitalizations, conditional on reporting one or more hospitalizations. We hypothesize that the Fund will do the following. First, it will include the very poor. The broader SEWA trade union organizes poor women working in the informal sector and seems to target quite effectively. By restricting membership in the Social Security Scheme (and hence the Fund) to members of the SEWA trade union, the Fund is likely to include women who are, on average, poorer than the general population. Second, it will increase the frequency of hospitalization among the insured by removing some component (that is, the maximum 1,200 rupees) of the financial barrier to seeking inpatient care. Note that the impact on utilization is likely to be lessened by the fact that women must first pay out of pocket--which often means borrowing money, selling valuables, or performing extra work--before seeking reimbursement from SEWA. Third, it will decrease the total annual hospital costs per person hospitalized. This basically assumes that (a) among insured women, some hospitalizations will be caused by conditions that are covered by the Fund, (b) women will actually seek reimbursement when they have been hospitalized for a covered condition, and (c) the Fund will reimburse women for some portion of the claims that are submitted. METHODS Data Collection and Analysis This was a cross-sectional cohort study; respondents were interviewed at only one point in time, and we fixed in advance the number of SEWA and uninsured house- holds (the two "cohorts"). Two-stage, random cluster sampling was used. The pri- mary sampling units (PSUs) were villages. Twenty villages were selected randomly (using random-number tables); the probability of selection was equal for all villages regardless of size. The secondary sampling units were households. Within each vil- lage, the insured were randomly selected from lists compiled by SEWA, and the uninsured were randomly selected from census or voting lists. In 10 villages, 14 SEWA households and 14 uninsured households were sampled, and in 10 villages, 14 SEWA households and 28 uninsured households were sampled (20 villages × 14 SEWA households = 280 SEWA households; 10 villages × 14 controls + 10 villages × 28 controls = 420 controls; therefore, 700 households are included in this analy- sis).1 The household questionnaire was administered between February 14, 2000, and May 6, 2000. An attempt was made to interview the female head of household. Data were double entered into a Microsoft Access Database. Analysis was con- ducted using Stata. Special statistical tools (the "svy function" in Stata) have been used to correct for clustering and stratification. This means that all measures of central tendency, association, and variance have been weighted or adjusted to account for the different probability of a household's being selected in each of the primary sampling units. 278 Health Financing for Poor People: Resource Mobilization and Risk Sharing The survey estimators are based on maximum likelihood estimation. We pre- sent adjusted Wald tests for all models; this is equivalent to the F-test of the sig- nificance of the regression. A p-value of 5 percent or less was used as criterion for significant association. The analyses in this chapter are restricted to women of ages 18 to 58 years, as only they are eligible for participation in SEWA's Medical Insurance Fund. Models What was the population reach of the Fund? The model for looking at socio- demographic determinants of membership is a logit model, written as follows: (8.1) ln(p/(1 ­ p)) = X + where p is the probability of being a member in the Fund, given female gender and age 18 to 58 years, and X represents a set of independent variables that are hypothesized to affect membership in community-based schemes. Did the Fund affect hospital utilization over the past year? The model is a logit model. It estimates the probability of an individual's being hospitalized during the one-year period preceding the interview. It can be written as follows: (8.2) ln(p/(1 ­ p)) = X + where p is the probability of hospitalization, given female gender and age 18 to 58 years, and X represents a set of independent variables that are hypothesized to affect individual patterns of hospital utilization. Did the Fund affect net annual hospital costs per person hospitalized? The model is a log-linear model that estimates the net costs incurred for all hospital- izations (over one year), conditioned on positive hospitalization. Costs were net of reimbursement by insurance schemes, including the Fund. The model can be written as follows: (8.3) lnY = X + where Y is the net annual hospital costs per person, given female gender, age 18 to 58 years, and one or more hospitalizations over one year, and X represents a set of independent variables hypothesized to affect individual patterns of hospi- tal expenditure.2 Equations 8.2 and 8.3 are equivalent to the "two-part" (utilization and expen- diture) model developed as part of the Rand Health Insurance Experiment and used more recently by Yip and Berman (2001) in their study of the impact of Egypt's School Health Insurance Programme. Dependent Variables Table 8.1 describes the independent variables included in the analyses. We include a number of household-level demand-side factors. Independent of insurance, TABLE 8.1 Independent Variables Included in the Regression Analyses Variables Model 1 2 3 Characteristics of the household ESI1 to ESI5 = quintiles of economic status index, this is an approximation of HH wealth based on assets, ESI1 being the poorest and ESI5 the wealthiest. (These variables are exhaustive, ESI1 is left out of the models.) HINDU = 1 if Hindu religion, 0 if Muslim or Christian BKWDCASTE = 1 if scheduled caste, scheduled tribe, and other "backward caste," 0 if castes that have not been identified by government as "backward" (Bhakshipanch, Brahmin, Patel, Shah, etc.). HHSIZE1 = 1 if 1 to 2 people in HH HHSIZE2 = 1 if 3 to 4 people in HH HHSIZE3 = 1 if 5 to 9 people in HH HHSIZE4 = 1 if > 10 people in HH (These variables are exhaustive, HHSIZE1 is left out of the models.) Characteristics of the individual NON-INS = 1 if not insured by SEWA and not living with someone who is insured by SEWA SEWA-INS = 1 if covered by SEWA's Social Security Scheme SEWA-FAM = 1 if uninsured but living in the same household as someone insured by SEWA (These variables are exhaustive, NON-INS is left out of the models.) AGE1 = 1 if 18 to 29 years of age AGE2 = 1 if 30 to 39 years of age AGE3 = 1 if 40 years of age or older (These variables are exhaustive, AGE1 is left out of the models.) LITERATE = 1 if person can read and write a simple letter, 0 if not MARRIED = 1 if married, 0 if never married, widowed, divorced, separated, or other DAILYWAGE = 1 if unskilled worker being paid daily wage (agricultural or factory worker) DOMESTIC = 1 if primary occupation is domestic work or housework OTHERWORK = 1 if other than unskilled daily wages or domestic work (These variables are exhaustive, OTHERWORK is left out of the models.) NUMBACUTE = number of acute illness episodes reported during the past 30 days (ranged from 0 to 3), intended to control for general level of health. We include this variable as a proxy, based on the hypothesis that those who are more sickly will have experienced illness episodes within the past month. This variable was included only in model 1 as it was collinear with SEWA-INS, the independent variable of interest in models 2 and 3. Characteristics of the hospitalization PUBLIC = 1 if government or ESIS hospital PRIVATE = 1 if private for-profit hospital NONPROF = 1 if "trust" or charitable hospital (These variables are exhaustive, PUBLIC is left out of the models.) SHORT = 1 if 0 to 3 days hospitalized MEDIUM = 1 if 4 to 7 days hospitalized LONG = 1 if 8 days or more hospitalized (These variables are exhaustive, SHORT is left out of the models.) OB/GYN = 1 if cause of hospitalization was pregnancy, delivery, or family planning, 0 if other 279 280 Health Financing for Poor People: Resource Mobilization and Risk Sharing wealth is hypothesized to be positively associated with rates of hospitalization and with net costs of hospitalization. As a proxy for wealth, we construct an economic status index (ESI) based on household assets, allowing the weights of these assets to be determined by the statistical procedure of principal compo- nents (Filmer and Pritchett 2001). The other household-level variables con- trolled for are religion, caste, and number of people living in the household. A number of individual-level, demand-side variables are controlled for. In all models, we control for age, literacy, marital status, and primary occupation. For models 2 and 3, individuals are classified as SEWA insured, uninsured but living in a household with at least one other insured person, and uninsured and not liv- ing with someone who is insured by SEWA. It was not uncommon for some adult women in a household to join the scheme while others abstained. We hypothe- size that uninsured women living in the same households may also have increased rates of utilization, due to the information and education provided by SEWA and the positive wealth effect of having insured people in the household. In model 1, we control for the number of acute illness episodes reported during the past 30 days as a proxy for general level of health (unfortunately, we did not collect information on whether or not individuals had chronic diseases). We hypothesize that those in poorer health are more likely to join the Fund. In model 3 only, we control for characteristics of the hospitalization. We hypothesize that use of private for-profit and private nonprofit hospitals (gener- ally perceived to be of higher quality) will be associated with higher net costs of hospitalization than use of government facilities. Women who report longer episodes of hospitalization are expected to have experienced higher net costs. Finally, we anticipate that women hospitalized for pregnancy, delivery, and fam- ily planning will generally have experienced an uncomplicated hospitalization without major surgical procedures and for this reason will have lower net costs. RESULTS In total, 242 SEWA households and 381 control households were included in the analyses (some households were dropped from the analyses due to misclassifica- tion). In the 242 SEWA households, there were 270 members and 125 women ages 18 to 58 who were nonmembers. In the 381 control households, there were 607 women ages 18 to 58. The demographic data (before controlling for any potential confounders) sug- gest that the SEWA-insured were of lower socioeconomic status than the unin- sured in control households (see table 8.2). They ranked lower on the ESI. They were almost twice as likely as the uninsured to be of a "backward caste" and tended to be from smaller households. They were older (mean 40.1 versus 35.0 years), less likely to be literate, more likely to report primary occupation as unskilled labor for daily wages, and almost 60 percent more likely to have reported illness within the past 30 days (our proxy for frequency of chronic disease). The SEWA Medical Insurance Fund in India 281 TABLE 8.2 Sample Characteristics Variable SEWA-INS SEWA-FAM NON-INS Number of households 242 381 Number of individuals 270 125 607 Mean ESI 0.29 0.84 Cat: Quintiles of ESI % in 1st quintile 13.1 15.8 % in 2nd quintile 23.9 13.8 % in 3rd quintile 30.1 22.0 % in 4th quintile 13.8 18.9 % in 5th quintile 19.1 29.5 Religion % Hindu 80.7 79.3 % Muslim 7.4 18.5 % Christian 12.0 2.1 % ST, SC, or other "backward" caste 54.0 29.4 Mean number of household members 5.8 7.0 Cat: Number of HH members % 1­2 6.1 3.5 % 3­4 30.4 20.8 % 5­9 53.0 52.6 % >10 10.5 23.1 Mean age 40.1 26.3 35.0 Cat: Age % 10­<20 0.1 12.1 6.3 % 20­<30 15.9 61.4 34.0 % 30­<40 29.7 18.2 21.2 % 40+ 54.3 8.3 38.4 % Literate 41.5 55.5 51.6 % Married 79.3 74.5 79.8 % Working for daily wages 25.0 11.4 16.9 % Doing domestic work 56.7 70.2 71.5 In the SEWA households interviewed, more than two-thirds of women ages 18 to 58 were enrolled in the Fund (see table 8.2). The SEWA-insured, in compari- son to the uninsured living with SEWA members, were older (mean 40.1 versus 26.3 years), less likely to be literate, more than twice as likely to report that pri- mary occupation was unskilled labor for daily wages, and more than four times as likely to have reported illness within the past 30 days. 282 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 8.3 Hospital Utilization and Expenditure per Hospitalization by SEWA Coverage SEWA-INS SEWA-FAM NON-INS (n = 270) (n = 125) (n = 607) Hospital utilization Total hospitalizations reported 28 12 56 Women with > 0 hospitalizations, 1 year 26 12 51 Probability of hospitalization 0.095 NS 0.105 NS 0.063 Hospital costs Total hospitalizations reimbursed 5 0 0 Women with > 0 reimbursement 5 0 0 Mean total hospital costs, 1 year 2,425 NS 3,532 NS 4,977 US$1 is approximately equal to 44 Rs. Note: T-tests were performed to compare rates and expenditures of the SEWA-INS with NON-INS, and the SEWA-FAM with NON-INS. Among the SEWA-insured who experienced hospitalizations during the one- year recall period, the frequency of reimbursement by SEWA was low (see table 8.3). There were 28 hospitalizations among Fund members; only 5 of these hos- pitalizations (18 percent) were reimbursed by SEWA. For the five members who were reimbursed, the costs before reimbursement were 4,431 rupees and after reimbursement 3,434 rupees. Before controlling for sociodemographic variables, SEWA-insured were 1.5 times more likely than uninsured women in control households to have been hospitalized (0.095 versus 0.063). Among those hospitalized, the net annual hos- pital costs for the SEWA-insured were less than half those for the uninsured women in control households. (Clearly, this difference cannot be attributed to the Fund, given that even among the very few people reimbursed, the mean reimbursement amounted to less than one-quarter.) Regression Analyses Controlling for other sociodemographic variables, only older age and higher fre- quency of illness episodes within the past month were significantly associated with membership in the Fund (see table 8.4). Results were the same for the "full" and "best fit" models. Wealth, proxied by quintiles of ESI, was not significantly associated with membership in the Fund; there was a trend suggestive of higher levels of membership among the second and third income quintiles (compared with the first, or poorest, quintile), but this did not reach significance at the 95 percent level. Women of ages 30 years and above were 3.4 times as likely to join the Fund as those ages 18 to 20 (full model). Each additional illness reported within the past month (acute illnesses as well as exacerbations of chronic dis- ease) was associated with a 70 percent (full model) to 80 percent (best fit) increase in the probability of joining the Fund. The SEWA Medical Insurance Fund in India 283 TABLE 8.4 Regression Results for Equation 1, the Odds of Being SEWA-INS Based on Sociodemographic Variables: Logit Model (N = 987) Odds ratios (t-statistics) Odds ratios (t-statistics) Full model Best fit model ESI2 1.906a ESI2 1.837a (2.060) (1.880) ESI3 1.922 ESI3 1.793 (1.280) (1.090) ESI4 0.961 ESI4 0.988 (­0.150) (­0.030) ESI5 1.300 ESI5 1.287 (0.700) (0.650) HINDU 0.720 HINDU -- (­0.550) -- BKWDCASTE 2.450a BKWDCASTE 2.563a (2.000) (1.910) HHSIZE2 0.821 HHSIZE (cont.) 0.940 (­0.320) (­1.220) HHSIZE3 0.631 (­1.120) HHSIZE4 0.454 (­1.100) AGE2 3.356c AGE (cont.) 1.040c (4.840) (6.220) AGE3 3.423c (4.930) LITERATE 1.166 LITERATE 1.166 (0.410) (0.470) MARRIED 0.970 MARRIED -- (­0.160) -- DAILYWAGE 0.672 DAILYWAGE 0.888 (­1.720) (­0.470) DOMESTIC 0.601 DOMESTIC 0.675 (­1.640) (­1.390) NUMBACUTE 1.695b NUMBACUTE 1.799c (2.690) (3.090) Adjusted Wald Test, F = 55 97 P-value = 0.003 0.000 Percent of predictions correct = 72.8% 72.9% a. 10% (borderline) significance level. b. 5% significance level. c. 1% significance level. 284 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 8.5 Regression Results for Equation 2, the Probability of Being Hospitalized within the Last Year: Logit Model (N = 987) Odds ratios (t-statistics) Odds ratios (t-statistics) Full model Best fit model SEWA-INS 2.042 1.668 (1.220) (0.960) SEWA-FAM 1.639 1.999 (1.000) (1.640) ESI2 0.564 0.522 (­1.370) (­1.540) ESI3 0.801 0.762 (­0.460) (­0.540) ESI4 0.798 0.690 (­0.540) (­0.890) ESI5 0.275a 0.253a (­1.960) (­1.940) HINDU 1.578 1.638 (1.100) (1.100) BKWDCASTE 0.785 -- (­0.700) -- HHSIZE2 2.328 2.729 (1.020) (1.200) HHSIZE3 1.687 2.307 (0.680) (1.010) HHSIZE4 0.846 1.233 (­0.190) (0.240) AGE2 0.450 -- (­1.270) -- AGE3 0.386a -- (­1.990) -- LITERATE 0.585 0.766 (­1.190) (­0.620) MARRIED 1.522 1.453 (0.670) (0.750) DAILYWAGE 0.733 0.719 (­0.470) (­0.520) DOMESTIC 1.254 1.499 (0.390) (0.690) Adjusted Wald Test, F = 36 21 P-value = 0.027 0.002 Percent of predictions correct = 91.6% 91.6% a. 10% (borderline) significance level. The SEWA Medical Insurance Fund in India 285 Neither membership in the Fund nor any of the sociodemographic variables tested were significantly associated with the probability of having been hospital- ized (see table 8.5). Again, results were similar for the full and best fit models. There was a trend suggestive of higher rates of hospitalization among Fund mem- bers (and even women living in the same households as Fund members), but this association was not significant. There were also trends toward lower frequency of hospitalization among higher ESI quintiles and lower frequency of hospitalization with increasing age, but again these were not significant at the 95 percent level. Results of the model of annual hospital costs per person hospitalized varied somewhat with changes in the variables included and the removal of outliers (see table 8.6). In some models, hospital expenditures were significantly lower among the SEWA insured (models 3B, 3C, and 3D). Interestingly, this finding was not sensitive to removal from the model of the five cases of hospitalization that were reimbursed; in model 3D, being insured by SEWA was associated with a decrease in hospital expenditures of 54 percent ( = ­0.789) even though the five reimbursed hospitalizations were removed from the calculations. Consistent in the various iterations of model 3 were the findings that hospital expenditures varied directly (and significantly) with quintiles of ESI, were significantly higher for private than for public hospitalizations, and were significantly lower for pregnancy, delivery, or family planning than for other causes. DISCUSSION Strengths and Limitations of the Study Perhaps the greatest limitation of this study was its small sample size. The study shows trends toward higher rates of utilization and lower spending per episode of hospitalization among SEWA members (significant in some models). Had the study been larger, these associations might have been statistically significant (or in the case of spending, consistently statistically significant). Insufficient sample size arose in part because there were fewer Fund members in "insured" house- holds than had been expected and because of the problem of misclassification of households--that is, households were identified as including a Fund member when in fact they did not. Such households were dropped from the analysis without replacement. It is difficult to say how accurately the economic status index reflected house- hold "wealth." A very similar index developed for Indian survey data (Filmer and Pritchett 2001) was closely correlated with state domestic product (SDP) and poverty rates data. Using data from Indonesia, Nepal, and Pakistan, they also showed their asset index to be consistent with consumption expenditures. Com- parison of our asset index with the interviewers' assessments of wealth and with daily household expenditures on food suggested strong correlation (data not presented here). Nonetheless, it is possible that some of the "negative results" in this study were due to insufficiently controlling for wealth. For example, if, as 286 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 8.6 Regression Results for Equation 3 Odds ratios (t-statistics) 3A full model 3B best fit 3C outliers removed 3D reimbursed removed N = 77 N = 81 N = 74 N = 70 SEWA-INS ­0.630 ­1.744b ­0.812b ­0.789b (­1.190) (­2.350) (­2.290) (­2.140) SEWA-FAM 0.560 0.289 0.000 ­0.048 (0.900) (0.300) (0.000) (­0.100) ESI2 1.662b 1.908c 1.356c 1.353c (2.270) (2.930) (4.000) (3.960) ESI3 1.779b 1.237c 0.550a 0.551 (2.800) (2.900) (1.740) (1.690) ESI4 2.171b 1.273c 1.235b 1.234b (2.810) (3.180) (2.640) (2.630) ESI5 2.443b 2.356b 1.266b 1.345b (2.7Z0) (2.580) (2.480) (2.620) HINDU 3.630c -- -- -- (3.300) -- -- -- BKWDCASTE 0.465 1.103b 0.270 0.268 (1.700) (2.120) (1.560) (1.450) HHSIZE2 ­1.301c -- -- -- (­3.440) -- -- -- HHSIZE3 ­1.610c -- -- -- (­3.500) -- -- -- HHSIZE4 ­0.581 -- -- -- (­1.490) -- -- -- AGE2 ­0.246 -- -- -- (­0.570) -- -- -- AGE3 0.403 -- -- -- (1.090) -- -- -- (continued) hypothesized, SEWA membership was inversely associated with wealth, and wealth was directly associated with hospital spending, then failure to fully con- trol for wealth could result in an observed estimate of effect diluted toward the null. Though not presented in this chapter, all of the models were run using two other indicators of wealth (interviewers' assessments and daily per capita food consumption) with no major changes in the results. Several questions were not included in the household questionnaire that, in retrospect, should have been included. For example, it is common in such analy- ses to control for the presence of "chronic diseases," but these data were not available from our questionnaire. It would have been both interesting and infor- mative to know, among the SEWA-insured women who had undergone hospital- ization, the number who had submitted claims but were still awaiting a response The SEWA Medical Insurance Fund in India 287 TABLE 8.6 Continued Odds ratios (t-statistics) 3A full model 3B best fit 3C outliers removed 3D reimbursed removed N = 77 N = 81 N = 74 N = 70 LITERATE 0.152 -- -- -- (0.420) -- -- -- MARRIED 2.861b 0.659 ­0.019 ­0.004 (2.620) (0.560) (­0.020) (0.000) DAILYWAGE 0.885 0.412 ­0.223 ­0.247 (0.950) (0.430) (­0.450) (­0.510) DOMESTIC 0.793 0.674 0.325 0.345 (0.980) (0.830) (0.510) (0.540) PRIVATE 4.306c 4.453c 2.600c 2.601c (10.710) (5.140) (8.220) (8.300) NONPROF 3.115a 3.058a 1.763b 1.737 (2.080) (2.080) (2.290) (2.070) MEDIUM (4 to 7 days) 0.206 1.092a 0.249 0.242 (0.420) (1.810) (0.740) (0.690) LONG (> 7 days) 0.716 1.433a 1.445c 1.453c (1.370) (1.750) (4.800) (4.640) OB/GYN ­1.379b ­1.357c ­1.312c ­1.334c (­2.430) (­4.170) (­4.190) (­3.940) Adjusted Wald Test, F = 95 47 8 7 P-value = 0.081 0.001 0.031 0.069 R-squared = 79.32% 66.19% 67.24% 67.76% a. 10% (borderline) significance level. b. 5% significance level. c. 1% significance level. or had been unsuccessful in their claims. Finally, for purposes of triangulation (that is, verifying the ESI), we could have collected data on household expendi- tures on a small number of items, as a proxy for total household expenditures (Morris and others 2000). It is possible, though unlikely, that observation bias affected the study results. Interviewer bias may have occurred if investigators elicited or interpreted infor- mation differently among the insured and the uninsured. It was impossible to blind interviewers to the insurance status of the household. Certainly, the inter- viewers came to make generalizations about households (for example, that SEWA households tended to be very poor). Thus there may have been some bias in the way they recorded household asset information. It is also possible that the interviewers probed more carefully into health care seeking and spending 288 Health Financing for Poor People: Resource Mobilization and Risk Sharing among poorer households. Study subjects may also have reported events in a noncomparable manner (recall bias). For example, SEWA members may be like- lier to recall episodes of hospitalization or to remember how much they have paid for hospitalization as they have been sensitized to the subject by the infor- mation, education, and communication from SEWA (or they have spent months collecting and processing the related paperwork). Perhaps the lower hospital expenditures reported by the SEWA insured is a function of more accurate recall (that is, a lower probability of accidentally inflating the figures). Unlike many other CBHI schemes, SEWA has not excluded the very poor. What design factors have facilitated inclusion of the poor in the SEWA scheme? The fact that the SEWA Integrated Social Security Scheme is nested within the larger development organization (the SEWA workers union) has undoubtedly been an important factor. Bennett, Creese, and Monasch (1998, p. 20) hypothe- size that "communities may be more willing to participate actively in health insurance schemes (initiated by NGOs involved in broad community develop- ment activities) since they consider that their priority needs--for a stable income, for instance--are also being addressed." Other factors that are likely to have facilitated inclusion of the poor include an affordable premium, village- level representatives who are themselves poor, self-employed women, and efforts to serve geographically isolated villages. The positive associations between older age and higher frequency of illness and membership in SEWA's insurance scheme suggest that adverse selection may be occurring. Bennett, Creese, and Monasch (1998), in their review of community- based health insurance schemes, found that adverse selection affected schemes covering hospital inpatient care, in particular. The fact that membership in the SEWA scheme is voluntary and individual may enable adverse selection. How- ever, the waiting period after joining and the exclusion of preexisting or chronic diseases are meant to limit adverse selection. It is likely that adverse selection is to some extent encouraged by scheme functionaries, insofar as poor households with limited expendable income may be encouraged to insure the household member most likely to fall ill. Furthermore, the scheme does fall somewhere on the spectrum between health insurer (strictly defined) and "social service" in that the scheme does aim to improve access to hospital care among the poor and to protect the poor from the costs of hospitalization. As such, adverse selection may be viewed in a positive light. If it is decided to try to deter adverse selection under the scheme, additional methods that could be used include (a) making the house- hold, or even the village, the unit of membership and enforcing this rule strictly; (b) stipulating that if a village is to be allowed to enter a scheme a certain propor- tion of households in the village must join; and (c) making the scheme compul- sory (Bennett, Creese, and Monasch 1998, p. 56). We found no significant association between membership in SEWA's Medical Insurance Fund and frequency of hospitalization, although there was a non- significant trend toward higher rates of hospitalization among SEWA members. Table 8.7 summarizes the results of other studies that have examined the impact The SEWA Medical Insurance Fund in India 289 TABLE 8.7 Summary of Studies Looking at the Impact of CBHI Schemes on Utilization and Out-of-Pocket Expenditures Study Description of study Utilization Expenditure Rwanda (Schneider and Fifty-four prepayment Increased probability (6.6 Decreased (by approx. 60%) Diop 2001) schemes in 3 districts times) of "at least one visit "total out-of-pocket payment covering some outpatient to professional health care per illness episode." and inpatient costs. What provider." does the scheme cover? Sample size of 11,583 (2,518 HH). Data collection, yr. 2000. Senegal (Jütting 2001) Three Mutual Health Increased "proportion of Decreased "out-of-pocket Insurance Schemes that sample with at least one spending on hospitalization" cover part of the costs of hospitalization." (48%). hospitalization. Sample size of 2,987 (346 HH). Data collection, yr. 2000. India (Gumber 2001) Self-Employed Women's Decreased likelihood (down No change in "total annual Association (SEWA); covers by 63%) of seeking cost (direct and indirect) of hospital costs only, to 1,200 ambulatory care in case of health care use." Neither rupees. Sample size 1,200 HH illness. (Perhaps women are ambulatory nor inpatient. in nonrandomly selected "jumping the queue"?) clusters. Data collection, yr. 1998­99. Democratic Republic District-level scheme that Rates of hospitalization -- of Congo (Criel and Kegels covers 80% of hospitalization were consistently higher 1997) costs at referral hospital. (1.5 to 2 times) among the Routinely collected insured. hospitalization data. Ghana (Atim 1999) Nkoranza community-financing Members are consistently -- scheme. Covers 100% of more likely (2.3 to 4 times) hospital costs for referred to be admitted to a hospital patients. Hospital data from than nonmembers. 1992 to 1994. China (Bogg and others 1996) Comprehensive health Evidence for increased -- insurance system in Jintan utilization of inpatient care. County. Partial reimbursement for drugs, outpatient and inpatient visits. Data from 1986 to 1994. Niger (Diop and others 1995) Boboye district. Mandatory The number of initial visits Total illness-related taxation. Coverage of (outpatient) increased by expenditure dropped by "pharmaceutical products." nearly 40% during the year 48%. Longitudinal data. following implementation. Note: HH = household. Source: Format adapted from Jakab and others (2001). 290 Health Financing for Poor People: Resource Mobilization and Risk Sharing of CBHI schemes on rates of hospitalization. Almost all other studies found that community-based insurance covering the costs of hospitalization increases hos- pital utilization.3 This may reflect a publication bias, wherein the most success- ful schemes are the ones most likely to have been studied and reported on. If we accept these findings as valid, then a question arises: Why has the SEWA scheme not resulted in significantly increased rates of utilization? The scheme's failure to affect hospital utilization is attributable to the factors that have prevented women from using the Fund (that is, the factors that have prevented women from submitting insurance claims). Data from qualitative interviews (Ranson 2001) suggested that members of the Fund are sometimes unaware of their mem- bership or the benefits of the scheme. Furthermore, among those who do know about the scheme, rates of reimbursement may be considered low by members (as the Fund does not cover, for example, transportation or bribes), while the costs of submitting a claim (for example, transportation to the SEWA office, opportunity cost of missed work, bribes paid to doctors for hospital certificates) are potentially quite high. There were significant (but not consistently so) associations between SEWA membership and lower costs of hospitalization. Very few other studies have looked at whether CBHI has actually resulted in decreased out-of-pocket expen- ditures (see table 8.7). In a small study of four mutuals in Senegal (carried out in 2000), Jütting (2001) found that "being a member reduces the expenditure for hospitalization by 48 percent in comparison to nonmembers holding all other variables constant." Other studies (Schneider and Diop 2001; Diop, Yazbeck, and Bitran 1995) have found decreased spending (both outpatient and inpatient) per illness episode. Interestingly, in our study the association between membership in SEWA's Fund and decreased hospital spending was not due to reimbursement by the scheme, as the associations remained even after we removed the five reimbursed hospitalizations from the sample. The difference then may result from methodological problems, such as failure to adequately control for wealth or other potential confounders (most important, severity of illness resulting in hospitalization). Alternatively, it may be that the scheme confers protection from hospital costs in some other way. For example, doctors may charge SEWA members less as the doctors know the scheme is restricted to the poor, self- employed (unlikely, given that providers are usually unaware of SEWA, far less a woman's membership status in the organization or insurance scheme). It may also be that SEWA members are more sensitive to the costs of hospitalization and as a result are more likely, or better able, to seek out low-cost hospital care. Conclusions and Policy Recommendations Members of the Fund were similar to the general population in terms of wealth. The Fund's success at including the poor was probably due to its being nested within a development organization committed to serving poor, self-employed women. Members of the Fund were older and sicker than the general popula- The SEWA Medical Insurance Fund in India 291 tion, suggestive either of adverse selection or effective targeting of those most in need of inpatient care. In either case, the Fund can facilitate risk pooling by broadening its membership to include younger and healthier individuals, by requiring, for example, enrollment of all eligible women in a household. Relatively few of those who were members of the Fund and were hospitalized were reimbursed through the scheme. This suggests that either women are not submitting claims even when they might be eligible for reimbursement or that the claims are not eligible for reimbursement (for example, if the hospitalization resulted from certain chronic illnesses). Given the low rate of utilization of the Fund by those who are members, it is not surprising that the Fund had no dis- cernable affect in terms of health care utilization. Rates of Fund utilization may be increased by providing members with information and education about their membership in the Fund and its benefits, and by making the process of claims submission easier, faster, and less expensive. Acknowledgments: The author of this report is grateful to the World Health Organization (WHO) for having provided an opportunity to contribute to the work of the Commission on Macroeconomics and Health and to the World Bank for having published the material in this chapter as an HNP Discussion Paper. This research was supported by the British Department for International Development (DfID). The author wishes to thank members of the Self-Employed Women's Association for their support and Anne Mills and Kara Hanson for stimulating review comments. The findings, interpretations, and conclusions expressed in the chapter are entirely those of the author and do not necessarily represent the views of the World Bank, its Executive Directors, or the coun- tries they represent. NOTES 1. Note that this study was actually designed to look at two community-based insurance schemes, SEWA and the Tribhuvandas Foundation (TF), and a total of 1,120 house- holds were actually interviewed. However, TF households and their controls are not included in this analysis. 2. In the log-linear model, the coefficient for a continuous independent variable gives the relative change in the mean value of Y for a unit change in X. To obtain the relative change in mean Y for a dummy variable, one must take the antilog (to base e) of the estimated dummy coefficient and subtract it from 1 (Gujarati 1995, p. 525). 3. Note that the other study of the SEWA scheme, by Gumber (2001), has not examined the impact of the scheme on hospital utilization REFERENCES Atim, C. 1999. "Social Movements and Health Insurance: A Critical Evaluation of Volun- tary, Nonprofit Insurance Schemes with Case Studies from Ghana and Cameroon." Social Science and Medicine 48(7): 881­96. 292 Health Financing for Poor People: Resource Mobilization and Risk Sharing Bennett, S., A. Creese, and R. Monasch. 1998. Health Insurance Schemes for People Outside Formal Sector Employment. Geneva: World Health Organization. Bogg, L., D. Hengjin, W. Keli, C. Wenwei, and V. Diwan. 1996. "The Cost of Coverage: Rural Health Insurance in China." Health Policy and Planning 11(3): 238­52. Criel, B., and G. Kegels. 1997. "A Health Insurance Scheme for Hospital Care in Bwa- manda District, Zaire: Lessons and Questions after 10 Years of Functioning." Tropical Medicine and International Health 2(7): 654­72. Diop, F., A. Yazbeck, and R. Bitran. 1995. "The Impact of Alternative Cost Recovery Schemes on Access and Equity in Niger." Health Policy and Planning 10(3): 223­40. Filmer, D., and L. H. Pritchett. 2001. "Estimating Wealth Effects Without Expenditure Data--or Tears: An Application to Educational Enrollments in States of India." Demog- raphy 38(1): 115­32. Gujarati, D. 1995. Basic Econometrics, 3d ed. London: McGraw-Hill. Gumber, A. 2001. Hedging the Health of the Poor: The Case for Community Financing in India. HNP Discussion Paper. Washington, D.C.: World Bank. Jakab, M., A. S. Preker, C. Krishnan, P. Schneider, F. Diop, J. Jütting, A. Gumber, K. Ranson, and S. Supakankunti. 2001. Social Inclusion and Financial Protection through Community Financing: Initial Results from Five Household Surveys. Report submitted to Working Group 3 of the Commission on Macroeconomics and Health, Jeffrey D. Sachs (Chair- man). WHO, Geneva. Jütting, J. 2001. The Impact of Community-Based Health Financing on Financial Protection: Case Study Senegal. HNP Discussion Paper. Washington, D.C.: World Bank. Morris, S., C. Carletto, J. Hoddinott, and L. Christiaensen. 2000. "Validity of Rapid Esti- mates of Household Wealth and Income for Health Surveys in Rural Africa." Journal of Epidemiology and Community Health (London) 54: 381­87. Musau, S. 1999. Community-Based Health Insurance: Experiences and Lessons Learned from East and Southern Africa. Technical Report No. 34. Partnerships for Health Reform Pro- ject, Abt Associates, Inc., Bethesda, Md. Ranson, M. K. 2001. "The Consequences of Health Insurance for the Informal Sector: Two Non-Governmental, Nonprofit Schemes in Gujarat." Ph.D. thesis in progress. London School of Hygiene and Tropical Medicine. Schneider P., and F. Diop. 2001. Synopsis of Results on the Impact of Community-Based Health Insurance (CBHI) on Financial Accessibility to Health Care in Rwanda. HNP Discussion Paper. Washington, D.C.: World Bank. WHO (World Health Organization). 2000. The World Health Report 2000: Health Systems: Improving Performance. Geneva. Yip, W, and P. Berman. 2001. "Targeted Health Insurance in a Low Income Country and Its Impact on Access and Equity in Access: Egypt's School Health Insurance." Health Eco- nomics 10(2): 207­20. CHAPTER 9 The Potential Role of Community Financing in India Anil Gumber Abstract: This chapter reviews the existing community-based and self-financing health insurance schemes in India that cater to the general population and address the needs of the society's poor and vulnerable section. It discusses critical issues of accessibility and use of health care services, out-of-pocket expenditure on treatment, and the need for health insurance for poor rural and urban households pursuing varied occupations. The chapter examines in detail the determinants of enrollment in the community-based financing scheme, using the household-level data from the pilot study undertaken in Gujarat. It also investigates the issue of how much health insurance mitigates the house- holds' burden of health care expenditure. The findings suggest that the community plan fairly addresses equity in enrollment but that, in terms of providing financial protection, social insurance coverage is much more successful. M ore than 90 percent of India's population and almost all its poor have no health insurance coverage. They primarily meet health care needs through direct out-of-pocket expenditure on services provided by the public and private sectors. Furthermore, various health care services studies show that the poor and other disadvantaged groups (scheduled castes and scheduled tribes) are forced to spend a higher proportion of their income on health care than better-off groups. For the disadvantaged, the burden of treat- ment, especially inpatient care, is disproportionately heavy (Visaria and Gumber 1994). In addition, this group suffers a high incidence of morbidity, which adversely affects their household budgets in two ways: in the large amounts of money and resources they have to spend on medical care, and in the earnings they have to forgo during periods of illness. Thus they often have to borrow funds at very high interest rates to meet both medical expenses and other house- hold consumption needs. One possible consequence of this is that these families could be pushed into a zone of permanent poverty. There are also concerns about problems in accessibility and the use of subsi- dized public health facilities. Most poor households, especially rural ones, reside in backward, hilly, and remote regions, where neither government facilities nor private medical practitioners are available. These households have to depend heavily on poor quality services provided by local, often unqualified, practitioners and faith healers. Wherever accessibility is not a constraint, the primary health centers are generally found to be either dysfunctional or providers of low-quality 294 Health Financing for Poor People: Resource Mobilization and Risk Sharing services. The government's claim to provide free secondary and tertiary care does not stand up; in reality, patients are charged for various services (Gumber 1997). Estimates based on a large-scale health care utilization survey of 1993 suggest that, overall, about 6 percent of household income is spent on curative care, which amounts to Rs. 250 per capita per year (Shariff and others 1999). However, the burden of expenditure on health care is unduly heavy on households in the informal sector, indicating the potential for voluntary comprehensive health insurance schemes for these segments of society. Overall, health insurance coverage is low. Only 9 percent of the Indian work- force is covered by some form of health insurance (through the Central Govern- ment Health Scheme [CGHS], the Employees' State Insurance Scheme [ESIS], and Mediclaim), and most of those insured belong to the organized sector (Gumber 1998). Health insurance coverage is so sparse because government policy has been to provide free health services through public hospitals, dispensaries, and clinics. As noted above, however, public sector providers charge patients for var- ious services, and outreach is also poor. According to estimates based on the National Sample Survey (NSS) 1986­87, nationally 42 percent of inpatients and 30 percent of outpatients using public sector facilities had paid for various ser- vices; the percentages varied substantially between rural and urban areas and among states (Gumber 1997). Furthermore, health care costs have increased enormously. A comparison of NSS data for 1986­87 and 1995­96 suggests that the cost of inpatient care and outpatient care grew annually at 26­31 percent and 15­16 percent, respectively, putting severe strains on efforts to achieve equity in health care (Gumber 2001). Nongovernmental organizations (NGOs) and charitable institutions (not-for- profit) have played an important role in the delivery of affordable health services to the poor, but their coverage has always been small. The issue is how to reach the unreached and how to ensure that the uninsured receive at least a minimum of affordable quality services. The public insurance companies so far have paid very little attention to volun- tary medical insurance because of low profitability, high risk, and lack of demand. From the consumer point of view, insurance coverage is low because consumers lack information about the private insurance plans, and the mechanisms used by the health insurance providers are not suitable for consumers. Furthermore, com- pared with ESIS and with the community-based schemes, the private plans offer a modicum of benefits (table 9.1), that is, only hospitalization and that with many exclusions (such as preexisting conditions). One analysis suggests that the exist- ing voluntary health insurance plans cover only between 55 and 67 percent of total hospitalization costs and, on average, only 10 to 20 percent of the total annual out-of-pocket expenditure on health care (Gumber 2000a). Gender bias in use of health care persists. For various socioeconomic and cul- tural reasons, men have better access than women. Poor women are most vul- nerable to diseases and ill health because they live in unhygienic conditions, carry a heavy childbearing burden, place little emphasis on their own health The Potential Role of Community Financing in India 295 TABLE 9.1 Type of Health Care Burden on Households Covered by Health Insurance Schemes Type of care or cost ESIS SEWA Mediclaim Inpatient Medical Transport and other direct cost Loss of earnings Outpatient Medical Transport and other direct cost Loss of earnings Preventive and promotive Immunization Prenatal and postnatal care Maternity care Family planning Note: Self-Employed Women's Association (SEWA) and Mediclaim are reimbursement plans (subject to the sum assured) while ESIS is a facility-based plan. care needs, and encounter severe constraints in seeking health care for them- selves. Thus far institutional arrangements to correct these gender differentials have been lacking. A pioneering study undertaken by Gumber and Kulkarni (2000) looked into issues related to the availability and needs of health insur- ance coverage for the poor, especially women, and the scope and likely problems in extending current health insurance benefits to workers in the informal sector. This chapter attempts to review existing community-based and self-financing health insurance schemes in India that serve the general population and address the needs of the poor and vulnerable. It discusses some critical issues concerning accessibility and use of health care services, out-of-pocket expenditure on health care, and the need for health insurance for poor rural and urban households pur- suing varied occupations. The chapter examines in detail the determinants of enrollment in the community-based financing scheme, using household-level data from the pilot study. It also investigates the issue of how much health insur- ance mitigates the households' burden of health care expenditure. COMMUNITY FINANCING IN INDIA AND THE SEWA PROGRAM Community and self-generated financing programs are usually run by NGOs or nonprofit organizations. These organizations rely on financing from various sources, including government, donor agencies, and community and self-generated sources. Among the many innovative methods being used to finance health care services are progressive premium scales, community-based prepayment and insurance schemes, and income-generating schemes. These organizations' target populations for health care services are primarily workers and families outside the formal sector. Program revenue comes from the following sources: 296 Health Financing for Poor People: Resource Mobilization and Risk Sharing · User fees, defined as the payment made by the beneficiaries directly to the health care provider (for example, fees for services or prices paid for drugs and immunizations). This mode of financing is not common. · Prepayment-insurance schemes, including payment by members for drugs at either subsidized rates or cost. · Commercial for-profit schemes actively run by health care finance organizations. · Fund-raising activities by organizations to pay for health care services. This type of revenue represents more than 5 percent of some organizations' total funding. · Contributions in-kind (for example, rice, sorghum, community labor). Because this method is hard to manage, it is not very popular. Other sources of community-based and self-financing include the Tribhuvandas Foundation, which provides health care through village milk cooperatives, and Amul Union (the milk cooperative organization), which puts a tax on milk col- lection to pay for health care. Tables 9.2 and 9.3 describe select schemes. Most of the successful case studies (Dave 1991) happen to be in the states of Assam, Gujarat, Maharashtra, Nadu, Orissa, Tamil, and West Bengal. The experience of such schemes could be useful for understanding their merits and disadvantages and their potential for replica- tion in other states. The most pertinent points about these schemes are that they are rurally oriented and have the ability to mobilize resources in a village com- munity. However, most of these schemes have served only a small segment of the population, and their health coverage has been restricted to elementary, pre- ventive, and maternal and child health (MCH) care. Microcredit Linked Health Insurance Schemes To break the vicious circle of poverty, malnutrition, disease, low productivity, and low income, several NGOs and governments in developing countries have started micro- credit schemes for vulnerable groups. Microcredit is considered not only an effective tool for reducing poverty but also an instrument for empowering the poor, especially women. This operation generates income for the poor by extending small credits for self-employment and other economic activities. However, loan repayments by these groups fell far below the expected level. The experience suggested that ill health, expenditures on treatment, and associated consumption needs were the prime rea- sons for defaults. To stop the erosion of borrowers' income by health care needs, some NGOs (such as Grameen Bank in Bangladesh and the Self-Employed Women's Associ- ation [SEWA] in India) have introduced health insurance schemes for their members. The Grameen Bank Health Program was started in 1994 to adopt disease-prevention measures, to arrange for minimum-cost treatment, and to build a nonprofit primary health care system. Under this scheme, the borrowers were asked to pay a fixed annual amount of 60 taka per family as a premium and a small sum at the time of use. A 2000 study found that the scheme had met the desired objectives (Rahman 2000). TABLE 9.2 Salient Characteristics of Selected NGO-Managed Health Insurance Schemes Voluntary Health service Total organizations/ Date Service delivery/ Population annual location started provided organization served cost (Rs.) Sevagram/ Hospital--1945 500-bed hospital Trained male VHW -- -- Wardha, Community health Outreach community provides basic 19,457 69,459 Maharashtra program--1972 health program curative, preventive, and promotive health care. Mobile with doctor and ANM provides care every 2 months. Bombay 1947 Health activities: Outpatient and -- 120,175 Mother and two maternity inpatient maternity (health and Child Welfare hospitals (40 beds care nonhealth Society each) with child Outpatient pediatric combined) (BMCWS)/ welfare centers care, including Chawla, Nonhealth activities: immunization Bombay day care centers, convalescent home Raigarh Hospital--1969 Federation of 3 RAHA functions 400,000 30,000­50,000 Ambikapur Community health referral hospitals and include management (cost range of Health services--started 65 independent of insurance scheme, individual Association 1974 health centers with training, and support health centers (RAHA)/Raigarh, outreach community for health centers. of which there Madhya Pradesh care Health centers staffed are 65) by nurse provide outpatient care, run MCH clinic. VHWs provide community- based care. Christian Hospital--1954 120 bed hospital Outpatient/inpatient -- 1,911,740 Hospital/ Outreach Community project care, specialties (hospital only) Bissamaucuttak, community currently not include obstetrics, Orissa care--1980 operational gynecology, surgery, ophthalmology UPASI/ 19th century Association of tea CLWS provides 250,000 300,000 Coocnoor, CLWS--1971 growers runs training and Tamil Nadu comprehensive labor management support welfare scheme to health programs of (CLWS) individual tea estates. Tea estates have small cottage hospital and outreach care provided by local workers. Goalpur 1964 Dispensary, periodic Doctor provides 1,247 32,000 Co-operative community health outpatient care Health Society/ services twice weekly Shanthiniketan, West Bengal (continued) 297 TABLE 9.2 Continued Voluntary Health service Total organizations/ Date Service delivery/ Population annual location started provided organization served cost (Rs.) Students 1955 Polyclinic plus 28 Polyclinic has 20 beds, 550,000 2,950,745 Health Home/ regional clinics provides outpatient West Bengal and inpatient care. Regional clinics provide outpatient care only, health education campaigns, and blood donation camps. Saheed 1978 Dispensary, Doctors provide -- 87,780 Shabsankar occupational health outpatient care Saba Samithi activities, rural health weekly at MCH (SSSS)/ program, school clinic. Burdwan, health program, and West Bengal fair-price medicine shop Arvind Eye 1976 2 urban hospitals Outpatient and -- 10,987,700 Hospital/ (100 beds), 2 rural inpatient eye care Madurai, hospitals (500 beds), provided. Regular eye Tanil Nadu and outreach program camps organized. Tribhuvandas 1980 Community-based CHWs provide basic 300,000 1,080,000 Foundation/ health program linked curative, preventive, (health and Anand, Gujarat with milk cooperatives, and promotive care. nonhealth regional rehabilitation Field supervisors combined) centers, and Balwadis provide support to women's income- CHWs, milk society generating scheme. building used as base for coordinating health services. SEWA/ Union--1972 Union of self- Health centers 63,000 391,850 Ahemadabad, Health employed women established in urban (health Gujarat program--1984 helps organize women slums and rural program only) into cooperatives of villages. CHWs various traders and provide basic care; provides credit doctors provide facilities. It provides support twice health care as a weekly. support that stocks rational generic drugs. CINI/Daulatpur, 1975 Community-based CHWs provide MCH 70,000 1,900,000 West Bengal health programs, care through (community dispensary, and Mahila Mandals. health outreach rehabilitation Doctors run daily project) center. Other activities: OPD, weekly MCH income-generating clinic, supplementary schemes, farm, health feeding. training, research Notes: -- not available; ANM, auxiliary nurse and midwife; MCH, maternal and child health; VHW, village health worker. Source: Dave (1991). 298 TABLE 9.3 Prepayment and Insurance Mechanisms in Selected NGO-Managed Health Insurance Schemes Tribhuvandas Students Features Sevagram RAHA Foundation Goalpur Health Home SSSS Coverage Household Individual Household Household Institutional Individual provided and individual Annual 8 payali sorghum Rs. 5 or Rs. 2 rice Rs. 10 Rs. 18 in cash or Rs. 2-- Rs. 2 or Rs. 5 subscription (landless) and 2 in kind (rice or institutions fee payali sorghum labor) Rs. 6--individuals per acre extra (landholders), or equivalent cash Number At least 75% 75,000 Approximately 150 out of 175 630 institutes; 6,800 of members of households 1/5 to 1/6 of households in total 350,000 (23 villages all households village students covered). Total in villages, (319 covered insured: 14,390 villages covered) Member Community care: Community care: Community care: Dispensary: Polyclinic/regional Outpatient entitlement free CHW free CHW free services, free doctor clinics: free clinic: free services, drugs, services and subsidized drugs. consultations consultations, consultations, and mobile drugs. Free Hospital: 50% and drugs at drugs, diagnostic drugs at cost, (doctor + ANM) health center subsidy cost; free tests, operations, and free MCH services. services, periodic public and bed stay at care Hospital: free including health activities nominal charges care for unphased MCH clinic. illness episodes; Hospital: free 25% subsidy for care after paying anticipated entrance fee up illness episodes, to ceiling of Rs. such as pregnancy 1000 and chronic ailments Nonmember Nonmembers Nonmembers Nonmembers Nonmembers Nonmembers Nonmembers entitlement not entitled to charged for have same charged for not entitled to are not entitled use community drugs (over cost), emoluments to drugs (over cost) avail themselves to avail health services not entitled to community of services themselves of attend MCH services as services clinic members but not to hospital care Management VHW Individual health VHW services Village health Institutions Able to enroll of fund responsible for centers responsible responsible for communities-- enrolled throughout the membership for membership membership funds collections once a year. year. No waiting collections. collections. collections. once a year. Individuals' period between Collections once Collections once Collected once enrollment enrollment and a year at harvest a year. New a year when ongoing (no service time. Compulsory members waiting bonus payments waiting period). entitlements. that 75% of period 2 months distributed villages covered. before services (nonadult society entitlements. members can Rs. 3 retained by also enroll in center, Rs. 2 to scheme). RAHA for referral fund. Source: Dave (1991). 299 300 Health Financing for Poor People: Resource Mobilization and Risk Sharing In India, SEWA is a trade union of 215,000 female workers in the informal sector. It organizes them at the household level toward the goals of full employment and self-reliance. Full employment includes social security, which in turn incorporates insurance. SEWA's experience revealed that women's efforts to escape from poverty through enhanced employment opportunities and increased income were repeatedly frustrated by crises such as sickness, a breadwinner's death, and accidental damage to, or destruction of, their homes and work equipment. Too often maternity also becomes a cri- sis for a woman, especially if she is poor, malnourished, and living in a remote area. One SEWA study observed that women identified sickness of themselves or their family members as the major stress events in their lives (Chatterjee and Vyas 1997). Sickness was also a major cause of indebtedness among women. From the start, the health insurance program was linked to SEWA's primary health care program, which includes occupational health services. Thus insured members also have access to preventive and curative health care with health education. Health insurance accounts for most of the claims and for 50 percent of the premiums paid out to the insurance program by SEWA members. The SEWA Bank introduced the scheme in March 1992, with an initial enrollment of 7,000 women from Ahmedabad City (Chatterjee and Vyas 1997). Later extended to cover rural woman from nine districts of Gujarat, it now has 30,000 women enrolled, half of them rural dwellers. Health insurance is an integral part of SEWA's insurance program. The main motivation for initiating a women's health insurance scheme was the recogni- tion that maintaining an active, health-seeking behavior is vital for ensuring a good quality life and women tend to place a low priority on their own health care needs. The SEWA health insurance program includes maternity coverage, hospital- ization coverage for a wide range of diseases, and coverage for occupational illnesses and diseases specific to women (table 9.4). It covers diseases not cov- ered by the GIC's Mediclaim plan and also provides life and asset insurance for the woman and for her husband or, in the case of widowhood or separa- tion, for other household members. Administrative procedures under the plan are simplified. The SEWA health insurance scheme functions in coordination with the Life Insurance Corporation of India (LIC) and the New India Assurance Company (NIAC). SEWA has integrated the schemes of LIC and NIAC into a comprehen- sive health insurance package to address women's basic needs. The claimants are needy health-benefits seekers, and as the insurance is an additional benefit, the beneficiaries willingly pay the premium. Most of the insurers opt for a fixed deposit of Rs. 500 or Rs. 700 (depending upon the type of coverage) with the SEWA Bank; accrued interest on the deposit goes toward the annual premium. The SEWA Bank's large membership and assets have enabled it to provide this insurance coverage at low premiums. The Potential Role of Community Financing in India 301 TABLE 9.4 Coverage under SEWA Scheme Coverage Premium Provider Description of coverage amount (Rs.) (Rs.) New India Assurance Accidental death of the woman member 10,000 3.50 Loss of assets Accidental death of a member's husband 10,000 3.50 SEWA Loss during riots, fire, floods, theft, and so forth 8.00 (a) of work equipment 2,000 (b) of the housing unit 3,000 Health insurance, including coverage for: 1,200 30.00 (a) gynecological ailments (10) (b) occupational health-related diseases (5) Maternity benefits 300 -- Life Insurance Corporation of India Natural death 3,000 15.00 Accidental death 25,000 Note: Total premium for the entire package is Rs. 60 plus a service charge of Rs. 5. RESEARCH DESIGN AND METHODOLOGY Research Design This chapter is based on a primary household survey undertaken in Gujarat's Ahmedabad District in 1998­99. The survey covered about 1,200 households from rural and urban areas. The households were stratified into four categories according to health insurance status. About 360 households belonged to a con- tributory plan known as the Employees' State Insurance Scheme (ESIS) for indus- trial workers. Another 120 households subscribed to a voluntary plan (Mediclaim), and 360 households belonged to a community-based financing scheme run by an NGO, the Self-Employed Women's Association (SEWA). The remaining 360 households were uninsured and were purchasing health care ser- vices directly on the market. This last subsample served as a control group. The idea behind selecting such stratification was to understand the varying health needs, access to health services, treatment pattern, and the types of benefits received by sample households in the different health insurance environments. The survey was conducted in eight slum-dominated localities in the city of Ahmedabad and six neighboring villages. On average, 60 households per village and 90 households per urban locality were selected. The criterion for selecting a village or an urban locality was that the settlement should have a cluster of households covered by the SEWA and ESIS plans. The sample canvassed from each settlement included about equal numbers of households from the ESIS, SEWA, and uninsured categories (20 each from a village and 30 each from an 302 Health Financing for Poor People: Resource Mobilization and Risk Sharing urban locality). The sample was purposive, and no house listing prior to the sur- vey was carried out. The sample of Mediclaim/Jan Arogya beneficiaries belonging to Ahmedabad city was selected from the list of subscribers obtained from the offices of two companies, United India Insurance and New India Assurance. Methodology Determinants of Participation in Mutual Health Organization (9.1) Prob (membership > 0) = X + , where X represents a set of independent variables that are hypothesized to affect membership in community-based schemes. These variables include income, gender, age, marker on chronic illness, and disability. is a vector of coefficient estimates and is the error term. Level of Financial Protection Provided by SEWA To assess the impact of mutual health organizations on financial protection of members, two aspects have to be taken into account: the probability of visiting a health care provider and the out-of-pocket expenditure borne by the individ- ual. We use a two-part model developed as part of the Rand Health Insurance Experiment: · a logit model that assesses the probability of visiting a health care provider (9.2) Prob (visit > 0) = X + u, where X stands as a vector for individual, household, and community characteristics; · a log-linear model that estimates the incurred level of out-of-pocket expen- ditures per episode, conditioning on positive use of health care services: (9.3) Log (out-of pocket expenditure / visit > 0) = X + e, where X represents a set of independent variables hypothesized to affect individ- ual patterns of utilization and expenditure on treatment. Variables Used in the Model Table 9.5 gives an overview of the variables included in the analysis. RESULTS Determinants of Participation A multinomial logit model is used to identify various determinants of being enrolled in the SEWA health insurance plan among members of SEWA. Out of the total 645 SEWA members above 15 years of age in the sample, 236 (36.6 per- cent) were enrolled in the plan. Out of 10 variables used in the model, 3 depicted The Potential Role of Community Financing in India 303 TABLE 9.5 Overview of Variables Used Variable Description Individual characteristics Gender Male and female Age Completed years of age at the time of last birthday; broad age groups are used in the model Marital status Never married, currently married, and widowed/divorced/separated Education level Years of schooling: broadly classified as illiterate, below primary, primary, middle, secondary, graduate, and above Activity status Usual activity status during the past year: broadly classified as nonworker, self- employed in agriculture, casual laborer, home-based production worker, trade or sales worker, salaried worker in organized sector, salaried worker in unorganized sector, and subsidiary status worker Health characteristics Acute morbidity Episode of illness during past 30 days not involving hospitalization Chronic morbidity Prevalence of any chronic disease/ailment Hospitalization Any illness resulting in hospitalization during the past 365 days Childbirth Childbirth during past two years Duration of illness Number of days the person was ill and also categorized into groups Source of care Source of treatment; broadly categorized as no treatment, use of public facility, including ESIS, and private facility Cost of treatment Cost of treatment includes: direct out-of-pocket payments toward fees, medicines, diagnostic tests, surgery, bed charges, transportation, and special diet. Indirect costs include income or wage loss of the patient and the caring person as well as interest payments on amount borrowed to meet treatment expenses. Health insurance enrollment Community-based plan (SEWA), social insurance (ESIS), private plan (Mediclaim), and uninsured Household characteristics Income and expenditure Annual household income from different sources; categorized into quintile groups. Monthly household expenditure by broad items--but not considered in the model. Household size Number of members usually residing in the house and sharing food from the common kitchen Community characteristics Area of residence Usual place of residence: rural or urban area health status (whether suffering from any chronic ailment, hospitalized during last 365 days, and had delivery during the past two years), 4 personal character- istics (age, education, marital status, and activity status), 2 household character- istics (household size and income quintile), and 1 community variable (area of residence). The description of these variables is provided in table 9.5. The mean value of the enrollment rate varied across these characteristics. The enrollment 304 Health Financing for Poor People: Resource Mobilization and Risk Sharing rate was higher among women who had reported suffering from any chronic ail- ment or had been hospitalized in the previous 365 days but not among those who had reported delivery during the past two years. Among personal character- istics, the mean enrollment rate was found to be higher in the middle age groups, 36­45 years and 46­55 years, than the other age groups; it was also higher among currently married women. However, with level of education the mean enroll- ment rate tended to decline. The rate was found to be much lower among non- workers or subsidiary status workers than among home-based production or salaried workers. The rate tended to decline with the size of household and did not vary much across income quintiles, except in the top quintile, where it was marginally higher. Overall, the enrollment rate was higher among urban women than rural women, mainly due to better access to information as well as to the SEWA Bank (located in Ahmedabad city), which manages the scheme. The alternative results of multinomial logit models (interchanging activity status and income variables) are presented in table 9.6. The explanatory power of the model (Pseudo R2) ranged between 0.185 (without income variable) to 0.218 (with inclusion of both income and activity status variables). The follow- ing are the main findings. There was no adverse selection in terms of whether the member had been suffer- ing from any chronic ailment or had been hospitalized before. However, maternity, a predictable event, had increased the likelihood of enrollment to take advantage of a benefit allowance of Rs. 300 and coverage of the high risk of hospitalization. Among the personal attributes, the odds of being enrolled were five to seven times higher among middle-age groups than in the 16­25 years age group. For currently married women, the odds were twice as high as for never-married women. Education level turned out to be an insignificant predictor. The type of activity pursued by a SEWA member was found to be a highly significant predic- tor (the predictive power was much higher than that of the income effects). The odds ratios were much higher for self-employed, home-based, or agricultural workers than for nonworkers. The odds were found to be insignificant for salaried workers in the formal sector. Household size showed an inverse relationship with enrollment, and the odds ratios tended to decline significantly in medium-size and large households. Income was not found to be a significant predictor. When activity status was not taken into account, women in the top income quintile were twice as likely to enroll as women in the bottom quintile. There seemed to be an urban bias in enrollment, which may be due to better outreach and accessibility factors. An urban woman was three times more likely to enroll than a rural woman. Determinants of Financial Protection in Community Financing A multinomial logit model is used to identify various determinants of utilization of services for ambulatory care, and an attempt is also made to explore predictors The Potential Role of Community Financing in India 305 for choosing a private facility for ambulatory and inpatient care over a public one. This model uses the cases of illnesses reported by all households, irrespec- tive of health insurance status (SEWA, ESIS, Mediclaim, and uninsured). Out of the total 1,327 illnesses reported by the sample population during the previous 30 days, treatment was sought for 1,271 ailments (96 percent). The first model uses all illnesses (excluding hospitalization) reported during the previous 30 days and predicts the probability of seeking treatment (only for 56 illnesses was the treatment not sought). The second and third models are subsets of the first model (treated in public facility versus no treatment--383 cases--and treated in private facility versus treated in public facility--1,271 cases). The third model shows that of the total treated ambulatory cases, nearly 74 percent relied on the private facility, thus suggesting the dominant role of the private sector in han- dling the ambulatory care burden. The last model is exclusively for hospitaliza- tion cases during previous 365 days (that is, treated in a private hospital versus public hospital--362 cases). Here the inpatient load was almost equally distrib- uted between the private and public sectors (53 percent of inpatients used public hospitals). Of 11 variables used in the model, 3 depicted health characteristics (whether suffering from any chronic ailment, duration of illness, and type of health insur- ance coverage), 5 personal characteristics (gender, age, education, marital status, and activity status), 2 household characteristics (household size and income quintile), and 1 community variable (area of residence). The mean value of uti- lization and the proportion using private health service facilities both for ambu- latory and inpatient care varied considerably across these characteristics. The results of multinomial logit models for utilization and private/public choice for ambulatory care are presented in table 9.7. The explanatory power of the utilization model (Pseudo R2) was 0.148. For the other two models, it was 0.372 (treated in a public sector facility versus no treat- ment) and 0.226 (treated in a private versus a public sector facility). The follow- ing are the main findings. Of 11 variables used for predicting utilization rate, only 3 (illness duration, type of health insurance enrollment, and area of residence) were found to be sig- nificant. None of the personal and household attributes exerted significant influ- ence on utilization rate. The odds of being untreated were higher among those enrolled with the community plan (SEWA) as well as among rural residents. In the case of choosing a private facility over the public facility, including ESIS for ambulatory care, seven variables had a significant impact. The odds of choos- ing a public facility were higher if the person had a chronic ailment, a salaried work status, or coverage under social insurance. Males, educated graduates and above, and those covered by a private plan (Mediclaim) tended to choose a private facility for ambulatory treatment. Patients from small households in urban areas tended to choose a public facility. The income effect for opting out the private facility was clearly discernable. Members of the SEWA plan also tended to choose the private facility for ambulatory care. 306 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 9.6 Determinants of Being Enrolled in SEWA Health Insurance Plan among SEWA Members Model 1 Variables Odds ratio Coefficient Significance Whether chronic ailment 1.155 0.144 0.655 Whether hospitalized 1.602 0.471 0.152 Whether had children in the past 2 years 1.835 0.607 0.047 Urban resident 3.096 1.130 0.000 Activity status (nonworker) Agricultural 3.357 1.211 0.011 Casual labor 3.006 1.101 0.001 Home-based worker 4.095 1.410 0.001 Trade or sales worker 2.475 0.906 0.013 Salaried worker--organized 2.257 0.814 0.136 Salaried worker--unorganized 2.753 1.013 0.006 Other worker--subsidiary status 2.016 0.701 0.089 Education level (0) 1­4 std. 1.400 0.336 0.292 5­7 std. 0.668 ­0.404 0.147 8­9 std. 0.540 ­0.617 0.122 10­12 std. 0.721 ­0.328 0.334 Graduate and above 1.483 0.394 0.530 Age (16­25 years) 26­35 2.235 0.804 0.008 36­45 5.444 1.694 0.000 46­55 6.729 1.906 0.000 56 and above 4.453 1.494 0.002 Marital status (never married) Currently married 2.089 0.737 0.099 Widowed/divorced/separated 1.154 0.143 0.799 Household size (1­4) ­0.375 5­6 0.687 ­0.907 0.170 6­8 0.404 ­0.925 0.004 9­10 0.397 ­1.391 0.015 11 and above 0.249 0.005 Annual household income quintile (lowest) 2 0.867 ­0.143 0.659 3 1.182 0.167 0.643 4 1.094 0.090 0.785 5 (top) 1.872 0.627 0.098 Constant ­2.929 0.000 Pseudo R2 0.218 The Potential Role of Community Financing in India 307 Model 2 Model 3 Odds ratio Coefficient Significance Odds ratio Coefficient Significance 1.164 0.152 0.630 1.121 0.114 0.719 1.716 0.540 0.087 1.528 0.424 0.193 1.480 0.392 0.184 1.761 0.566 0.063 2.720 1.001 0.000 3.131 1.141 0.000 3.316 1.199 0.012 2.777 1.021 0.002 4.835 1.576 0.000 2.599 0.955 0.009 2.328 0.845 0.115 2.802 1.030 0.004 1.811 0.594 0.145 1.173 0.160 0.600 1.511 0.413 0.190 0.715 ­0.336 0.208 0.701 ­0.355 0.196 0.484 ­0.725 0.062 0.578 ­0.549 0.163 0.599 ­0.512 0.111 0.805 ­0.217 0.512 1.058 0.056 0.927 1.813 0.595 0.325 2.752 1.012 0.000 2.203 0.790 0.009 6.294 1.840 0.000 5.801 1.758 0.000 6.746 1.909 0.000 6.878 1.928 0.000 3.334 1.204 0.010 4.867 1.582 0.001 2.251 0.811 0.061 1.939 0.662 0.134 1.299 0.262 0.629 1.122 0.116 0.835 0.698 ­0.359 0.176 0.792 ­0.233 0.370 0.455 ­0.787 0.009 0.482 ­0.729 0.011 0.470 ­0.755 0.039 0.497 ­0.700 0.044 0.264 ­1.333 0.005 0.367 ­1.001 0.026 0.885 ­0.122 0.699 1.202 0.184 0.594 1.203 0.185 0.564 2.106 0.745 0.041 ­2.490 0.000 ­2.936 0.000 0.185 0.212 TABLE 9.7 Determinants of Being Treated and Use of Public or Private Facility for Ambulatory Care (Multinomial Logit Model) Treated vs. untreated Predictor Odds ratio Coefficient Significance Illness duration (1­3 days) 4­7 0.9975 ­0.003 0.995 8­14 1.5286 0.424 0.369 15­29 7.6682 2.037 0.057 30 and older 5.4784 1.701 0.095 Whether chronic ailment 0.2392 ­1.430 0.154 Whether male 0.8775 ­0.131 0.697 Age (0­14 years) 15­24 2.9836 1.093 0.337 25­34 0.9822 ­0.018 0.990 35­44 0.8176 ­0.201 0.893 45­54 0.5992 ­0.512 0.730 55 and older 0.4542 ­0.789 0.588 Marital status (never married) Currently married 1.2941 0.258 0.853 Widowed/divorced/separated 1.3710 0.316 0.828 Education level (illiterate) 1­4 Std. 0.9347 ­0.068 0.852 5­7 Std. 1.1325 0.124 0.765 8­9 Std. 1.09E+14 32.323 1.000 10­12 Std. 1.8429 0.611 0.388 Graduate and above 1.0233 0.023 0.984 Activity status (nonworker) Agricultural 1.5643 0.447 0.590 Casual labor 1.0278 0.027 0.956 Home-based worker 0.5168 ­0.660 0.426 Trade or sales worker 0.7356 ­0.307 0.627 Salaried worker--organized 1.23E+14 32.441 1.000 Salaried worker--unorganized 1.34E+14 32.527 1.000 Other worker--subsidiary status 1.1006 0.096 0.905 Health insurance enrollment (uninsured) Community plan--SEWA 0.3669 ­1.003 0.035 Social insurance--ESIS 1.1657 0.153 0.671 Private plan--Mediclaim 9.50E+13 32.185 1.000 Urban resident 2.0618 0.724 0.019 Household size (1­4) 5­6 0.8953 ­0.111 0.750 7­8 1.6893 0.524 0.293 9­10 1.1413 0.132 0.832 11 or more 1.0451 0.044 0.951 Annual household income quintile (lowest) 2 1.8603 0.621 0.168 3 0.9483 ­0.053 0.902 4 1.8409 0.610 0.227 5 (top) 1.3011 0.263 0.613 Constant 1.915 0.000 Pseudo R2 0.148 Number of cases (dependent valued as): 1 1,271 0 56 Note: Out of 1,327 illness episodes reported during the past 30 days, 56 were not treated. Of the treated episodes, the public facility, including ESIS, was contacted in 327 cases; in the remaining 944 cases a private facility was contacted. Figures in brackets refer to the reference category of the variable. 308 Public contact vs. untreated Private vs. public contact Odds ratio Coefficient Significance Odds ratio Coefficient Significance 0.6281 ­0.465 0.426 0.8946 ­0.111 0.661 1.4671 0.383 0.564 0.8339 ­0.182 0.492 8.3126 2.118 0.081 0.5712 ­0.560 0.065 4.6462 1.536 0.311 1.4982 0.404 0.304 0.5453 -0.606 0.678 0.2108 ­1.557 0.000 0.7512 ­0.286 0.561 1.4326 0.359 0.046 10.0297 2.306 0.076 0.6249 ­0.470 0.151 1.2636 0.234 0.855 0.6186 ­0.480 0.329 2.1255 0.754 0.617 0.4862 ­0.721 0.148 0.3484 ­1.054 0.471 0.6338 ­0.456 0.376 0.5865 ­0.534 0.707 0.4829 ­0.728 0.149 0.8231 ­0.195 0.883 1.9124 0.648 0.124 0.9973 ­0.003 0.999 1.7128 0.538 0.286 0.6011 ­0.509 0.307 0.9139 ­0.090 0.696 1.4707 0.386 0.487 0.6471 ­0.435 0.052 2.48E+15 35.448 1.000 0.7109 ­0.341 0.252 0.9979 ­0.002 0.998 0.8000 ­0.223 0.432 0.1647 ­1.804 0.292 3.2589 1.181 0.036 0.6132 ­0.489 0.699 0.8316 ­0.184 0.745 3.4850 1.248 0.097 0.6730 ­0.396 0.160 1.6128 0.478 0.668 0.9354 ­0.067 0.894 2.0847 0.735 0.397 0.5980 ­0.514 0.213 5.35E+15 36.217 1.000 0.2761 ­1.287 0.000 4.54E+15 36.051 1.000 0.5682 ­0.565 0.083 7.4579 2.009 0.080 0.2318 ­1.462 0.001 0.0798 ­2.529 0.003 2.1073 0.745 0.043 5.9587 1.785 0.000 0.1715 ­1.763 0.000 8.40E+15 36.666 . 4.4984 1.504 0.050 8.1439 2.097 0.000 0.2525 -1.377 0.000 2.0201 0.703 0.166 0.5795 ­0.546 0.006 7.4421 2.007 0.004 0.4030 ­0.909 0.000 2.0905 0.737 0.386 1.0896 0.086 0.806 4.4154 1.485 0.159 0.5556 ­0.588 0.138 0.6607 ­0.414 0.513 1.6769 0.517 0.033 0.1623 ­1.818 0.008 2.2582 0.815 0.002 0.1336 ­2.013 0.016 3.3027 1.195 0.000 0.1134 ­2.177 0.007 2.6790 0.985 0.001 ­0.550 0.481 3.116 0.000 0.372 0.226 327 944 56 327 309 310 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 9.8 Determinants of Using Private Facility for Inpatient Care (Multinomial Logit Model) Private vs. public hospital Predictor Odds ratio Coefficient Significance Illness duration (1­3 days) 4­7 0.4685 ­0.7583 0.026 8­14 0.1998 ­1.6105 0.000 15­29 0.1299 ­2.0410 0.000 30 or more 0.3191 ­1.1423 0.027 Whether male 1.1257 0.1184 0.719 Age (0­14 years) 15­24 1.2172 0.1966 0.725 25­34 3.6314 1.2896 0.069 35­44 1.5422 0.4332 0.591 45­54 3.2025 1.1639 0.132 55 or more 1.1200 0.1134 0.877 Marital status (never married) Currently married 0.5864 ­0.5337 0.345 Widowed/divorced/separated 0.7650 ­0.2679 0.730 Education level (illiterate) 1­4 Std. 0.9722 ­0.0282 0.947 5­7 Std. 0.5129 ­0.6678 0.089 8­9 Std. 0.7450 ­0.2944 0.575 10­12 Std. 0.6452 ­0.4383 0.336 Graduate and above 2.0286 0.7074 0.419 Activity status (nonworker) Agricultural 0.5824 ­0.5406 0.537 Casual labor 0.5931 ­0.5224 0.259 Home-based worker 0.6837 ­0.3802 0.622 Trade or sales worker 2.0432 0.7145 0.382 Salaried worker--organized 1.4176 0.3489 0.553 Salaried worker--unorganized 0.4821 ­0.7297 0.158 Other worker--subsidiary status 0.5435 ­0.6097 0.443 (continued) For inpatient care, the results of the multinomial logit model for choosing a public or private hospital are presented in table 9.8. Of the 10 variables used in the model, only 4 (illness duration, type of health insurance enrollment, area of residence, and income) showed a significant influence. The odds of choosing a public hospital for inpatient care were much higher for illnesses requiring a longer stay in a hospital. This is entirely due to price considerations because for longer stays, out-of-pocket expenditure would be huge if treatment were in a pri- vate hospital. People covered under social insurance tended to use public and ESIS hospitals much more. Only patients from households in the top quintile, who could afford treatment, chose private hospitals over public for inpatient The Potential Role of Community Financing in India 311 TABLE 9.8 Continued Private vs. public hospital Predictor Odds ratio Coefficient Significance Health Insurance Enrollment (uninsured) Community plan--SEWA 0.7946 ­0.2299 0.630 Social insurance--ESIS 0.2143 ­1.5404 0.000 Private plan--Mediclaim 5.1690 1.6427 0.171 Urban resident 0.2855 ­1.2536 0.000 Household size (1­4) 5­6 0.6137 ­0.4882 0.169 7­8 0.8579 ­0.1533 0.722 9­10 0.9079 ­0.0966 0.859 11 or more 0.5836 ­0.5386 0.447 Annual household income quintile (lowest) 2 1.5507 0.4387 0.315 3 1.4488 0.3707 0.409 4 1.8937 0.6385 0.159 5 (top) 4.7391 1.5558 0.003 Constant 1.8405 0.001 Pseudo R2 0.228 Number of cases (dependent variable coded as): 1 171 0 193 care. As public hospitals are located mainly in urban centers, urban residents have better access and thus showed greater reliance on public services than did their rural counterparts. Determinants of out-of-pocket expenditures on treatment of ailments, for both ambulatory and inpatient care, are presented in table 9.9. The dependent variable is expressed as the log of out-of-pocket expenditure on treatment. Overall, the direct cost of treatment for ambulatory care was Rs. 286 per episode. For inpatient care, it was Rs. 2,771. Of the 12 variables used in the OLS regression model, 4 depicted health characteristics (chronic ailment, duration of illness, type of provider--private or public--and type of health insurance coverage), 5 personal characteristics (gender, age, education, mari- tal status, and activity status), 2 household characteristics (household size and income quintile), and 1 community variable (area of residence). The mean value of direct out-of-pocket expenditures per episode varied significantly across these characteristics. The regression results for both ambulatory and inpatient cares are presented in table 9.9. The explanatory power of the model (R2) was 0.284 for ambulatory care and 0.413 for inpatient care. The following are the main findings. 312 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 9.9 Determinants of Out-of-Pocket Expenditure on Treatment by Type of Care (Dependent Variable in Log Form) Ambulatory care Hospitalization Predictor Coefficient Beta Significance Coefficient Beta Significance Private provider 0.5502 0.3181 0.000 0.4175 0.3230 0.000 Days of illness 0.0075 0.1735 0.000 0.0111 0.2357 0.000 Whether chronic ailment 0.0830 0.0483 0.108 Whether hospitalized Whether male 0.0982 0.0646 0.025 0.0298 0.0230 0.672 Age 0.0081 0.2212 0.159 0.0098 0.2939 0.266 Age squared ­0.0001 ­0.2117 0.095 ­0.0001 ­0.2696 0.257 Marital status (never married) Currently married ­0.0253 ­0.0167 0.767 ­0.0500 ­0.0382 0.657 Widowed/divorced/separated ­0.0528 ­0.0207 0.648 0.1315 0.0623 0.421 Education level (0) 1­4 Std. ­0.0297 ­0.0148 0.590 ­0.0342 ­0.0187 0.708 5­7 Std. ­0.0328 ­0.0168 0.574 0.0115 0.0073 0.893 8­9 Std. 0.0722 0.0283 0.322 0.1183 0.0540 0.287 10­12 Std. 0.0539 0.0253 0.417 0.1245 0.0769 0.182 Graduate and above 0.0508 0.0132 0.637 0.1505 0.0449 0.358 Activity status (nonworker) Agricultural ­0.0806 ­0.0177 0.499 0.1211 0.0275 0.546 Casual labor ­0.0896 ­0.0354 0.204 0.1648 0.0808 0.111 Home-based worker ­0.2041 ­0.0389 0.119 0.3035 0.0873 0.050 Trade or sales worker ­0.0670 ­0.0174 0.509 0.1527 0.0439 0.342 Salaried worker--organized ­0.0179 ­0.0066 0.827 0.0641 0.0314 0.588 Salaried worker--unorganized 0.0655 0.0216 0.420 0.0633 0.0289 0.574 Other worker--subsidiary status 0.2297 0.0490 0.050 ­0.2883 -0.0694 0.118 Health insurance enrollment (uninsured) Community plan--SEWA 0.0145 0.0048 0.855 0.0656 0.0296 0.535 Social insurance--ESIS ­0.3719 ­0.2220 0.000 ­0.4274 ­0.3065 0.000 Private plan--Mediclaim 0.1461 0.0353 0.177 0.3449 0.0784 0.080 Urban resident ­0.2437 ­0.1477 0.000 ­0.1786 ­0.1290 0.007 Household size ­0.0172 ­0.0560 0.049 0.0182 0.0673 0.199 Annual household income quintile (lowest) 2 0.0666 0.0363 0.265 0.0535 0.0337 0.564 3 0.1132 0.0575 0.074 0.1634 0.0981 0.086 4 0.1446 0.0784 0.024 0.1059 0.0697 0.268 5 (top) 0.0515 0.0272 0.466 0.0785 0.0452 0.485 Constant 1.7059 0.000 2.5908 0.000 R2 0.284 0.413 Number of illness episodes 1,274 363 The Potential Role of Community Financing in India 313 Of the 12 variables used for determining direct out-of-pocket expenditure on ambulatory care, only 7 (type of provider, illness duration, gender, type of health insurance enrollment, household size, income, and area of residence) were found to be significant. Of these, the most important explanatory variables were type of provider, duration of illness, social insurance coverage, and area of residence. Cost of treatment turned out to be higher if treatment were in the private sector and of long duration when the patient was male and resided in a rural area. The cost of care was inversely related to household size and relatively higher among patients in the third and fourth income quintiles. Only social insurance coverage, and not the community plan, provided financial protection. Both the commu- nity plan and the private Mediclaim plan cover hospitalization only. In the case of out-of-pocket expenditures on inpatient care, only 4 of the 11 variables had a significant impact. The cost of treatment for inpatient care was higher if treatment were in a private hospital and of long duration and the patient resided in a rural area. Income effects were not found to be significant. In this case, both social insurance and Mediclaim plans succeeded in providing financial pro- tection whereas the community plan did not meet expectations. Another way of looking at the financial protection is to explore determinants of annual per capita expenditure on health care at the household level (after obtaining the annual estimates of expenditure on ambulatory care, inpatient care, delivery, and maternal and child health care). Alternatively, one can also estimate the burden of ill health on the household (annual per capita expendi- ture on health care as a proportion of annual per capita income) and explore how much of this burden is protected through a health insurance mechanism. Acknowledgments: This chapter further explores household-level data of the NCAER-SEWA study on "Health Insurance for Workers in the Informal Sector." The author is thankful to the Ford Foundation (the sponsor) and SEWA as collaborator in the primary fieldwork in Ahmed- abad. The author would also like to thank Ms. Shakun Datta for providing assistance in extract- ing results of multinomial logit models through STATA software. The author is grateful to the World Health Organization for having provided an opportunity to contribute to the work of the Commission on Macroeconomics and Health and to the World Bank for publishing the mater- ial in this chapter as an HNP Discussion Paper. REFERENCES Ahmad, E., J. Dreze, J. Hills, and A. Sen, eds. 1991. Social Security in Developing Countries. Oxford: Clarendon Press. Chatterjee, M., and J. Vyas. 1997. Organizing Insurance for Women Workers: The SEWA Expe- rience. Ahmedabad: Self-Employed Women's Association (SEWA). Dave, P. 1991. "Community and Self-Financing in Voluntary Health Programmes in India." Health Policy and Planning. 6(1): 20­31. Gumber, A. 1997. "Burden of Disease and Cost of Ill Health in India: Setting Priorities for Health Interventions During the Ninth Plan." Margin 29(2): 133­72. 314 Health Financing for Poor People: Resource Mobilization and Risk Sharing ------. 1998. "Facets of Indian Healthcare Market--Some Issues." Saket Industrial Digest 4(12): 11­17. ------. 2000a. "Health Care Burden on Households in the Informal Sector: Implications for Social Security Assistance." Indian Journal of Labor Economics 43(2): 277­91. ------. 2000b. "Structure of the Indian Health Care Market: Implications for Health Insur- ance Sector." WHO Regional Health Forum 4(1/2): 26­34. ------. 2001. "Economic Reforms and the Health Sector: Toward Health Equity in India." Paper presented at the National Seminar on Economic Reforms and Employment in the Indian Economy, organized by the Institute of Applied Manpower Research and Planning Commission, New Delhi, March 22­23. Gumber, A., and V. Kulkarni 2000. Health Insurance for Workers in the Informal Sector: Detail Results from a Pilot Study. New Delhi: National Council of Applied Economic Research. Rahman, K. 2000. "Poverty, Microcredit and Health--What Role Can WHO Play?" WHO Regional Health Forum 4 (1/2): 68­80. Shariff, A., A. Gumber, R. Duggal, and M. Alam. 1999. "Health Care Financing and Insur- ance Perspective for the Ninth Plan (1997­2002)." Margin 31(2): 38­68. Visaria, P., and A. Gumber. 1994. Utilization of and Expenditure on Health Care in India, 1986­87. Ahmedabad: Gujarat Institute of Development Research. CHAPTER 10 Impact of the Thailand Health Card Siripen Supakankunti Abstract: The health insurance card scheme was introduced as the Health Card Project (HCP) in 1983. This program was based on the risk sharing of health expenditures with no cost sharing in a voluntary health insurance prepayment scheme. Frequent adjustments in both the strategies and the objectives of the program have included voluntary risk shar- ing with cost recovery in addition to service provision. The HCP needs a large enough number of enrollees to ensure a sufficient pool of risks. However, the newly elected gov- ernment of Thailand has committed to rapidly extending health care coverage to all Thai citizens. All uninsured Thai citizens will have access to required health services for a flat fee of 30 baht (US$0.67), regardless of the type of disease treated. Even though the HCP was suppressed and replaced by this program in October 2001, this study of the determi- nants of demand for the prepaid health card is still important. The HCP can be assessed as relatively progressive, serving rural areas, poor, and near-poor groups. This study has found that employment, education, and the presence of illness are sig- nificant factors influencing health insurance card purchase. The third factor is related to the program's problem of adverse selection: families with symptoms of sickness are more likely to buy cards and increase their use of health services. The results also show an improvement in accessibility to health care and a high level of satisfaction among card- holders, both key objectives of the program. Problems of program performance include issues of program and financial management, marketing, quality control and cost recov- ery, ineffective referral systems, and lack of limits on episodes and ceilings for expenses. There is a need for an efficient and consistent health policy, which would involve revised criteria for card use, standard reimbursement agreements with hospitals, government subsidies, and an overall strengthening of the program. T he objective of this chapter is to assess the application of voluntary health insurance, in this case the Health Card Program (HCP) of Thailand, and provide greater understanding of how a voluntary health insurance pro- gram performs and how to improve and sustain it more efficiently. Utilizing data collected in Khon Kaen, where the program was implemented in 1983, the fol- lowing study explores the development, problems, and health service capabili- ties of a voluntary health insurance scheme, identifies factors influencing project outcomes (accessibility, sustainability, and efficient use of resources), investigates health card purchase and dropout patterns, and evaluates the factors affecting the discrimination between health card purchase and nonpurchase and also affecting dropouts from the continued health card member groups. To analyze any type of health care financing, it is first necessary to understand its background. 316 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 10.1 Health Service Utilization Pattern for Reported Ill Persons (Comparing Different Surveys) Choice of Outlet 1970 1979 1985 No treatment 2.7 4.2 -- Traditional medicine/healers 7.7 6.3 2.4 Self-prescribed drugs 51.4 42.3 28.6 Public health centers 4.4 16.8 14.7 Public hospitals 11.1 10.0 32.5 Private clinics/hospitals 22.7 20.4 21.8 Source: Tangcharoensathien (1995), originally from Health Planning Division, MOPH (1970,1979); Institute for Population and Social Research, Mahidol University (1988). Health Delivery System In Thailand, from 1970 to 1985, there was an increasing expectation and use of pub- lic outlets staffed by physicians and a decreasing trend in using self-prescribed drugs and traditional medicines, or attendance of healers. This reflects the aim of the Fifth National Health Development Plan (1982­86) to achieve 100 percent district cover- age in the country and the three-year compulsory service program at the Ministry of Public Health (MOPH) district hospitals imposed on all medical graduates in 1972. The consequence was an increase in the number of hospitals and doctors at the dis- trict level, which undoubtedly began to meet the previously unmet demand for health services in rural areas and led to a significant threefold increase in the use of public hospitals in 1985 compared with 1979 (Tangcharoensathien 1995). Table 10.1 shows the consistent trend in choices of outlets used by ill individuals. When looking at the employment status and type of health service utiliza- tions in 1986, it is clear that the percentage of health service utilization by self- treatment was highest in all occupational groups. Between the professional and administration groups and those of farmers and miners there were some differ- ences in proportions, that is, self-treatment was more than 50 percent in the farmers' group but lower than 50 percent in the professional group. The profes- sional group has relatively much higher use of private clinics and hospitals than the farmer group, while utilization rates in public facilities were similar. Trends in Health Expenditure In 1987, 5.7 percent of the gross national product (GNP) was spent on health, including private household payments and public health expenditures (see table 10.2). Health expenditure has been steadily increasing at a higher rate than the growth of the GNP. In 1984, a 12 percent real-term increase in health expendi- ture was due to the high capital investment for constructing district hospitals as stated above (Tangcharoensathien 1995). In 1994, 1996, and 1998, 3.53 percent, 3.72 percent, and 3.88 percent of the gross domestic product (GDP) was spent on health respectively (NESDB, 1998). Impact of the Thailand Health Card 317 TABLE 10.2 Trend of Total per Capita Health Expenditure (Public and Private Spending) 1987 Prices Year % of GNP on health Per capita expenditure (Baht)a % Increase health expenditure % GDP growth (Baht) 1978 3.4 680 -- -- 1979 3.6 710 4.4 5.05 1980 3.9 738 3.9 4.57 1981 4.2 798 8.1 5.96 1982 4.6 864 8.3 3.90 1983 4.8 939 8.7 6.76 1984 5.2 1,052 12.0 6.65 1985 5.6 1,132 7.6 3.40 1986 5.6 1,192 5.3 4.30 1987 5.7 1,282 7.6 7.74 Current prices 1994 3.53 2,186 1996 3.72 2,858 1998 3.88 2,924 Note: The GDP growth rate in 1988 and 1989 was 12.0 percent and 10.8 percent, respectively. a. 25 Baht:US$1 Source: Modified from Tangcharoensathien (1995), originally from Social Development Project Division, NESDB (1990). NHA Project phase II (2000). The percent source of health care financing in 1984, 1986, and 1987 at 1987 prices is shown in table 10.3, demonstrating trends in sources of finance during that decade. Public sources of funding play a minor role in financing health services with a decreas- ing trend from 27.9 percent of total health expenditure in 1984 to 24.2 percent in 1987. The most important source is private out-of-pocket expenditures by households, with an increasing trend from 69.3 percent in 1984 to 73.2 percent in 1987. It must be noted that some of the private household expenditures on health could have been reimbursed from the Ministry of Finance, for government employees and dependents, while others may have been reimbursed from employers. Among the public services, the MOPH is the major provider for the country, providing comprehensive health care mainly in areas outside the Bangkok metropolitan area (Tangcharoensathien 1995). Private household expenditures on health were mainly to pay for curative care either through user fees at government facilities or charges at private clinics and drug stores. In 1996, of the total 843,200 million baht government budget, 9.1 percent was for agriculture, 10.2 percent for transportation and communication, 20.1 percent for education, 14.3 percent for social services, 12.8 percent for maintenance of national security, and 7.53 percent for public health (see table 10.4). The budget for health remained consistent at the rate of 7.1­7.5 percent of the total public spending dur- ing fiscal years 1994­96, compared with 4.1­4.5 percent in the previous decade. All expenses are financed by the government's receipts of revenue and borrowings. In 318 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 10.3 Percent Source of Health Care Financing in 1984, 1986, and 1987 1987 Prices Sources of funding 1984 1986 1987 Public sources: 27.9 26.0 24.2 MOPH 17.4 15.3 14.1 Other ministries 6.9 6.5 6.0 Public employee medical benefits 3.6 4.2 4.1 Workmen's compensation fund 0.5 0.4 0.4 State enterprise employee medical benefits 0.8 0.9 0.8 Private insurance 0.8 0.7 0.7 Foreign aid 0.8 0.8 0.7 Private households 69.3 71.2 73.2 Total: Percent 100.0 100.0 100.0 Million Baht 53,032.9 62,099.9 67,771.3 Sources: Nittayaramphong and Tangcharoensathien (1994) originally from Health Planning Division, MOPH; National Accounts; Workmen's Compensation Fund, Ministry of Labour and Social Welfare; Controller-General's Department, Ministry of Finance; Financing Health Service and Medical Care in Thailand (1987 report), MOPH. TABLE 10.4 Budget Expenditures Classified by Program, Fiscal Years 1994­96 (Unit: Million Baht) Budget Expenditure (percent) Program 1994 1995 1996 Economic development 127,846.4 149,114.3 186,788.3 (22.15) Social development 243,404.2 277,923.7 353,363.7 (41.91) Education 124,457.9 137,641.4 169,560.7 (20.11) Public health 44,335.0 52,372.7 63,452.2 (7.09) (7.33) (7.53) Social services 74,611.3 87,909.6 120,350.8 (14.27) Maintenance of national security and maintenance 125,063.9 131,886.1 148,304.3 of internal peace and order (17.59) General services 70,172.7 111,345.1 106,751.7 (12.66) Debt services 58,512.8 44,730.8 47,992.0 (5.69) Total 625,000 715,000 843,200 (100) Source: Bureau of the Budget, Office of the Prime Minister. Impact of the Thailand Health Card 319 fiscal year 1996, of the total 843,200 million baht government's estimated receipts, 87.4 percent was from taxes, 7.0 percent from state enterprises, 3.0 percent from sales of assets and services, and 2.6 percent from miscellaneous sources. Only 36.56 percent of government tax revenue came from personal income tax and corporate income tax. All indirect taxes accounted for 71.31 percent: general sales tax 30.54 percent, export-import duties 18.09 percent, and specific sales tax 21.98 percent (calculated from Thailand in Figures, 1996). Health Insurance Development Health insurance is a means of financial protection against the risk of unexpected and expensive health care. In countries such as Thailand, where the use of gov- ernment health services is heavily subsidized, governments are implicitly cover- ing the risk of incurring high cost care. This limits the demand for more explicit types of risk-sharing arrangements. Explicit forms of health insurance are wide- spread in the industrialized or middle-income countries. In some countries, such as Canada and Japan and most European countries, coverage is compulsory and universal and the insurance program is financed from either general government revenue or payroll taxes. In the United States, insurance coverage is voluntary, administered by third parties, not universal, and usually financed by a combina- tion of employer and employee contributions. The expansion of health insurance coverage is usually intended to increase health sector revenues, reduce financial barriers to care, and improve the efficiency of resource allocation and use (Kutzin 1995). How to achieve universal coverage when the proportion of the population in the formal sector is relatively low, for example in a country such as Thailand, is one of the major problems in expanding coverage. Health insurance schemes in Thailand can be classified into three types: welfare and fringe benefit, compul- sory, and voluntary health insurance, as discussed below by Piyaratn and Jan- jaroen (1994) and Tangcharoensathien and Suphachutikul (1993). There are four health insurance schemes in the welfare and fringe benefit cate- gory: Civil Service Medical Benefit Scheme (CSMBS), Free Medical Care for the Low Income Household Scheme (FC/L), Free Medical Care for the Elderly Scheme (FC/E), and the School Health Insurance Scheme (SHI). The CSMBS, initiated in 1980, aims to provide medical care benefits to civil servants and employees, retired pensioners, and their dependents. Dependents include parents, spouses, and up to three children under 20 years of age. Benefits of this scheme include medical consultations, medical treatment, operations and other therapeutic care, drugs, inpatient care, and obstetric delivery expenses. The total population coverage was estimated to be 6.4 million, or 11 percent of the total Thai population in 1993. The FC/L was initiated in 1975 with the twin objectives of creating more equitable access to health services and improv- ing the health status of the poor. The target populations are single persons with income less than 2,000 baht (US$80) per month, 24,000 baht (US$960) per year, and married couples with income less than 33,600 baht (US$1,344) per year per person. The benefits are free medical services from public outlets and hospitals. The popula- tion coverage was 11.7 million, or 21 percent of the total population in 1993. 320 Health Financing for Poor People: Resource Mobilization and Risk Sharing The FC/E was implemented in 1992 with the aim of increasing accessibility to health services and improving the health status of the elderly. The target popula- tion is those citizens 60 years old and above who are not covered by other schemes. The benefits include outpatient and inpatient care provided at public facilities. The population coverage was 3.5 million, or 6 percent of the total population in 1993. Finally SHI, the school health insurance scheme, guarantees medical service to schoolchildren from grade 1 to grade 9 (around 6­14 years old). The popula- tion coverage in this scheme was 6.7 million, or 11.5 percent of the total popu- lation in 1993. The benefits cover outpatient and inpatient care at public service units. In some areas, dental services are also provided. Compulsory insurance consists of three insurance schemes: the Workmen Compensation Scheme (WCS), the Social Security Scheme (SSS), and Car Acci- dental Insurance (CAI). The WCS was introduced in 1974. The objectives of the scheme are to protect workers from illness, injuries, death, and disability caused by work or work-related conditions. The target population is employees in firms with more than 10 workers. Benefits include medical compensation for work- related illness and injuries, temporary and permanent disability benefits, sur- vivor's pension, funeral grants, and rehabilitation expenses. Population coverage in 1992 was 2.8 million, which was 5 percent of the total population. The SSS was enacted in 1990 and implemented in February 1991. The objec- tives are to protect workers from nonoccupational illness and injuries and to compensate for maternity, disability, and death. The target population is firms with more than 10 employees. Population coverage was 3.8 million, or 7 percent of the total population in 1993. The CAI scheme was implemented in 1992, with the main objectives of guar- anteeing medical treatment for victims of vehicular accidents. In theory or by law, every vehicle owner, including motorcycle owners, must have this insur- ance; in practice many are uninsured. Two health insurance schemes are voluntary: the Voluntary Health Insurance Scheme (VHIS) and Private Health Insurance (PHI). The PHI scheme was intro- duced in 1978 when the Thai Medical and Health Company Limited was estab- lished. The scheme's main objective was to improve security and provide better health care, often by combining life insurance and medical insurance for people in the upper-middle and high-income groups who can afford the premiums. Popula- tion coverage was only 0.6 million, or 1 percent of the total population in 1992. The last scheme is VHIS, commonly known as the Health Insurance Card Scheme, first introduced as the Health Card Project in 1983. The three main objectives of this scheme are to promote community development under the primary health care pro- gram, foster the rational use of health services via a referral system, and increase health resources based on a community-financing concept. The Voluntary Health Insurance Scheme has been continuously monitored and evaluated. Frequent adjust- ments of its strategies and objectives have included voluntary risk sharing with cost recovery in addition to service provision (see table 10.5). The target population is the near-poor and middle-income class in rural areas or those who can afford the premium. TABLE 10.5 Modification of Health Card Program: Rationale, Objectives, and Activities 1983 1984­1986 1987­1991 1992­present Rationale 1. MCH + FP PHC + Voluntary health (Community Voluntary health insurance Financing) insurance 1. MCH+FP 1. PHC To provide security To provide security 2. Referral system 2. Referral system to the people to the people 3. Integrated PHC Coverage all services services 4. MOPH decentralized decisionmaking to provincial level 5. Community Financing 6. To reduce OP visits at provincial hospital 1.1 Implementing 7 provinces 1. At each province At each province 68 provinces area 18 villages selecting 2 villages selecting 3 villages from one subdistrict from 8 subdistricts in 1. Khon Kaen 2. Chiang Mai (GTZ) each district 2. Roei In 1984 implementing 3. Lamphun in 4 districts, 6 4. Nakhon Sawan subdistricts, and 33 5. Phetchaburi villages 6. Ratchaburi In 1985 whole 7. Songkhla province 1.2 Coverage N/A 34% of population 36% of households 20% of population 55% of households 21% of households 1.3 Type and price Medical care and Family $8 Family $12 Family $20 of card MCH $8 Individual $8 (member < 5) Medical care $4 MCH $4 MCH $4 (MCH included) MCH $4 (Individual is an option) 1.4 Criteria for -- 8 episodes 6 episodes No limit card use (ceiling $80 per (ceiling $80 per episode) episode) 1.5 Level: health Village Village Village District card fund Province 1.6 Subsidy N/A N/A $13.50 per card $20 per card 1.7 Share of health -- Health service unit N/A Health service unit card fund 75% 80% Health center 15%, Incentive & adm. Community hospital 20% of total 30%, Provincial and monetary amount regional hospital of card sales 30% revenue (the Health personal formula differs 10% slightly from year Program managing to year) 15% Source: Modified from Singkaew (1995, pp. 19­21). 321 322 Health Financing for Poor People: Resource Mobilization and Risk Sharing At present, the price of a health insurance card is 1,000 baht (US$40) per year per household of not more than five members. A household contributes half of the price; the other half is subsidized by general tax revenue through the Ministry of Public Health. The benefits include outpatient care for illness and injuries, inpatient care, and mother and child health services. There is no limit on utilization of the ser- vices. The beneficiaries, however, can only go to health care provider units under the MOPH. The first contact is either the health center or community hospital; patients must then follow a referral line for higher levels of care. There is a spe- cific time for card selling at each health card cycle. At present the cycle is one year, and the specific time for card selling depends on the seasonal fluctuations in income. The premium is collected when cash incomes are highest, when, for example, crops are harvested. In 1992 the population coverage by the health card program was 3.6 million, or about 5 percent of the total population. Summary Information on the Health Card Program The Health Card Program, introduced in 1983, is based on the risk sharing of health expenditure with no cost sharing in a voluntary health insurance prepay- ment scheme. Ideally, in a prepayment-insurance scheme, members enroll when healthy, and only those who fall ill make use of services. The success of risk shar- ing in prepayment-insurance schemes depends upon enrolling a large enough number of people to ensure a sufficient pooling of risks. Mechanisms for community financing of health resources, one of HCP's three main objectives, include revolving drug funds, nutrition funds, and other funds, which at present are health card funds. The health card program was gradually implemented in rural areas and later, in 1985, started in some urban areas of six pilot provinces. The MOPH health facilities are responsible for providing care to health cardholders. During the HCP implemented nationwide before 1992, the MOPH decentralized decisionmaking to the provincial level, allowing health agencies to define their own prices for health cards and policies of disease cover- age, number of episodes, number of members, the level of compensation to providers, and the percentage of HCF kept in the community. The number of members covered by a card varies by provinces, but the premium is generally the same (Tangcharoensathien 1995). In 1995, there were many adjustments in the program. Some were that no HCF is kept in the community, the maximum num- ber of members per card is five, no limit is set on the number of episodes, and no ceiling is set on health care expense. The price of a card is fixed at 500 baht (US$20 in 1995). MOPH set a general formula for allocating HCF, that is, to com- pensate the providers for administrative costs and for incentives to sell cards, but kept decentralized decisionmaking at the provincial levels, allowing those levels to design details of HCF allocation. In 1985­86, there was strong support from MOPH for the Health Card Program. In 1986, the program was described to the parliament as a key strategy toward a voluntary health insurance. Unfortunately, Impact of the Thailand Health Card 323 HCP was a low priority for MOPH support from 1989 until 1995. This reflects the overall HCP situation during 1988­1992. In 1990, the area coverage rate in terms of population by HCP was 4.49 percent in 1988 and decreased to 2.6 percent in 1992. HCP covered 20.3 percent of villages in 1988 and increased to 33.6 percent in 1992; 48.2 percent of subdistricts in 1988 decreased to 30.2 percent in 1992; covered 79.7 percent of districts in 1988 and decreased to 58.8 percent in 1992 (Health Insurance Office report 1992; Kiranandana and Apairatr 1990). Assessments and Studies of HCP Demand for health cards has changed over time due to various factors in both the demand and supply sides, particularly government health policies. There are likewise many studies on the performance of the program. The Health Card Program in Chiang Mai is the pilot phase of a national health insurance system in Thailand. In 1985, 1986, and 1988, the pilot project for the Health Card Program and Social Research Institute (SRI) of Chiang Mai University conducted three household surveys. The surveys investigated house- hold health care expenditure and utilization patterns, health card knowledge, attitudes of the target groups regarding the health card system, and fund man- agement. In 1988, Adeyi studied the Health Card Program in Chiang Mai and found three main factors affecting health card purchase: the expectations of free treatment, reduced waiting time at referral centers, and subsidized drugs. Since the present program has many changes, some of these factors may no longer affect the health card purchase. The main reason for buying a health card has been examined in various studies. Kiranandana (1990) evaluated the HCP situation using a national census of all HCFs and health facilities and found that the main reason was not to ensure future cover- age for the possible illness of a household member, but either awareness that a fam- ily member was currently becoming sick or influence from others. In 1994, Veeravongs analyzed the HCP in the Phuket Province and found that the influenc- ing factors of health card purchase were coverage by alternative health insurance, accessibility to health care, card purchaser's place of origin and sex, age of the house- hold head, level of education of the household head and religious group affiliation, marital status of the household head, number of household members, number of sick household members, and problems experienced with health expenditure. In summary, the reasons to buy a card are economic conditions, benefits of the pro- gram, family size, presence of illness in the family, expectation of received health care, and the way the program is publicized (Hongvivatana and others 1986; San- tampol 1990; Supakankunti, Janjaroen, and Sritamma 1996). When HCP has been implemented in its current revised form (1995 to date), a research question arises: What are the current reasons for a household to buy the health card? These prob- lems of moral hazard and adverse selection are also found in other countries with similar programs. For example, in Burundi, if insurance is sold for short periods to accommodate families' fluctuations in income, then people may buy the "health 324 Health Financing for Poor People: Resource Mobilization and Risk Sharing card" entitling them to services only when they are already sick or can anticipate a medical need (McPake, Hanson and Mills 1992). The reasons not to buy a card have also been studied and identified as follows: trend of the number of card users decreases because people do not have enough money to pay for the card, because of the health workers or the inefficiency of the referral system, because they do not receive enough information about the card, because they have alternative health insurance, never used a previous card, no one sells the card, or it is difficult to use card (Manopimoke 1993). The attitude of households toward the program is a main factor for program sustainability. Hongvivatana and Manopimoke (1991) found that more than 50 percent of the people knew about the benefits of the program. When there were changes in the price of the card, including the benefits and criteria for card use in 1993, more than 50 percent of the people knew about the changes. Nonethe- less, between card users and nonusers there was a significant difference, with the card user group having better knowledge than the nonuser group. Behavior of card users had been examined in various studies, but there is no clear evidence of the changes in behavior before and after the implementation of the program. In addition, the episodes of card use and the referral systems have been studied in various ways. Even though the health center and commu- nity hospital can screen patients before referring them to a higher level, Kiranan- dana and Apairatr (1990) found that more than 50 percent of the patients at community and regional hospitals did not follow the referral system. The health center bypass rate (without a referral letter) was 52 percent for outpatients (OPs) and 40 percent for inpatients (IPs), respectively. Health centers could screen 62 percent of the OP consultations (38 percent were referred), while community hospitals could screen up to 97 percent of the OP visits, with only 3 percent referred. These findings are consistent with those in other studies, which have found that people still do not follow the referral system. The sustainability of the program also depends on program financial viability, which is in turn dependent upon relatively stable and adequate demand to ensure a sufficient pooling of risks. Therefore the cost of providing health care at all levels, usage rates, and utilization patterns of cardholders, apart from pur- chase patterns, have been studied. Different unit costs in different regions and years from several studies are shown in table 10.6. Cost recovery also has been studied in various regions and years. The cost recov- ery for community and provincial hospitals is still quite low. Permpoonwatanasuk (1985) looked at cost recovery in the Ratchaburi Province and found it to be 32.62­47.76 percent; while in the Health Card Program in Chiang Mai (1989) it was 31 and 39 percent in 1987 and 1988, respectively (only material cost). Kiranan- dana and Apairatr (1990) found that cost recovery for drugs was 40 percent for community hospitals and 46 percent for provincial hospitals. Manopimoke (1993) found that for outpatients at community and provincial hospitals cost recovery was 60 percent and 59 percent, respectively, for an average of 70 percent. The inpa- tient average was 54 percent, and cost recovery at health centers was 108 percent. Impact of the Thailand Health Card 325 TABLE 10.6 Unit Cost at Health Service Unit Used by Health Card Member (Unit: Baht) Health Service Units Case study Index Health center Com. hosp. Prov. hosp. Reg. hosp. Ratchaburi Cost/outpatient visit 23.61 77.35 86.11 -- (1985) Cost/inpatient day -- 283.85 466.25 -- Ubon- Cost/outpatient 10.11a 69.31d -- 137.34 Ratchathani 7.17b 60.01e -- -- (1984) 15.09c 63.10f -- -- Cost/inpatient -- 733.99d -- 1,431.34 -- 419.04e -- -- -- 556.09f -- -- Chiang Mai* Cost/visit 23.85 66.12 52.62 -- (1987) Cost/admission -- 364.15 724.99 -- Cost/green card/year outpatient 74.87 112.40 4.21 -- inpatient -- 29.13 28.96 -- Chiang Mai** Cost/visit 26.11 67.27 50.39 -- (1987) Cost/green card/year outpatient 65.28 107.63 9.07 -- inpatient -- 27.30 77.84 -- (1988) Cost/visit 30.44 64.60 76.01 -- Cost/green card/year outpatient 32.27 38.76 28.88 -- inpatient -- 39.28 91.21 -- Chiang Mai*** Cost/outpatient 100 182 212 -- (1996) Cost/outpatient visit 34 96 118 -- Cost/inpatient -- 1,100 1,102 -- Cost/inpatient day -- 303 522 -- 9 provinces Cost/person/year 9.98 34.27 17.84 17.74 (samples) Cost/card/year 36.71 114.84 79.09 79.09 Note: Districts in Ubon Ratchathani: a. Sa Tuek, b. Non Rung, c. Som Sa Ard, d. Tra Karn Puech Phol, e. Khueng Nai, f. Dech Udom. Sources: Ratchaburi: Chaisak Permpoonwatanasuk (1985); Ubon Ratchathani: Penjant Pradabmuk (1984); Chiang Mai*, Thai-German Technical Cooperation for Health (1987); Chiang Mai**, Health Card Program in Chiang Mai (1989); Chiang Mai***, Pravuth Vechrak (1996); sample of 9 provinces: Vachiraphan Chantamars (1989). The card usage rate from the Thai-German Technical Cooperation for Health (1987) found that the averages of outpatient visits per case at health centers, com- munity hospitals, and provincial hospitals were 0.5, 0.38, and 0.12, respectively. Kiranandana and Apairatr (1990) found on average, in a health card fund, 207 out- patient visits were made, 44 percent at a health center, 33 percent at a community hospital, and 23 percent at a provincial hospital (nine inpatients were admitted to the hospitals). Thus the health card usage rate was 6.63 services per card per year. The success of the program depends on several factors implicit in health ser- vice utilization and health card purchasing patterns, as well as government pol- icy: personal health management, the effectiveness of HCP, and community 326 Health Financing for Poor People: Resource Mobilization and Risk Sharing participation. A review of this scheme identified four critical factors for success: (1) strong government support, (2) managerial capability and supervision by local health staff, (3) strong community involvement and capacity, and (4) administra- tive simplicity of the scheme (Wibulpolprasert 1991 cited in Kutzin 1995). Before 1995, the main difference between HCP and other health insurance programs was that the financial management of the HCP was in the hands of responsible com- mittees at village level, under the supervision of health workers. The committee could manage revenue collected as a revolving fund for income-generating activi- ties initiated by villagers to promote PHC (Veeravongs 1994). According to the records, some villages had success in managing the fund, while some encoun- tered many problems. Therefore this was removed under the most recent form of the HCP. The committee is now at the district and provincial levels and cannot use the fund to generate income. In addition to this change, the price and type of card and the criteria for card use have changed. Therefore this study is different from previous studies, which analyzed the former program. METHODS Study Area Khon Kaen Province has had, and has developed, experience with voluntary health insurance since 1983. In 1995, Thailand had 76 provinces in total. Khon Kaen is a large province located in the northeastern part of the country. In 1995, it had a population of 1,652,030, population rank per total was 4/76, and popu- lation per square kilometer was 152. The population age structure in the year 2000 has been estimated for age groups 0­9, 10­19, 20­39, 40­59, and 60+ as 19.0, 19.1, 37.0, 18.0, and 6.9, respectively. Its area accounts for 10,886 square kilometers and area rank per total was 15/76. For the economic status, Khon Kaen's gross provincial product (GPP) in 1993 was 38,688 million baht. GPP rank per total was 12/73, and GPP per capita was 23,519 baht. The industry that has the highest share in the GPP is the wholesale and retail trade sector, with growth in the transport, communication, and service sectors, and a decrease in the agri- culture sector. Khon Kaen was designated a major city in the northeastern part of Thailand along the country development plan. In the province there are a regional university and several public health facilities: 1 regional hospital, 7 spe- cialist centers or health services, 19 community hospitals, 212 health centers, and 1 municipality health center. There are also many private clinics and hospitals. The card project involved in this study was implemented in six districts in Khon Kaen Province. The target population was identified by the research team and the provincial health office. The provincial and district health officers and the research team went to the six districts to explain the program to the communities and to inves- tigate the readiness of the communities. A sample of 1,000 households from the target population were selected by health officers. In this study, investigations were con- ducted first to examine differing characteristics of card purchasers and nonusers as Impact of the Thailand Health Card 327 well as of card dropouts and continuing card users. There are four groups of house- holds in the sample: (1) individuals who had not purchased a health card during the period 1993­1995, or card nonpurchase; (2) individuals who had purchased a health card for the first time in 1995, or new card purchase; (3) individuals who had repur- chased a health card, or continued card purchase; and (4) individuals who had not repurchased a health card, or health card dropouts. Further examination of the attitudes toward the health card program of card and non­card users at health centers and community hospitals will be carried out, using information from the questionnaires. Data Sources This chapter is based in part on the data collected from the research project on voluntary health insurance in Khon Kaen Province. The project implemented in Khon Kaen was a pilot project, the main objective of which was to provide health insurance for the uninsured in the province to achieve universal cover- age. The study period was 1994­95. The unit of study was households in selected rural areas. Health officers at both provincial and district levels were trained to conduct field interviews in six districts (study areas). There were four types of questionnaires used in the study: · Interview questionnaire of subdistrict and village leaders and volunteer health workers · Interview questionnaire of households in the sample areas · Interview questionnaire about the attitude toward the program of card and non­card users · Cost data obtained from the health center and community hospital. In addition to these primary data, there is secondary data, namely the statistics of card-usage rate at all levels, utilization rates, retrospective reimbursement from providers, and the number of insured and uninsured in the province before and after the implementation of the program, by type of insurance schemes. Data from the reports from each district in which the card project was imple- mented are also used in the study. The primary data of households and health facilities consists of: · Socioeconomic information, that is, number of household members, marital status, gender, age, number of dependents, number of unemployed members, use of alternative health insurance, education, occupation, income, type of expenditure and income, problem of health care payment, presence of illness (in past three months), number of members having chronic illness · Health care seeking behavior, that is, choice of providers, type of communica- tion, health card knowledge, decision to buy health card · Health care services utilization, that is, health card utilization rate, hospital utilization rate 328 Health Financing for Poor People: Resource Mobilization and Risk Sharing · Source of health care financing, that is, out-of-pocket, relative, health insurance · Attitude of the target groups regarding the health card system · Utilization rate and reimbursement of health care expenditure from providers. Analytic Methods This study employed both qualitative and quantitative methods. The data were entered from the interview questionnaires into coding form, and verification of the database was carried out using spreadsheet software. Data processing involved use of the statistical package for Social Science (SPSS/PC+, Marija J. Norusis/SPSS Inc., 1993). The analysis of the data involved the investigation of the factors influencing health card purchases and dropout patterns including continuity of card use as well as attitude of card users. The nonparametric test, that is, Chi-square statistic and parametric t-test statistic, were used initially to identify the various variables in both socioeconomic and psychological factors. These significant factors pro- vide more understanding of demographic and socioeconomic patterns between card purchase and nonpurchase groups, and of subsequent repurchase and non- repurchase groups. The logistic regression model was then used to identify sig- nificant predictors of health card purchase and nonpurchase patterns as well as the continuation of card purchase. Card utilization patterns and attitudes toward the health card will also be analyzed for a better understanding of the future prospect of voluntary health insurance in Thailand. Results The total number of response households was 1,005. The number of households that reported having not purchased a health card at any time over the 1993 to 1995 period was N = 495 (49.3 percent) as non­card purchase and N = 510 (50.7 percent) as card purchase. Of the 510 cards purchased, N = 297 (58.2 percent) as new card purchase, N = 132 (25.9 percent) as continued card purchase, and N = 81 (15.9 percent) as health card dropout (no longer cardholders). Description of the Data The demographic, socioeconomic, and cultural characteristics of card purchase and nonpurchase, and of subsequent repurchase groups are shown in tables 10.7­10.10. Results shown in the tables may be interpreted in different ways, since the program has undergone many changes, for example, in price and type of card, card usage criteria, and financial management, as stated above. Some characteristics between the health card nonpurchase and new card purchase groups are different (see table 10.7). The results show that compared with the nonpurchase group, those in the health card new purchase group were older, Impact of the Thailand Health Card 329 TABLE 10.7 Demographic and Socioeconomic Characteristics of Health Card New Purchase and Health Card Nonpurchase Demographic Socioeconomic Characteristics Health Card Nonpurchase (495) Health Card New Purchase (297) Marital status Single/widowed/divorced 2.4 2.7 Married 97.6 97.3 Educational level ** Lower 6 grade 87.1 93.9 Higher 12.9 6.1 Number of household members Under 5 66.9 70.2 6 and above 33.1 29.8 (average) (5.03) (5.01) Average number of employed members 2.32 2.83 Average household income/baht/year 60,253.17 50,580.00 Presence of illness*** Yes 47.5 57.2 No 52.5 42.8 Number of sick members with chronic illness None 78.4 78.5 1­2 20.6 21.5 More than 2 1.0 0.0 Family economic problems during sickness of family members Have problems 13.6 17.5 Never have problems 86.4 82.5 Note: *p<.05, **p<.01, and ***p<.001 had a higher average number of employed members in the family, had a higher percentage of presence of illness, had a higher percentage of family economic problems during sickness of family members, and had a lower level of education and household income per year. Only two of these factors are significant, namely educational level and presence of illness. In addition, among the health card dropout group and continued health card purchase group, the results show that compared with the health card dropout group, the continued card purchase group tended to get married at a higher rate, have a lower educational level, have a higher average number of employed members in the family, have a smaller number of household members, have a higher household income per year, have more members with chronic illness, have a higher percentage of presence of illness, have a higher percentage of hav- ing family economic problems during sickness of family members, and have a higher percentage of card use (see table 10.8). But only one factor is significant-- presence of illness. Interestingly, these findings are consistent with the above comparison between health card nonpurchase and health card new purchase 330 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 10.8 Demographic and Socioeconomic Characteristics of Health Card Dropout and Continued Card Purchase Demographic socioeconomic characteristics Health card dropout (81) Continued card purchase (132) Marital status Single/widowed/divorced 4.9 3.1 Married 95.1 96.9 Educational level Lower 6 grade 87.5 93.1 Higher 12.5 6.9 Number of household members Under 5 63.0 71.8 6 and above 37.0 28.2 (average) (5.15) (4.86) Average number of employed members 2.52 2.88 Average household income/baht/year 44,964.35 52,207.26 Presence of illness** Yes 48.1 66.7 No 51.9 33.3 Number of sick members with chronic illness None 82.7 77.3 1­2 16.0 19.7 More than 2 1.2 3.0 Family economic problems during sickness of family members Have problems 13.5 15.4 Never have problems 86.5 84.6 Note: *p<.05, **p<.01, and ***p<.001 groups except for the data on income per year, marital status, and number of sick members with chronic illness. Between the health card nonpurchase group and the health card dropout group, the results show that compared with the health card nonpurchase group, the health card dropout group had a lower income per year and a lower number of sick members with chronic illness (see table 10.9). As expected, there were no significant differences between these groups; both groups had similar character- istics, and they did not purchase the card during the current year. It is important to note that the association between chronic illness and health card purchase or insurers has been previously observed. For example, Hongvi- vatana and others (1986) found that there was significantly more chronic illness in a health card user family, as did Veeravongs (1994). But in this study even though there were differences--a higher percentage of sick members with chronic illness in the new health card purchase group than in the nonpurchase group, likewise a higher percentage in the continued card purchase group than in the health card dropout group, and also higher in the health card nonpur- Impact of the Thailand Health Card 331 TABLE 10.9 Demographic and Socioeconomic Characteristics of Health Card Nonpurchase and Dropout Groups Demographic socioeconomic characteristics Health card non-purchase (495) Health card dropout (81) Marital status Single/widowed/divorced 2.4 4.9 Married 97.6 95.1 Educational level Lower 6 grade 87.1 87.5 Higher 12.9 12.5 Number of household members Under 5 66.9 63.0 6 and above 33.1 37.0 (average) (5.03) (5.15) Average number of employed members 2.32 2.52 Average household income/baht/year 60,253.17 44,964.35 Presence of illness Yes 47.5 48.1 No 52.5 51.9 Number of sick members with chronic illness None 78.4 82.7 1­2 20.6 16.0 More than 2 1.0 1.2 Family economic problems during sickness of family members Have problems 13.6 13.5 Never have problems 86.4 86.5 Note: *p<.05, **p<.01, and ***p<.001 chase group when compared with the health card dropout group--there were no significant differences between any of these groups. After testing for the mean of numbers of chronic illness in the family between the health card nonpurchase group and the combined other three groups, no significant difference was found. For further investigation of the factors influencing the purchase pattern, the response households were divided into two groups: the nonpurchase (never pur- chase) group and the purchase either current or continued purchase group. In table 10.10, the results show that when the nonpurchase group was compared with the purchase (currently purchase, continued, and dropout) group, the pur- chase group was older, had a higher average number of employed members in the family, had a higher percentage of presence of illness, had a higher percent- age of family economic problems during sickness of family members, and had a lower average income per year and educational level. In terms of access to health facilities, the results show that the purchase group had more convenient access to a community hospital when needed than the nonpurchase group, and no dif- ferent access for other places, such as health centers, private clinics, or hospitals. 332 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 10.10 Demographic and Socioeconomic Characteristics of Health Card Nonpurchase and Health Card Purchase (Dropout/Continued/New Purchase Groups) Demographic socioeconomic characteristics Health card nonpurchase (495) Health card purchase (510) Marital status Single/widowed/divorced 2.4 3.2 Married 97.6 96.8 Educational level** Lower 6 grade 87.1 92.7 Higher 12.9 7.3 Number of household members Under 5 66.9 69.4 6 and above 33.1 30.6 (average) (5.03) (4.99) Average number of employed member*** 2.32 2.80 Average household income/baht/year** 60,253.17 50,099.12 Presence of illness*** Yes 47.5 58.2 No 52.5 41.8 Number of sick members with chronic illness None 78.4 78.8 1­2 20.6 20.2 More than 2 1.0 1.0 (average) (0.23) (0.22) Family economic problems during sickness of family members Have problems 13.6 16.3 Never have problems 86.4 83.7 Note: *p<.05, **p<.01, and ***p<.001 Only four of these factors are significant, namely educational level, income per year, number of employed members in the family, and presence of illness. Interestingly, the significant factors between the health card nonpurchase group and the health card new purchase group are educational level and presence of illness. The latter factor also was significant between the health card dropout and continued card purchase group and between the purchase and nonpurchase groups. As expected, it was not significant between the health card nonpurchase and the health card dropout group. This strongly confirms the problem of adverse selection among health card purchase and nonpurchase groups. The family with presence of illness tended to purchase and repurchase the health card. This crucial finding indicates a factor that will jeopardize the sustainability of the HCP in the future if the program continues to be implemented without improvement. Some of the psychological factors affecting card use were found to be particu- larly related to the continued and dropout groups. The continued card purchase group reported greater knowledge regarding the referral system, greater ability to Impact of the Thailand Health Card 333 seek treatment, greater convenience in buying a card, greater use of the health card, and greater satisfaction in using a card. They were more likely to have been per- suaded by a health officer rather than a HCF committee or village leader to buy a health card. By receiving a clear explanation about the card in advance from a health officer, they were able to make a decision to buy a card immediately. They had better access to health centers, while those in the card dropout group had bet- ter access to a community hospital. Finally, they wanted to buy a card, although no one persuaded them when compared with the dropout group, which has a higher percentage of having been persuaded to buy a card at home. But only a few factors are significant: convenient access to health care, persuasion by a health officer to buy a card, reception of clear advance information about the health card from a health officer, health card usage in the last year, satisfaction with the health card, and persuasion by a neighbor to buy a card. These factors clearly demonstrate how to make cardholders continue to buy a card. The effort needs active health officers who will explain about the health card, but it also depends on how convenient health care access is for people, and what card use experience they have had. Logistic Regression Model In the first part of the analysis, the factors influencing card purchase and non- purchase were identified by using t-test and Chi-square analyses. The analysis was performed separately for the following four pairs: · The nonpurchase and purchase group · The health card dropout and continued card purchase group · The nonpurchase and new card purchase group · The nonpurchase and health card dropout group. Only five significant demographic and socioeconomic factors were identified, namely, educational background, number of employed members in the family, household income per year, presence of illness, and convenient access to health care--that is, health center, community hospital, private health facilities, provincial hospital. These factors were included in the logistic regression models to estimate their relationships to the following health card purchase patterns: (1) the nonpurchase versus purchase (dropout/continued/new purchase) group, (2) the nonpurchase and health card dropout group versus continued and new card purchase group, (3) the nonpurchase versus new card purchase group, (4) the health card dropout versus continued card purchase group, and (5) the new card purchase and continued card purchase versus health card dropout group. In models 4 and 5, nine more factors were added to estimate the pattern of card purchase. The additional factors added were who persuaded the cardholder to buy a card (health officer, village health volunteer, village leader); who gave a clear explanation of the health card; health card usage; satisfaction with the health card; and persuasion by a neighbor to buy a health card. 334 Health Financing for Poor People: Resource Mobilization and Risk Sharing The following factors were used to estimate the various models of the logistic regressions to identify the best set of predictors for each model: Marital status · Single/widowed/divorced · Married Age in years · Under 40 · 41 and above Educational level · No school/primary · Secondary · Higher Proportion of employed persons to total family members Household income per year · Less than 33,600 baht · More than 33,600 baht Presence of illness Number of sick members with chronic illness in the family Family economic problems during sickness of family members The most convenient place to access health care · Health center · Community hospital · Private hospital/private clinic/drugstore · Other Who influenced card purchase · Village health volunteer · Health center personnel · Village head · Other Who explained clearly and in advance about health card program · Village health volunteer · Health center personnel · Village head · Other How the person persuaded cardholder to buy a card · Arranged meeting at village · Came to your home Impact of the Thailand Health Card 335 · None · Other Who made the decision to buy a health card · Household head · Spouse or household members · Card bought from whom · Village health volunteer · Health center personnel · Village head · Other Health card usage · Satisfaction with the card · Persuaded neighbors to buy a card The Coefficients of the Logistic Regression Model In logistic regression the probability of an event occurring, such as card pur- chase, can be directly estimated from the model. For the case of multiple inde- pendent variables, the logistic regression model can be written as (10.1) Prob (card purchase) = 1/1 + e­Z, where Z is the linear combination Z = B0 + B1 X1 + B2 X2 +....................+ BPXP. The probability of the event not occurring is estimated as Prob (not purchase) = 1 ­ Prob (purchase). Xi are the independent variables in the model, such as the demographic- socioeconomic factors or psychological factors among health card purchasers and nonpurchasers. The interpretation of the logistic regression coefficient is not straight- forward as in the regression model. The logistic model can be rewritten in terms of the odds of an event occurring. The odds of an event occurring are defined as the ratio of the probability that it will occur to the probability that it will not. The value of the coefficient for each variable indicates the changes in the log odds when the value of a particular variable changes by one unit and the values of the other inde- pendent variables remain the same (SPSS/PC+, Marija J. Norusis/SPSS Inc. 1993). DISCUSSION Results from the various models of the health card groups described above are interesting and should prove valuable from several aspects. Tables 10.11­10.14 show that the statistically significant factors distinguishing purchase groups (dropout, continued, and newly purchase groups) from nonpurchase groups are educational level, proportion of employed persons to total family members, 336 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 10.11 Variables that Predict People Who Ever Purchased a Health Card Versus Nonpurchase Group Variable Coefficient Standard error Significance Exp. ( ) odds ratio Marital status Single 0.7467 1.4994 0.6185 2.1100 Married 0.4360 1.4416 0.7623 1.5465 Widowed 0.3409 1.4984 0.8200 1.4062 Separated ­0.4263 2.0675 0.8367 0.6530 Age ­0.0003 0.0079 0.9727 0.6530 Gender 0.0888 0.1662 0.5934 1.0928 Educational level Primary 2.0272 0.8406 0.0159 7.7924 Secondary 1.5163 0.8711 0.0817 4.5552 Proportion of employed in household ­0.7015 0.1565 0.0000 0.4959 Household income Quintile 1 0.6477 0.2435 0.0078 1.9111 Quintile 2 0.3913 0.2355 0.0966 1.4789 Quintile 3 0.6207 0.2350 0.0082 1.8602 Quintile 4 0.5828 0.2305 0.0115 1.7910 Presence of illness in the past 3 months 0.4545 0.1515 0.0027 1.5754 Number of sick with chronic illness None 0.6365 0.6829 0.3513 1.8898 1­2 0.6049 0.6984 0.3864 1.8311 Economic problems during sickness 0.0329 0.2112 0.8763 1.0334 Most convenient access to health care Health center ­0.6580 0.2602 0.0115 0.5179 Community hospital ­0.7958 0.3046 0.0090 0.4512 Provincial hospital 0.0801 1.1992 0.9467 1.0834 Drugstore 0.2289 0.5545 0.6797 1.2572 Private hospital ­1.2076 1.4773 0.4137 0.2989 Private clinic ­2.0717 0.6338 0.0011 0.1260 University hospital ­5.3682 5.8082 0.3554 0.0047 Constant ­2.6297 1.8332 0.1514 household income, presence of illness, and a convenient community hospital. Interestingly, the significant differences related to purchase patterns in the cur- rent year between currently nonpurchase (never purchase and dropout groups) and currently purchase (continued and new card purchase) are the first four as well, but the fifth, the most convenient place to access health care, is no longer significant in the model. This is consistent with the model of card purchase between the nonpurchase and new purchase groups. The finding clearly and strongly demonstrates that health card purchase in Khon Kaen in the current year is influenced by the following four factors: proportion of employed to total in family, education, household income, and presence of illness. The last factor Impact of the Thailand Health Card 337 TABLE 10.12 Variables that Predict Health Card Nonpurchase and Dropout Groups Versus Continued and New Card Purchase Groups Variable Coefficient Standard error Significance Exp. () odds ratio Marital status Single 0.3945 1.4957 0.7920 1.4836 Married 0.3018 1.4401 0.8340 1.3523 Widowed 0.6496 1.4972 0.6644 1.9148 Separated ­0.3808 2.0320 0.8514 0.6833 Age ­0.0102 0.0080 0.2029 0.9898 Gender 0.1677 0.1681 0.3186 1.1826 Educational level Primary 1.7220 0.8331 0.0387 5.5957 Secondary 0.8842 0.8698 0.3093 2.4210 Proportion of employed in household ­0.6352 0.1563 0.0000 0.5298 Household income Quintile 1 0.3476 0.2469 0.1591 1.4157 Quintile 2 0.3002 0.2397 0.2105 1.3501 Quintile 3 0.5018 0.2370 0.0342 1.6517 Quintile 4 0.7239 0.2323 0.0018 2.0625 Presence of illness in the past 3 months 0.4972 0.1524 0.0011 1.6442 Number of sick with chronic illness None 0.6449 0.6863 0.3473 1.9059 1­2 0.8206 0.7020 0.2424 2.2718 Economic problems during sickness ­0.0054 0.2108 0.9797 0.9946 Most convenient access to health care Health center ­0.6941 0.2473 0.1640 0.7088 Community hospital ­0.3441 0.2983 0.0200 0.4995 Provincial hospital ­6.5534 11.0855 0.5544 0.0014 Drugstore 0.5353 0.5292 0.3117 1.7080 Private hospital ­0.7154 1.4422 0.6199 0.4890 Private clinic ­1.4901 0.6292 0.0179 0.2254 University hospital ­5.8800 9.4998 0.5359 0.0028 Constant ­2.4282 1.8287 0.1842 demonstrates the problem of adverse selection in the program, particularly sig- nificant since the health card program was introduced in the country in 1983, and Khon Kaen was one of seven provinces in the implementing areas in that year, has thus had experience with voluntary health insurance, and has under- gone development of its program throughout the years. The various previous studies of card purchase patterns showed the important fac- tors for card purchase to be gender, age, chronic illness, alternative health insurance, family size, income, and health service satisfaction. Veeravongs (1994) found that females tended to purchase health cards more than males because this related to greater maternal and child health care card use, and that there was also an association 338 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 10.13 Variables that Predict Health Card Nonpurchase Versus New Card Purchase Group Variable Coefficient Standard error Significance Exp. () odds ratio Marital status Single 0.0611 1.5223 0.9680 1.0630 Married ­0.1779 1.4451 0.9020 0.8370 Widowed 0.2639 1.5052 0.8608 1.3019 Separated ­7.1747 36.6851 0.8449 0.0008 Age ­0.0065 0.0093 0.4838 0.9935 Gender 0.3009 0.2000 0.1324 1.3511 Educational level Primary 2.2894 1.1173 0.0405 9.8693 Secondary 1.5669 1.1540 0.1745 4.7918 Proportion of employed in household ­0.5981 0.1781 0.0008 0.5494 Household income Quintile 1 0.4567 0.2846 0.1085 1.5789 Quintile 2 0.2161 0.2749 0.4318 1.2412 Quintile 3 0.6155 0.2658 0.0206 1.8505 Quintile 4 0.4982 0.2655 0.0607 1.6457 Presence of illness in the past 3 months 0.4498 0.1744 0.0099 1.5680 Number of sick with chronic illness None 6.8014 16.1808 0.6742 899.0828 1­2 6.8769 16.1819 0.6709 969.5739 Economic problems during sickness 0.0405 0.2403 0.8660 1.0414 Most convenient access to health care Health center ­0.4297 0.3097 0.1653 0.6507 Community hospital ­0.4881 0.3571 0.1717 0.6138 Provincial hospital ­7.2429 36.6599 0.8434 0.0007 Drugstore 0.6634 0.5983 0.2675 1.9415 Private hospital ­7.2481 36.6596 0.8433 0.0007 Private clinic ­1.5246 0.7145 0.0329 0.2177 University hospital ­6.7581 15.5401 0.6636 0.0012 Constant ­9.0499 16.2870 0.5784 between gender and chronic illness. Another important factor is family size, which some researchers have indicated (Veeravongs 1994; Hongvivatana and Manopi- moke 1991; Suwanteerangkul 1992). This was not confirmed in this study since many aspects of the program have been changed. The gender and family-size fac- tors were not statistically significant, perhaps because the average Thai family size is five persons and the sample shows, on average, household size is also five. More important, the health card is now a household card, which allows five members per card. The family-size factor was also not significant between groups, as tested above. Therefore the proportion of employed persons to total family members fac- tor was selected in estimating the model to reflect the dependency among the fam- ily members. Interestingly, it was one of the significant factors found in this study. Impact of the Thailand Health Card 339 TABLE 10.14 Variables that Predict New and Continued Health Card Purchase Versus Health Card Dropout Group Variable Coefficient Standard error Significance Exp. () odds ratio Marital status Single ­5.8711 60.4411 0.9226 0.0028 Married ­4.9461 60.4371 0.9348 0.0071 Widowed 1.4894 62.3793 0.9810 4.4344 Separated 0.7680 85.4692 0.9928 2.1555 Age ­0.0533 0.0182 0.0034 0.9481 Gender ­0.2368 0.3581 0.5084 0.7892 Educational level Primary ­6.5208 60.4342 0.9141 0.0015 Secondary ­7.9935 60.4356 0.8948 0.0003 Proportion of employed in household ­0.4364 0.3287 0.1843 0.6464 Household income Quintile 1 ­0.8185 0.4989 0.1008 0.4411 Quintile 2 ­0.0655 0.5323 0.9020 0.9366 Quintile 3 ­0.1471 0.4944 0.7660 0.8632 Quintile 4 0.5668 0.5491 0.3020 1.7626 Presence of illness in the past 3 months 0.1946 0.3395 0.5665 1.2149 Number of sick with chronic illness None 1.2969 1.3407 0.3334 3.6579 1­2 2.0872 1.3838 0.1315 8.0627 Economic problems during sickness ­0.3714 0.4572 0.4166 0.6898 Most convenient access to health care Health center 0.1308 0.4563 0.7744 1.1397 Community hospital ­0.4435 0.5730 0.4389 0.6418 Provincial hospital ­9.3719 42.7032 0.8263 0.0001 Drugstore 0.8040 1.1644 0.4899 2.2345 Private hospital 6.5944 29.1878 0.8213 730.9802 Recommendation to neighbors ­0.3269 0.4092 0.4243 0.7211 Purchase health card from HCF committee member or village health volunteer ­1.5333 0.6830 0.0248 0.2158 Heath center personnel ­0.4324 0.6372 0.4974 0.6490 Village head ­1.7454 0.8440 0.0386 0.1746 Health card usage 0.1527 0.3580 0.6696 1.1650 Who made decision to buy health card Household head 0.0110 0.4140 0.9787 1.0111 Spouse ­0.2649 0.4630 0.5673 0.7673 Constant 15.5619 85.4793 0.8555 340 Health Financing for Poor People: Resource Mobilization and Risk Sharing The households that had a higher proportion of employed persons tended to pur- chase more cards than the households with a lower proportion. This might be because the former can afford the price of the card, which must be prepaid, although the income factor was not of overall significance in this study. The pro- portion of employed to total in family factor might represent the income class, a higher proportion reflects the lower income class, which tended to buy health cards. This would also relate to alternative insurance schemes such as the Elderly Scheme (FC/E) and the School Health Insurance Scheme (SHI) since unemployed persons in the family might be eligible for these schemes. Moreover, most of the employed persons in rural areas are not covered by any of the health insurance schemes. As stated above, income was not shown to be a strong determinant of card purchase. This was confirmed by Veeravongs (1994). However, other studies have found that economic status was a significant indicator of the ability to pur- chase a health card (Hongvivatana and Manopimoke 1991). There was not much difference in income among households in this study, the subjects of which were rural residents, even though there were observable trends by income among the groups. For example, the dropout group tended to have incomes lower than the continued card purchase group. The other significant factor is education. Those with lower levels of education tended to purchase cards, since lower education means lower income and thus usu- ally not covered by any of the health insurance schemes. The only health insurance for which these persons are eligible is the health card program. The number of members with chronic illness in the family, marital status, age, and problems with health expenditures were not significant in determining card purchase in this study. The satisfaction with health services factor was indicated in previous studies as a determinant factor for card purchase, but in this study the factor was not sig- nificant. The results found little difference among households and found most to be satisfied with the services. This is discussed in detail below. It was difficult to make a comparison between studies since the studies were of dif- ferent areas and conducted at different times. Of most importance, however, are the differences in health card rationale, type and price of card, criteria for card use, health card fund management, government subsidies, and share of the health card fund. Continuity of Card Purchase The sustainability of the program depends on various factors, a very important one being is satisfaction of the card users to continue to buy. The findings indicate that the continuity of card purchase in the study was associated with these factors: mar- ital status, lower levels of education, number of employed in the family, income, presence of illness, problems with health expenditure, most convenient place to access health care, health card usage, the persuasion of a household to buy the card at home, health center personnel, and persuasion by a neighbor to buy a card. Of these, the significant factors were persuasion by a neighbor to buy a card, age, edu- cation, income, health center personnel explained clearly about card, and persua- sion of a household to buy the card at home. Impact of the Thailand Health Card 341 To analyze the satisfaction of card users with health care services requires more data; therefore, this study utilized the data from another set of question- naires distributed when the program had been implemented for one year. The sample is health care seekers at health center and community hospitals: posses- sion or nonpossession of a health card. They were interviewed about their atti- tudes toward the health card program when seeking care at health centers and community hospitals and about their health-seeking behaviors in the past three months. Below is the summary of demographic and socioeconomic characteris- tics between individuals who held cards and those who did not hold cards. Results in table 10.15 show that the statistically significant factors related to health card users and non­card users are education, number of household mem- bers, presence of illness, problems with health expenditure, and the most conve- nient place to access health care and seek care when sick. Card users tended to have a lower education, a lower average income per year, a lower health expen- diture per year--but in terms of the proportion of total expenditure to total income it is higher. Moreover, card users had not had many problems with health expenditure and had more members in the family, had more presence of TABLE 10.15 Demographic and Socioeconomic Characteristics of Health Care Seekers* by Whether or Not They Possess Health Card Demographic socioeconomic characteristics Health card (non-possession) (464) Health card (possession) (500) Marital status Single/widowed/divorced 15.1 12.8 Married 84.9 87.2 Educational level*** Lower 6 grade 79.6 87.8 Higher 20.4 12.2 Number of household members** Under 5 72.9 70.2 6 and above 27.1 29.8 (average) (4.78) (5.07) Gender Male 35.3 38.6 Female 64.7 61.4 Average household income/baht/year 63,453.84 58,572.21 Average household expenditure/baht/year 41,099.33 39,401.24 Average household health expenditure/baht/year 4,294.71 3,945.20 Average proportion of total expenditure to total income 0.7372 0.9477 Average proportion of health expenditure to total expenditure 0.1157 0.1123 (continued) 342 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 10.15 Continued Demographic socioeconomic characteristics Health card (non-possession) (464) Health card (possession) (500) Presence of illness** Yes 42.1 52.0 No 57.9 48.0 (average) (0.4213) (0.5200) Having chronic illness Yes 19.5 25.2 No 80.5 74.8 Family economic problems during sickness of family members** Have problems 38.4 30.2 Never have problems 61.6 69.8 Most convenient place to access health care Health center 75.1 81.0 Community hospital 16.5 15.2 Private clinic/hospital*** 8.0 2.8 Other 0.4 1.0 Type of chronic illness (86) (127) Diabetes 38.0 62.0 High blood pressure 58.8 41.2 Asthma/respiratory 39.4 60.6 Other 38.9 61.1 Seek care** Yes 69.9 78.6 No 30.1 21.4 Note: *Patients who visited health center or community hospital. Note: * p<.05, **p<.01, and ***p<.001 illness, had more chronic illness, sought more care when sick, and had a health center as the most convenience place of access. This shows that the health card eased their health expenditure burden, despite the fact that there was more ill- ness among card users. The proportion of health expenditure to total expendi- ture on average was not different when compared with that of non­card users. Nonetheless there remained a problem of adverse selection among card users. To improve the program, officials must know how often and what factors influence health card utilization. Table 10.16 shows that 69.4 percent of card- holders have used the cards 1­5 times, while only 4.2 percent never used it. In addition, 74.5 percent of card users tended to visit health centers while 23.8 per- cent with a common illness such as fever, cough, flu, and ulcer visited commu- nity hospitals. Interestingly, 94.8 percent were satisfied with health services, leaving only 5.2 percent not satisfied. Of those not satisfied, 63 percent had dis- satisfaction with the referral system and 28 percent with the quality of care. Of the sample, 90 percent will purchase a health card again next year. Since this is related to the problem of illness, they definitely will utilize health cards. Impact of the Thailand Health Card 343 The results in table 10.17 show that non­card users tended to receive no treatment when sick, have more self-prescription, and sought care at private clinics or hospitals, but sought care at health centers and public hospitals less than the card users. This health care seeking pattern among card users and non­card users strongly supports the importance of accessibility to health care among the card user group. However, it still leaves risk-sharing and card-utilization pattern problems to consider in making the program more sustainable and effi- cient. Therefore the health card utilization patterns will be studied in detail to provide more information on card usage. TABLE 10.16 Frequency of Card Use among Card User Group N=451 Card utilization rate Frequency Percent Never used 19 4.2 1­5 313 69.4 6­10 91 20.2 More than 10 28 6.2 TABLE 10.17 Health Care-Seeking Pattern in the Past 3 Months among Card Users and Noncard Users Percent of Health Number Drug Expenditure Care Seekers at of Visits (Baht) Different Facilities (average) (average) Card Non­card Card Non­card Card Non­card users users users users users users (500) (464) (500) (464) (500) (464) Sick but no treatment 16.4 18.6 -- -- -- -- Self-prescription 12.0 19.5 2.16 2.37 89.00 72.80 Sick and received treatment at private clinics and hospitals 5.8 8.2 2.47 2.11 252.86 241.94 Sick and received treatment at health center 50.0 41.5 2.49 2.35 61.02 47.06 Sick and received treatment at public hospitals 18.6 14.4 2.32 2.46 645.42 257.02 The main purposes of estimating health card utilization patterns were twofold: investigate the factors affecting the frequency of card use through a multiple regression model, and investigate the factors affecting the use or nonuse of health cards through a logistic regression model. The variables listed below were selected in estimating health card utilization patterns: · Marital status · Gender · Educational level 344 Health Financing for Poor People: Resource Mobilization and Risk Sharing · Number of household members · Proportion of total household expenditure to total income · Proportion of total household health expenditure to total expenditure · Presence of illness · Having chronic illness · Family economic problems during sickness of family members · Most convenient place to access health care. Results Multiple Regression Model The model in table 10.18 shows that the significant factors related to card usage rate are having chronic illness and convenient access to a health center. This is not surprising since the chronically ill will seek care regularly and usually at a health center located in the village near their homes, where they will use cards. This is related to the problem of adverse selection in the program. TABLE 10.18 Multiple Regression Model Variable B SE Beta T Sig T Chronic 1.622212 0.563359 0.156944 2.880 0.0043 Health center 1.915122 0.593073 0.175999 3.229 0.0014 Constant 2.613482 0.555544 4.704 0.0000 Logistic Regression Model The results from table 10.19 show that the significant factor related to card use is the presence of illness; the others are not statistically significant. This confirms once again the problem of adverse selection in the program since the one who has the presence of illness tended to buy a card and affected the card usage. In addition, other models were estimated to investigate the impact of having or not having a card to the health care seeking behavior among the households. Interestingly, the results show that possession or nonpossession of a health card was not a significant factor at the beginning of the program. But when the pro- gram had been implemented for one year, among health care seekers, possession of a health card was a significant reason to visit the health centers or community hospitals. This evidence strongly supports the importance for cardholders of accessibility to health care. Problems concern the risk sharing among the target population and card overutilization. The problem of card overutilization, con- firmed in this study, has implications for the sustainability and efficiency of the program. The results show that among card users 41.6 percent tended to visit health facilities more than before having a card, 48.4 percent the same as before, only 7.2 percent less than before, and 2.8 percent do not remember. Impact of the Thailand Health Card 345 TABLE 10.19 Logistic Regression Model: Variables that Predict Health Card Usage among Card Users Variable Coefficient Standard error Significance Exp. (B) odds ratio Marital status 0.0245 1.0413 0.9812 1.0248 Gender 0.3981 0.6898 0.5639 1.4889 Age ­0.0275 0.0325 0.3973 0.9728 Educational level 1.2836 0.9249 0.1652 3.6097 Number of family members 0.0262 0.1986 0.8950 1.0266 Proportion of total household expenditure to total income 0.0270 0.1618 0.8673 1.0274 Proportion of total household health expenditure to total expenditure 0.2636 2.7682 0.9241 1.3016 Presence of illness 1.8629 0.8417 0.0269 6.4422 Having chronic illness ­0.4791 0.8084 0.5534 0.6193 Family economic problems during sickness of family members 0.8375 0.8334 0.3160 2.3064 Most convenient place to access health care Health center 1.9755 1.4112 0.1616 7.2100 Community hospital 1.2323 1.4958 0.4100 3.4292 Constant 0.4165 1.8298 0.8200 Attitudes toward the Health Care Program among Cardholders and Noncardholders Among the health care seekers, cardholders and noncardholders at health cen- ters and community hospitals were interviewed about their attitudes toward the health card program. The main results are shown in tables 10.20­10.21. Cardholders had greater prior knowledge about the health card program before health officers explained it to them than did noncardholders, and they had more satisfaction with the explanations from health officers. This confirms the idea that knowledge of the health card was a strong determinant of card pur- chase. Even more important, the attitude about the usefulness of the health card to them or their family was different in the two groups. The cardholder group had greater satisfaction with the health card than the noncardholder group, and the cardholders tended to be satisfied with the price of the card more than were the noncardholders. This strongly explains the role of attitude. Cardholders said they would buy a new card next year, while noncardholders were not likely to buy. It is important to note that the decision to purchase a health card was not dependent solely on the price of the card but was also influenced by other factors, such as the quality of drugs received, the quality of medical services provided, 5 8.5 5.4 3.8 8.7 1.8 2.0 7.9 8.6 7.2 10.4 4 9.2 21.7 16.9 19.6 24.0 12.2 27.5 28.0 29.1 14.7 scale of 3 41.5 34.7 40.8 28.5 35.9 35.9 39.7 42.9 37.5 37.5 Noncardholders Level 2 24.0 37.8 31.5 26.9 38.3 38.1 19.9 17.8 18.7 25.3 1 4.3 5.2 4.3 5.0 2.7 4.3 11.9 14.8 11.7 15.3 5 1.0 0.4 0.4 0.4 0.4 1.4 15.4 12.4 1.0 14.7 4 4.7 2.2 0.2 2.2 6.9 12.4 37.1 32.8 2.0 38.5 scale of 3 31.2 17.6 29.6 12.8 29.1 30.2 27.8 40.7 27.7 22.2 Cardholders Level 2 44.2 64.1 55.8 53.0 44.7 42.6 15.4 11.0 15.5 33.5 Noncardholders and 1 4.3 3.1 3.7 11.2 13.2 12.0 33.5 23.6 18.9 41.2 Cardholders unsatisfactory). of card (very card using card using unlikely Program when with most when care 5 Card for benefits received level services time and Health and drugs of limits medical waiting toward HCP quality satisfactory), longer family appropriate HCP to is low (very appropriate substandard Attitudes explained about explanation card baht) possible year likely year the expected with (500 1 receive next most clearly health 1 10.20 for of card will about about card knowledge seller of Level buy ABLET orried orried orried Context Prior Card Satisfaction Benefits Price Coverage W W W Will Note: 346 Impact of the Thailand Health Card 347 TABLE 10.21 Attitudes toward Health Card Program of Cardholders and Noncardholders (Mean Score) Cardholders Noncardholders Context Mean score Mean score Prior knowledge about HCP 2.479 3.063 Card seller clearly explained HCP limits and benefits 2.150 2.795 Satisfaction with explanation 2.233 2.872 Benefits of health card to family 1.809 3.092 Price of card (500 baht) is appropriate 2.112 2.448 Coverage for 1 year appropriate 2.294 2.546 Worried will receive substandard medical services when using card 3.440 3.135 Worried about the possible longer waiting time for care when using card 3.138 3.219 Worried about expected low quality of drugs received with card 3.450 3.226 Will buy card next year 1.882 3.269 and whether the subjects had alternative health insurance. This study found that the problem of expectations about substandard medical services, low quality of drugs, and longer waiting time for care with cards were similar in the two groups, except that the noncardholders tended to worry more than the cardholders. Even though some studies have indicated that health service satisfaction is the most important factor influencing the use of health services and dissatisfac- tion among health card users was one significant factor influencing the dropout group (Adeyi 1988; Silapasuwan 1989; Suwanteerangkul 1992). This was not confirmed with the study by Veeravongs (1994), which found that both contin- uing card users and dropouts were satisfied with their current health service and there were no differences between them in terms of card knowledge obtained from village health volunteers and health card committee members. This study had similar findings as stated above. Actually, knowledge regarding the benefits of the program, and better knowledge about the principle of health insurance and the health system among the households are more important to the pro- gram in terms of the efficient use of resources, encouragement of households to participate in the program, and the sustainability of the program. In organizing a voluntary health insurance program there are important implications concerning the allocation of resources used and the technical effi- ciency with which they are used. These relate to the type of medical services pro- vided in the program, its referral system, the use of medical technologies, and mechanisms to allocate the resources. Some studies have shown similar prob- lems in organizing health insurance, particularly for rural populations. Eco- nomic pressure is key to bringing about health care financing reform in various countries, and one way of doing so is through the health insurance schemes, either compulsory or voluntary. It is important to realize that the existence of any scheme will be unstable if the scheme attracts a special subset of the popula- tion with unusually high health care costs or with greater sickness. 348 Health Financing for Poor People: Resource Mobilization and Risk Sharing Health Card Fund The current formula to allocate HCF to compensate providers and for administra- tive costs is set at 80 percent and 20 percent of the monetary amount of health card sales revenue (this differs slightly from year to year). The revenue is gener- ated by health card sales (US$20 per card) and subsidized US$20 per card from the government budget. After one year of implementation of the health card pro- gram in Khon Kaen, operating cost data at health centers and community hospi- tals in six districts were collected; including the utilization rate with these data will provide better understanding about the health card fund and its viability. At community hospitals, the unit cost per case for inpatients is 1,356 baht, and the unit cost per visit for outpatients is 168 baht. At health centers, the unit cost per case is 71 baht, and the unit cost per visit is 37 baht. When card usage rates in this study are compared with those in the previous study (Kiranandana and Apairatr, 1990), the rates per case for outpatient visits at health centers and community hospitals are higher by as much as 50 percent and 94 percent, respectively (see table 10.22). The rates per card for outpatients at health centers and community hospitals are also higher by 15 percent and 48 percent, respec- tively. Likewise, the card usage rate for inpatients is higher both per case and per card, but not by such a high percentage. Nonetheless, considering the diffi- culties of comparing the present study directly with other studies (they were conducted in different areas, times, and most importantly, under different rationales, types, card prices, and criteria for card usage), the average card usage rate per card for all services at health centers and community hospitals was not greatly different. TABLE 10.22 Health Card Utilization Rate Community Health center hospital Utilization rate Card Case Card Case Outpatient 3.2 0.72 3.11 0.70 Inpatient -- -- 0.17 0.04 Inpatient days -- -- 0.66 0.15 However, the results did show that under the new criteria for card use, with no limit on episodes and with the first contact at either a health center or com- munity hospital depending on convenient location, the rate of usage is increas- ing greatly, especially at community hospitals. This result in the major workload and cost of care is borne by community hospitals. The cost of care compared with unit cost in Chiang Mai in the same year revealed a similar pattern. The data show that providers, health centers, and community hospitals need to reimburse their expenses (based on price schedules rather than cost) in amounts greater than the health card fund committee can allocate from the fund, which Impact of the Thailand Health Card 349 is financed from the total monetary amount of card sales revenue alone. The col- lected data show that health centers and community hospitals can receive only 58 percent of their reimbursement (the government subsidy was not included in this statistic because there was a delayed subsidy from the government in 1996). The existing program formula is set at 80 percent of the total monetary amount of card sales revenue. If no subsidy from the government was allocated to the providers, the expense would be borne by providers. There is a problem of equity since only the health cardholders are better off if community hospitals must subsidize cardholders by the former's own revenues. It is essential that the formula of reimbursement from HCF to compensate providers should be developed. HCF should establish reimbursement agree- ments with public hospitals based on a combination of expected outputs and costs, and all agreements might contain either a cost or a volume ceiling. This policy must be developed for the country as a whole to assist the HCF commit- tee in each province in applying a standard criteria for allocation of funds. Program Sustainability In Khon Kaen Province there were a series of meetings during the year of this study between the research team, provincial health officers, and district health officers, including the health care providers in the province. In the meetings, there were discussions of how to organize the program, how to improve the pro- gram, problems in the field, government policy, budget, and how to sustain the program. The discussions give valuable suggestions for improving the existing program in areas such as program management, marketing management, finan- cial management, and quality of services. Program management: · Build up an effective organization at all provincial and district levels. Prob- lems may arise in the large district areas. · Make the health card project more prominent among the various ongoing projects being implemented in the province. · Improve the objectives and plan more efficiently. · Create positive attitudes toward the health card program among managers and officers at all levels of the health system. Encourage effective coordina- tion among health officers at all levels of the system to promote greater knowledge and better understanding of the health card program. This will help address the problem of providers being reluctant to participate in the program because of insufficient reimbursements. The public facilities under MOPH must participate, even though some of them are unwilling. This indi- cates the problem of coordination among health officers at all levels. · Make the health insurance information system more effective. 350 Health Financing for Poor People: Resource Mobilization and Risk Sharing Marketing: · Work toward a goal of 100 percent coverage in an appropriate way. · Improve marketing strategies, for example, using the social marketing concept. · Develop a serious and continuous advertising campaign for the health card program. · Create incentives to sell the health card. Financial management: · Establish monitoring committees at all levels. · Improve the regulation of monitoring committees. · Set penalty rules. · Abandon the installment payment for health cards. Quality of services: · Strengthen the health care service units to achieve a high quality of care. · Create good attitudes toward provider; provide services of equity and merit. These problems are similar to those in other countries where voluntary health insurance for rural people is implemented. Dave (1991) studied the community and self-financing in voluntary health programs in India and found that mixed success with financing efforts. The financing methods, including user charges, community-based prepayment schemes, fundraising, commercial schemes, and in-kind contributions can be further strengthened with better planning, man- agement, monitoring, and evaluation. Chabot, Boal, and da Silva (1991) studied national community health insurance at the village level and found that the experience of a voluntary levy scheme in Guinea-Bissau may be feasible and manageable in rural parts of Africa, if the village population is allowed to decide on the amount of money and method of collection and if the government sup- ports the scheme by guaranteeing sufficient drugs, low prices, and effective con- trol measures. A review of case studies carried out by the Bamako Initiative in five countries noted that the extension of insurance coverage to rural popula- tions is a strategy that can improve equity (McPake, Hanson, and Mills 1992). The main problem areas concerned in the program are as follows: · Program management: marketing, financial, quality of care · The adverse selection or self-selection problem, that is, how to increase the number of people enrolled to ensure a sufficient pooling of risks · Operational policies: choices about the type of services included in the pro- gram, number of episodes per card, reimbursement from providers, and effec- tive referral systems · Communications: providing people with greater knowledge and better under- standing of HCP, that is, sustainability (strong promotion of the program). Impact of the Thailand Health Card 351 Problems of program management are related to the effectiveness of health offi- cers, government health policies, and many adjustments in the program as stated above. The problem of risk sharing with no cost sharing needs a sufficient number of people or communities participating in the program. Households have been encouraged to participate in the program in various ways. The percentage of HCF kept in community (village): In 1984­1986, there were village committees responsible for selling the health cards and collecting pre- miums. These funds were kept at the village because a HCF with a minimum enrollment base of 35 percent of the households in each village was required to ensure risk sharing. These committees also manage revenue collected as a revolving fund for income-generating activities initiated by villagers to pro- mote PHC. A fixed rate of 75 percent of HCF is reimbursed to compensate the providers in the referral line at the end of the year. In FY 1987, premiums col- lected were 182.9 million baht and the approximate medical expenses were 267.5 million baht (Tantiserani and Prompakdi, 1988, cited in Tangcharoen- sathien 1995); the health card fund decreased to 81.3 million baht in 1992. This also related to the MOPH's inconsistent support for the program, as stated above, which reflects an inadequate sales promotion. At present the HCFs are kept at district and provincial levels, with no income generating from the funds. The renewability of unused cards: This was abolished after 1985. No limit of criteria for card use from 1992 to present: There has been no limita- tion imposed on the number of episodes and no ceiling on expenses per episode. The problem of adverse selection or self-selection or selection bias is a com- mon problem in voluntary health insurance schemes. In this study, this prob- lem is quite apparent. The research results show that the presence of illness was one of the significant factors related to card purchase and card utilization pat- terns. This suggests that the selection bias from this source may influence card purchase. Another problem is the bypassing of health centers due to peoples' percep- tion of the low quality of services provided there. In any case, free services were provided to cardholders at all levels in the referral line, and thus cardholders chose free services at hospitals rather than at health centers. Some studies (Hongvivatana and others 1986) have suggested that cost sharing, either as a deductible or a fixed percent of the bill (coinsurance), should be imposed at community hospitals and higher levels but not at the health center. This would encourage greater use at the subdistrict level and more rational use of higher unit cost care at higher levels and would also generate more revenue to the providers. Yet it might deter the demand for cards. The renewability of unused cards, which was abolished after 1985, was a successful incentive for avoiding unnecessary use of health cards. In the present program, the first contact in which the cardholders can choose is either a health center or a community hos- pital, which might be one factor influencing health cardholders to bypass the health centers, as stated above. 352 Health Financing for Poor People: Resource Mobilization and Risk Sharing The problem of case-based reimbursement of providers, which is a retrospective reimbursement in the program, requires substantial administrative capacity and a highly developed information infrastructure. HCF should be aware of the sophis- ticated nature of such a system when establishing reimbursement agreements with public hospitals based on a combination of expected outputs and costs. Finally, a continuing problem of the HCP is dissemination of knowledge about HCP and a better understanding of health insurance in principle among the Thai people. This requires consistent sales promotions and strong annual campaigning about the HCP, a problem separate from that of the frequent adjustments in the program. CONCLUSIONS AND RECOMMENDATIONS Although the health card program has been in operation for more than 15 years and has undergone many adjustments, whether or not the project is a success is still unclear. Thailand's economic structure is changing toward more industrializa- tion with multiple effects on society. The transition poses difficult questions for planners and affects many key issues in the health sector. The country faces many problems, such as how to increase, or at least sustain, economic growth, income distribution, political instability, and incompetence. The overall performance of the economy remains a major concern. Likewise, in the health sector, financing is still the main problem. The objective of this chapter was to assess the future potential for application of voluntary health insurance, the Health Card Program scheme, in Thailand by utilizing data collected in Khon Kaen, where the recent program was implemented. This study has indicated the problems, development, and health service capability in the application of the Health Card Program, iden- tified factors influencing project outcomes, investigated the card purchase and dropout pattern, and evaluated the factors affecting the discrimination between health card purchase, nonpurchase, and dropouts from continued health card member groups. The results of the study will provide more understanding of how the program performs and how to sustain it more efficiently, as well as suggesting alternative ways to improve it. Both qualitative and quantitative statistical tech- niques have been applied to provide complementary approaches to investigating the factors affecting the performance of the program. The research results show that the statistically significant factors related to new card purchase and nonpur- chase groups are education level and presence of illness. Between the health card dropout and continued card purchase groups, the significant factor is again the presence of illness. Moreover, there were four significant factors related to non­card users and card users: educational level, income per year, number of employed members in the family, and presence of illness. A study of health card purchase patterns and health card utilization patterns (health utilization behavior) demonstrates clearly and strongly that health card purchase in Khon Kaen has been influenced by the following factors: proportion Impact of the Thailand Health Card 353 of employed persons to total family members, education, and presence of illness. The last factor confirms the problem of adverse selection in the program, partic- ularly significant in the health card program implemented in Khon Kaen, since the program was introduced in the country and that province in 1983, giving Khon Kaen extensive experience with voluntary health insurance and its devel- opment throughout the years. The sustainability of the program depends on various factors, one of which is a level of satisfaction among card users that leads them to continue to buy in the following years. The findings indicate that the continuity of card purchase in the study was associated with the following factors: age, education, and income, and obtained knowledge on HCP from health center personnel. Moreover, the card- holders who had been persuaded to buy a card at home and had been persuaded by a neighbor to buy a card were more likely to be continued users. This study also investigated the health care seeking pattern among card users and non­card users. The results indicate clearly the importance of accessibility to health care among the card user group. However, the program still has the risk- sharing and card utilization pattern problems to consider in order to become more sustainable and efficient. The results also show that the significant factors related to card usage rates are having chronic illness and convenient access to a health center. This is not surprising since the chronically ill will seek care regularly. Recommendations The uninsured have to pay out of pocket for medical services either in public or private facilities. A proper voluntary health insurance scheme is a choice for peo- ple, especially for the poor, who are not covered by any schemes and are thus not protected from financial difficulties due to the high cost of care. The health card program can be a choice for these people, who are in rural areas or some urban areas. To expand this program for urban residents, it would need many adjustments because there are no community hospitals provided in such areas, only municipal health centers, provincial hospitals, and regional hospitals. The cost of care and types of care are different. If the existing health card program continues to be implemented without any of the adjustments suggested above, it might destabilize the whole health sys- tem. This study suggests another possibility to adjust the program as a compul- sory program. It could be implemented in rural areas as a community-based compulsory insurance scheme. The services in the program would cover cata- strophic cases only, the household would pay out-of-pocket for outpatient care because the various studies, including the present one, show that the cost of such care is not very high. Low-income households will be eligible for the FC/L scheme, which provides free medical care for the low-income households. Aside from that there must be an essential package of health services provided free at the public facilities to guarantee basic care for the people. Given the low proba- bility of hospitalization and the compulsory enrollment, premiums will be low 354 Health Financing for Poor People: Resource Mobilization and Risk Sharing and affordable. For the urban areas it must be linked to health service facilities now provided and will require more information and further studies to estab- lish. This might be an alternative way to assist the uninsured for high cost care. To summarize, the findings reported in this study show that improvements to the existing health card program require: · Efficient and consistent health policy · Revision of the criteria for card use such as number of episodes, type of services, ceiling on expenses, and effective referral system (problem of bypassing the health centers) · Development of reimbursement agreements with public hospitals based on a combination of expected outputs and costs to assist the HCF committee in allocating funds to compensate the providers · A subsidy from the government budget · The strengthening of public health service units · The securing of the health card program as the base for universal health insurance. Finally, the findings in this study can provide more information on HCP perfor- mance and its prospects for the future beyond a pilot project of voluntary health insurance: a prepayment with no cost-sharing scheme. The project aims to achieve the universal coverage that the government has taken as its goal. It is hoped that the results and recommendations emerging from the careful investi- gations in this chapter may assist policymakers to improve and to expand the existing health card program as the base for universal insurance. Acknowledgments: The author is grateful to the World Health Organization (WHO) for hav- ing provided an opportunity to contribute to the work of the Commission on Macroeconomics and Health and to the World Bank for having published the material in this chapter as an HNP Discussion Paper. Part of this research was supported by the Health Insurance Office, Ministry of Public Health (MOPH), Thailand. The study was conducted under the auspices of the College of Public Health and the Centre for Health Economics, Faculty of Economics, Chulalongkorn Uni- versity. The author thanks the staff of the Department of Population and International Health, Harvard School of Public Health, for their valuable comments and support. The findings, inter- pretations, and conclusions expressed in the chapter are entirely those of the author and do not necessarily represent the views of the World Bank, its Executive Directors, or the countries they represent. REFERENCES Adeyi, O. 1988. "Requirement for the Health Card? Sustaining the Demand for Rural Health Insurance in Thailand: A Case Study from Chiang Mai Province." Master's the- sis, Liverpool School of Tropical Medicine. Buisai, S. 1995. "Urbanization and the Health Systems." Health System in Transition, Con- temporary Health Issue. No. 1. Health Systems Research Institute, Thailand. Impact of the Thailand Health Card 355 Chabot, J., M. Boal, and A. da Silva. 1991. "National Community Health Insurance at Vil- lage Level: the Case from Guinea-Bissau." Health Policy and Planning 6(1): 46­54. Dave, P. 1991. "Community and Self-Financing Voluntary Health Programmes in India." Health Policy and Planning 6(1): 20-31. Hongvivatana, T., and S. Manopimoke. 1991. A Baseline Survey of Preference for Rural Health Insurance. Bangkok: Center for Health Policy Studies, Mahidol University. Hongvivatana, T., P. Tantiserrani, P. Predaswat, S. Prompakdee, P. Leerapan, and others. 1986. Health Services Utilization under the Health Card Program. Mahidol University, Monograph of Center for Health Policy Studies No. 4. Bangkok. Janjaroen, W. S., and S. Supakankunti. 1994. Voluntary Health Insurance (Health Card): A Mechanism to Ease up Household Health Care Expenditure Burden. Paper prepared for the VII International Congress, World Federation of Public Health Associations in Bali, Indonesia, December 4­8. ------. 1996. Economic Evaluation of Voluntary Health Insurance: A Case Study from Chiang Mai Province. Bangkok: Chulalongkorn University. Kakwani, N., and M. Krongkaew. 1996. "Big Reduction in `Poverty'." Economic Review Year- End `96, Bangkok Post. Bangkok. Keeratipipatpong, W. 1996. "Expansion by Healthy Rate." Economic Review Year-End `96, Bangkok Post. Bangkok. Kiranandana, T. 1993. "Voluntary Health Insurance in Thailand." Proceedings of a National Workshop on Health Financing in Thailand. Bangkok. Kiranandana, T., and S. Apairatr. 1990. Evaluation of Health Card Project: A National Census. Bangkok: Military Welfare Organization. Kutzin, J. 1995. Experience with Organizational and Financing Reform of the Health Sector. WHO, SHS Paper No. 8. Geneva. Maddala, G. S. 1991. Limited-Dependent and Qualitative Variables in Econometrics. New York: Cambridge University Press. Manopimoke, S. 1993. Financial and Social Sustainability of the New Health Card Approach. Bangkok: Thai-German Cooperation for Health. McPake, B., K. Hanson, and A. Mills. 1992. "Experience to Date of Implementing the Bamako Initiative: A Review and Five Country Case Studies." London School of Hygiene and Tropical Medicine, Department of Public Health and Policy, Health Policy Unit. MOPH (Ministry of Public Health). 1989. Health Card Programme­HCP Chiangmai Pilot Pro- ject 1985­1988: Summary of Main Results and Conclusions. The Health Card Programme Center and the Thai-German Cooperation for Health (GTZ), Bangkok. ------. 1992­96. Health Development Plan under the Seventh. National Economic and Social Development Plan, Health Development Planning Committee, Bangkok. Musgrove, Philip. 1996. "Public and Private Roles in Health: Theory and Financing Pat- terns." World Bank, Discussion Paper No. 339. Washington, D.C. NESDB (National Economic and Social Development Board), UNDP (United Nations Development Programme), and TDRI. 1991. National Urban Development Policy Frame- work Final Report, vol. 1. Bangkok. 356 Health Financing for Poor People: Resource Mobilization and Risk Sharing Nittayaramphong S., and V. Tangcharoensathien. 1994. "Thailand: Private Health Care Out Of Control?" Health Policy and Planning 9(1): 31­40. Norusis, Marija J. 1993. SPSS/PC+ Advanced Statistics version 5.0. Office of the Prime Minister, National Statistical Office. 1986­90. Statistical Yearbook Thai- land. Bangkok: Ladprao Teachers' Association Press. ------. 1992a. Preliminary Report of the 1992 Household Socio-Economic Survey. Bangkok. ------. 1992b. Survey of Private Hospital and Clinic. Bangkok. Pempoonwatanasuk, C. 1985. The Health Card Programme in Ratchaburi Province: A Case Study. Master's thesis, Thammasat University. Piyaratn, P., and W. S. Janjaroen. 1994. Health Insurance in Thailand. Paper presented at the National Seminar on Health Sector Development in Hanoi, Vietnam, October 4­7. Pradabmuk, P. 1984. "The Health Card Program in Ubol Ratchatani: A Case Study." In T. Hongvivatana and others, eds. Health Services Utilization under the Health Card Pro- gram. Center for Health Policy Studies, Mahidol University. Puntachart, P., and others. 1988. An Evaluation of Chiangmai Health Card Project. Research Report No. 34. Research and Development Center, Payub University, Chiangmai, Thailand. Santampol, J. 1990. "Comparative Analysis of Expenses Related to Care Providing, at Sev- eral Levels of Health Card Program: A Case Study of Chiang Mai Province." Master's thesis, Chulalongkorn University. Silapasuwan, V. 1989. "Factors Affecting Health Card Program Utilization among People in Rural Thailand." Journal of Primary Health Care and Development 2 (December): 84­95. Singkaew, T. 1995. "Comparison between Voluntary and Compulsory Health Insurance in Thailand." Health Insurance Office Report 1992. Ministry of Public Health, Thailand. Supakankunti, S. 1994. Structural Change: Impact on Urbanization Process in Thailand. Fac- ulty of Economics. Bangkok: Chulalongkorn University. Supakankunti, S., W. S. Janjaroen, and S. Sritamma. 1996. Economic Analysis of Voluntary Health Insurance Schemes in Khon-Kaen Province. Bangkok: Chulalongkorn University. Suwanteerangkul, J. 1992. "Factors Influencing Drop Out in Health Cardholders: A Case Study of Mae Rim District, Chiang Mai Province, 1989­1991." Master's thesis, Chula- longkorn University. Tangcharoensathien, V. 1995. "Health Care Financing in Thailand." Health System in Tran- sition, Contemporary Health Issue, No. 1. Health Systems Research Institute, Thailand. Tangcharoensathien, V., and A. Suphachutikul. 1993. "Health Insurance in Thailand: Pre- sent and Future." In MOPH, Office of Undersecretary, Report of Health Insurance Office: 1992 Performance. Bangkok. Terdudomtham, T. 1996. "Analysts Taken by Surprise." Economic Review Year-End `96, Bangkok Post. Bangkok. Thailand Economic Information Kit. 1994. Thailand Development Research Institute. Bangkok. Thailand in Figures. 1996. Alpha Research Co., Ltd. Sukhum and Sons Co., Ltd. Bangkok. Impact of the Thailand Health Card 357 Thailand Public Health. 1995. Alpha Research Co., Ltd. and Manager Information Services Co., Ltd. Sukhum and Sons Co., Ltd. Bangkok. UNDP (United Nations Development Programme). 1995. Human Development Report. New York: Oxford University Press. Veeravongs, S. 1994. "Factors Influencing Health Card Purchase and Continuity: A Case of Voluntary Health Insurance in Phuket Province, Thailand." Takemi Research Paper No. 89. Takemi Program in International Health. Harvard School of Public Health. Harvard University. Wibulpolprasert, S. 1991. "Community Financing: Thailand's Experience." Health Policy and Planning 6(4): 354­60. World Bank. World Development Report 1993: Investing in Health. New York: Oxford Univer- sity Press. PART 3 Expenditure Gaps and Development Traps 11. Deficit Financing of Health Care for the Poor Alexander S. Preker, John C. Langenbrunner, and Emi Suzuki 12. Impact of Risk Sharing on the Attainment of Health System Goals Guy Carrin, Riadh Zeramdini, Philip Musgrove, Jean-Pierre Poullier, Nicole Valentine, and Ke Xu CHAPTER 11 Deficit Financing of Health Care for the Poor Alexander S. Preker, John C. Langenbrunner, and Emi Suzuki Abstract: What is the optimal amount of health care spending needed to achieve a given outcome, and how much is a country able and willing to afford? Health sector experts and policymakers have asked these questions for decades. In recent years, the Millennium Development Goals (MDGs) have become a quantitative set of targets for poverty reduction and improvements in health, education, gender equality, the environ- ment, and other aspects of human development. To help focus national and interna- tional priority-setting, the goals and targets selected were intended to be limited in number, be stable over time, and be easily communicated to a broad audience. This chap- ter attempts to estimate the expenditure needed to achieve the health-related targets set by the MDGs--reducing the number of children who die before their fifth birthday and women who die from complications of childbirth, and reversing the spread of HIV/AIDS, malaria, tuberculosis, and other major diseases--using a production frontier technique. The chapter compares this expenditure estimate with current expenditure trends in low- and middle-income countries. D uring recent years, the Millennium Development Goals (MDGs) have become a quantitative set of targets for poverty reduction and improve- ments in health, education, gender equality, the environment, and other aspects of human development (see box 11.1).1 This chapter attempts to esti- mate the expenditure needed to achieve the MDGs health-related targets-- reducing the number of children who die before their fifth birthday and women who die from complications of childbirth, and reversing the spread of HIV/AIDS, malaria, tuberculosis (TB), and other major diseases--using production frontiers as an analytical technique. The chapter compares this expenditure estimate with current expenditure trends in low- and middle-income countries to arrive at an estimate of the global health care expenditure gap. PROGRESS TOWARD ACHIEVING THE MDGS Progress since 1990 in achieving the MDGs has been uneven across countries and regions and uneven among the goals themselves (Devarajan, Miller, and Swanson 2002) and (Preker, Langenbrunner, and others 2001). The United Nations set 1990 as the baseline year for monitoring the targets and 2015 as the target date for achieving the goals. In 1990, 29 percent of the global population--1.3 billion people--lived in extreme poverty (first component of Goal 1). This had dropped to 22.7 percent of 362 Health Financing for Poor People: Resource Mobilization and Risk Sharing BOX 11.1 MILLENNIUM DEVELOPMENT GOALS (1990­2015) 1. Eradicate extreme poverty and hunger - halve the proportion of people living on less than US$1 a day - halve the proportion of people who suffer from hunger 2. Achieve universal primary education - ensure that boys and girls alike complete primary schooling 3. Promote gender equality and empower women - eliminate gender disparity at all levels of education 4. Reduce child mortality - reduce by two-thirds the under-five mortality rate 5. Improve maternal health - reduce by three-quarters the maternal mortality ratio 6. Combat HIV/AIDS, malaria, and other diseases - halt and reverse the spread of HIV/AIDS - halt and reverse the spread of malaria and tuberculosis 7. Ensure environmental sustainability - integrate sustainable development into country policies and reverse loss of environmental resources - halve the proportion of people without access to potable water - significantly improve the lives of at least 100 million slum dwellers 8. Develop a global partnership for development - increase official development assistance, especially for countries applying their resources to poverty reduction - expand market access - encourage debt sustainability the population--1.15 billion people--by 1999. During the same time period, the undernourished dropped from 22 percent of the global population to 18 percent-- 780 million people (second component of Goal 1). The global targets for both poverty reduction and hunger are broadly on target (see figure 11.1a). The income poverty indicator is important, since it correlates highly with the overall social indicators such as health status and financial protection against the cost of illness. Although growth translated into poverty reduction, the total head count of poor people depends significantly on historical levels of income distrib- ution, policy choices, and institutional constraints (World Bank 2000). The best progress globally has been in the East Asia and the Pacific region. Sub-Saharan Africa and South Asia, where the majority of the world's poor live, have experi- enced the worst progress. Sub-Saharan Africa failed to grow during the 1990s, leaving a large part of the population in poverty. In Eastern Europe and Central Asia and the Middle East and North Africa, hunger increased during the 1990s. Deficit Financing of Health Care for the Poor 363 FIGURE 11.1 Millennium Development Goals, Global Aggregate a. Eradicate Poverty and Hunger Ensure Environmental Sustainability People living on less than $1 a day Improved water source 30 (% of population with access) 29 23.4 90 88 20 85 14.5 81 10 80 76 0 75 1990 1998 2015 1990 1995 2000 2005 2010 2015 b. Universal Primary Education Promote Gender Equality Ratio of girls to boys in primary and Net primary enrollment secondary school (% of school age children) 100 100 100 100 95 80 89 83 60 90 40 85 20 84 81 80 0 1990 1999 2015 1990 1999 2005 Total government revenues as % GDP 100 · Governments in many countries often 80 raise less than 20% of GDP in public revenues; and 60 40 · The tax structure in many low-income countries is often regressive. 20 0 100 1,000 10,000 100,000 Per capita GDP (log scale) 364 Health Financing for Poor People: Resource Mobilization and Risk Sharing During this time period, people without access to water decreased from 27 percent to 21 percent of the population (first component of Goal 7; see figure 11.1a). This still left 900 million without access to clean water. Those without access to sanitation systems dropped from 56 to 48 percent of the population (second component of Goal 7). Improved policies and significant additional financial resources are required to address these problems. Data for the global indicator on education and gender equality (Goals 2 and 3) are not complete. Based on the available data, both goals appear to be mod- estly off target (see figure 11.1b). Net primary enrollment rates increased only from 81 percent to 84 percent, and the ratio of girls to boys education increased only from 82 percent to 86 percent in 1990­99. Improved policies and signifi- cant additional financial resources are needed to address these problems. Among all the MDGs, the health goals--maternal and child health--are the most seriously off track (see figure 11.1c). The MDG for maternal and child health calls for reducing maternal mortality by three-quarters and the under-five child mortality rates by two-thirds of their 1990 levels by 2015. Tracking progress in reducing maternal mortality is difficult. Deaths related to pregnancy and child- birth occur infrequently when compared with other health problems and are often outside the formal health system. This leads to a small sample size made more unreliable by underreporting. According to the last estimate of maternal mortality, for 1995, 500,000 women die annually during pregnancy and child- birth, most of them from conditions that could be prevented or treated in equipped medical facilities. Not surprisingly, maternal mortality is low in the Latin American and East European regions, where skilled attendants and equipped medical facilities are readily available, while high maternal mortality occurs in the African and South Asian regions where they are not. Tracking progress in reducing infant and child mortality is more reliable. Global progress toward this goal is seriously off track. This is particularly vexing because so much is known about the causes of infant and child mortality. Fur- thermore, progress already made in some countries, even at very low income levels, indicates that effective interventions are both readily available and affordable to most countries. Part of the problem is that progress in achieving under-five mortality targets relies significantly on both nonspecific intersectoral actions and specific health care interventions (preventive and curative services). The nonspecific activities include, for example, poverty reduction, nutrition, education, and gender equal- ity programs, improvements in access to clean water and sanitation systems, and usage of insecticide-treated bed nets. Many of these activities require focused government policies across different sectors, coordination, and ongoing moni- toring and evaluation of progress. This is often lacking at low-income levels and in settings with severely constrained management and institutional capacity. But even with good general hygiene and other preventive and health promo- tion measures, children get sick and need medical interventions. Many interven- tions are as simple as vaccination, oral hydration during diarrhea, and use of Deficit Financing of Health Care for the Poor 365 antibiotics to treat complicated upper respiratory infections. Health providers with only basic skills are able deal with most of the conditions of childhood using simple protocols such as those available through integrated management of childhood illness. Other interventions, however, require skilled birth attendants, inpatient care during complicated pregnancies, and knowledge about appropriate treatment and referral at the time of trauma and other more serious illnesses. Many of the old, and a few of the new, scourges of poverty are still ravaging low- and middle-income countries, threatening both human welfare and the potential for medium-term growth (see figure 11.2). In 2000, 34.7 million adults and 1.4 million children had HIV/AIDS, 300 million to 500 million cases of malaria resulted in 1 million to 2 million deaths, mainly in children under five years of age; and 8.4 million new TB cases were reported, between 10 and 15 per- cent of them in children. As measured by new cases reported, the incidence of HIV/AIDS and TB infections is still increasing, and the MDG global targets are far off track. Sub-Saharan Africa and South Asia are the regions most severely affected. If they are to be effective, health sector-specific programs need to be under- pinned by well-functioning health systems. These include strong government stewardship (policymaking, coordination, regulations, contracting, information dissemination, and monitoring and evaluation systems), health care financing (prepaid revenue collection, risk pooling, and resource allocation and purchasing FIGURE 11.2 Strong Correlation between Wealth and Health across Time Life expectancy in years 80 1990 About 1960 70 About 1930 60 About 1900 50 40 30 0 5,000 10,000 15,000 20,000 25,000 Income per capita (1991 US$) 366 Health Financing for Poor People: Resource Mobilization and Risk Sharing mechanisms), input generation and management (human resources, research, pharmaceuticals, medical technology, consumables, and capital), effective and responsive service delivery systems (public health and curative services), and the market forces created by demand for service by individuals and households. Many interventions known to be effective and affordable do not get to the chil- dren or households that need them due to failures in the health system. Once again, improved policies and significant additional financial resources are needed to address these problems in the underlying health systems. KEY DRIVERS OF ACCELERATED PROGRESS TOWARD ACHIEVING THE MDGS Increased Income from Economic Growth Is Necessary Richer countries do better across a wide range of health indicators. As a result of complex synergies among income levels and expenditure on education, health- seeking behavior, public policy, and health services, people all over the world live almost 25 years longer today than they did at similar income levels in 1900 (see figure 11.2). This relationship between income, health spending, and health outcomes in developing countries is now well-established (World Bank 1993, 1997; Pritchett and Summers 1996; WHO 2000). This story is as true for maternal and child mortality as it is for the major dis- ease challenges facing low-income countries (see figure 11.3). Child mortality decreases as incomes increase, a relationship some analysts have used as a perfor- mance indicator for a country's overall development policy (Wang and others 1999). The "black box" assumptions is that income mediates through a variety of inter- mediate factors, including nonhealth sector determinants of better health out- comes (such as education, nutrition, safe water, roads, and sanitation systems), health-enhancing policies (maintaining health and preventing disease), and health services (treating disease, palliative care, and preventing death). But recent work indicates that even at very high economic growth during the next few years, most countries will not reach the MDG targets of two-thirds reduction in under-five mortality rate (U5MR), a three-quarters reduction in maternal mortality (MMR), and a halving of the prevalence of underweight (UW) among children (Devarajan, Miller, and Swanson 2002; Alderman and others 2000). By projecting the current level of several health-related MDG indi- cators to 2015, using the historical elasticity of U5MR, MMR, and UW with respect to income, Wagstaff (2002a) demonstrates that even an unlikely 8 per- cent growth in income between 1990 and 2015 would reduce U5MR by only 20 percent, MMR by only 30 percent, and UW by only 40 percent, compared with the target reductions of 67 percent, 75 percent, and 30 percent, respectively. Furthermore, in another paper, Wagstaff (2002b) argues that the importance of extrasectoral programs may mislead because they often act by improving the effectiveness of more specific health programs. For example, investment in roads allows pregnant mothers to get to delivery services on time, receive vaccines, Deficit Financing of Health Care for the Poor 367 FIGURE 11.3 Income and Child Mortality Under-5 mortality rate 200 Ethiopia 180 Nigeria 160 Bangladesh 140 Pakistan 120 Predicted level 100 India Indonesia 80 Turkey 60 Brazil Iran Mexico China 40 20 Russia 0 0 1,000 2,000 3,000 4,000 5,000 6,000 Income per capita (US$) and reach health centers without having the cold chain broken. Education allows mothers to make the right choices when faced with illness episodes. The advantages of investments across sectors are therefore likely to have important synergies and complementarities across sectors that specific invest- ments in the health sector alone would fail to achieve. A disadvantage of this approach is that it may still fail to correct known constraints in the health sector that could be addressed effectively and at lower cost by focusing directly on the specific problems and their solutions. It is also a leap of faith to think that the marginal, untargeted dollar would always be spent on health-enhancing activi- ties related to the MDGs. The alternative choices, some of which might even shunt resources away from priority activities, are almost limitless. Income from Economic Growth Alone Is Not Enough to Achieve the MDGs As seen in figures 11.2 and 11.3, at any given income level, there is a wide range of performance in terms of child health outcomes. Furthermore, the high per- formers in absolute terms are not always the best performers in relative terms. For example, both China and Singapore do well in terms of absolute perfor- mance but lag behind progress made during the 1980s (Wang and others 1999). 368 Health Financing for Poor People: Resource Mobilization and Risk Sharing The variability in performance at any given income level is thought to be the result of a combined effect produced by several factors that include differences in both the effectiveness of interventions and the health systems delivering them (WHO 2000; Anell and Willis 2000; OECD 2001). Increased Total and Public Spending on Health Care A second driver of accelerated progress toward achieving the MDG targets is in- creased public and total spending on health or nutrition services--without increases in income.2 Past work has shown a positive, though modest, correlation between health spending and health outcomes even when controlling for income and possible confounding factors (Filmer and Pritchett 1999). However, caution should be used when attributing improvements in health outcomes to nontargeted health spending since it is very hard to fully isolate all the confounding factors. Notwithstanding this constraint, the current level of several health-related MDG indicators can be projected to 2015 using the historical elasticity of U5MR, MMR, and UW with respect to health expenditure. Wagstaff (2002a) demon- strated that between 1990 and 2015, U5MR would drop by an additional 40 per- cent, MMR by 30 percent, and UW by 35 percent, compared with the target reductions of 67 percent, 75 percent, and 30 percent, respectively. A sustained health expenditure growth rate of more than 15 percent would be needed to reach the U5MR target of 75 percent. These findings refute earlier work that suggested health care services are not significant contributors to health status relative to other measures such as sani- tation, income, and education (Newhouse and Friedlander 1980). An advantage of such broad, systemwide expenditure increases on health ser- vices is that it allows policymakers and managers to exercise decision rights over the allocation and use of funds in areas they think are most effective and where there is the greatest demand. This is likely to be highly context specific and not readily specified under a blueprint. A disadvantage of this approach is that it may fail to correct known con- straints in financing, inputs, and service delivery, and deliver programs and interventions that could be addressed more effectively and at lower cost through a more direct targeting of scarce public resources in priority areas. It is a leap of faith to think that, without some strategic priority setting and targeting, the marginal dollars spent on the health sector will always be spent in areas that will have the greatest impact on accelerating progress toward the MDGs. Increased Spending on Priority Populations, Priority Interventions, and Priority Health Programs A third driver of improved outcomes is knowledge about the determinants of poverty and poor health related to the MDG targets (for example, the links between hygiene and infections, maternal nutrition and low birth weight, diet and malnutrition, and poverty and health) and the implementation of effective health programs and interventions. Deficit Financing of Health Care for the Poor 369 Researchers have a much better understanding today of the determinants of poverty (World Bank 2000), the intersectoral synergies needed to achieve good health outcomes (Wagstaff 2002b), and the role health, prevention, and curative health services play in this story (World Bank 1993; WHO 2000; Van Doorslaer and others 2000; Wagstaff 2002c). Extensive work has been done in the area of targeting health programs to benefit the poor (Claeson and Waldman 2000; Gwatkin 2000; Gwatkin and Heuveline 1997). During the past few years, a considerable body of knowledge has also accu- mulated on the cost and effectiveness of alternative interventions for specific health conditions (Murray, Evans and others 2000; Evans 1990b). Early work using such techniques in developing countries looked mainly at the cost-effec- tiveness of specific interventions (Barnum 1986), disease control programs (Bar- num, Tarantola, and Setiady 1980), and investment projects (Barnum 1987; Prescott and De Ferranti 1985; Mills 1985a, 1985b). This type of work exploded following publication of the World Development Report 1993: Investing in Health (World Bank 1993; Jamison and others 1993) and subsequent extensive work by the World Health Organization in this area. Figure 11.4 provides a few examples of specific interventions that, if implemented well, would have a large impact on reducing the burden of disease, especially among the poor (World Bank 1997; Gwatkin and Heuveline 1997; Claeson, Mawji, and Walker 2000). FIGURE 11.4 Cost-Effective and Affordable Public Health and Clinical Services Indicative cost in US$ Integrated management 14% of childhood illness 40.00 1.60 Immunization (EPI plus) 6% 14.50 0.50 Prenatal and delivery care 4% 40.00 3.80 Family planning 3% 25.00 0.90 AIDS prevention program 2% 4.00 1.70 Treatment of STDs 1% 2.00 0.20 Treatment of tuberculosis 1% 4.00 0.60 School health program 0.1% 22.50 0.30 Tobacco and 0.30 0.1% 42.50 alcohol program Cost per Annual cost 0 2 4 6 8 10 12 14 16 DALY per capita Percentage of total global disease burden averted 370 Health Financing for Poor People: Resource Mobilization and Risk Sharing These interventions rank high in terms of their (a) potential to avert a large percentage of the global disease burden, (b) low cost per DALY averted, (c) low annual cost per capita, (d) potential impact on socially excluded and poverty groups, and (e) relevance to the MDGs. At the heart of the broadened policy use of cost effectiveness is a belief among health professionals that resources in the health sector should be allocated across interventions and population groups to generate the highest possible overall level of population health. Stated in other terms, allocative efficiency of the health sector could be enhanced by moving resources from cost-ineffective to more cost-effective interventions (World Bank 1993). Musgrove (1999) provides a decision tree for rational use of public financing in the health sector. It starts with the overarching issue of allocative efficiency by asking if the proposed expenditure is for public goods, generally population-based services. If the answer is "yes," the next step is to rank such expenditures in terms of cost- effectiveness--or even better, benefit-cost analysis--to decide what will be funded. If proposed expenditures do not meet public goods criteria, the tree asks whether signif- icant externalities are involved, whether risks of catastrophic costs are involved, and whether the proposed beneficiaries are poor. Thus allocative efficiency, risk, equity, and cost-effectiveness interact to determine public financing decisions in health. Eco- nomic principles govern each decision point, but many other factors are often weighed, so the outcomes will vary considerably from country to country. The over- riding principle is maximizing the potential impact on people, especially the poor. Several challenges to this wider use of cost-effectiveness analysis (CEA) have emerged, especially when it is used as a basis for priority setting or allocative effi- ciency within a given budget envelope or when it is used to undertake "bottom-up" costing to estimate marginal extra dollars spent on the health sector (M. Williams 1997; Filmer, Hammer, and Pritchett 2000, 2002; Jack 2000). Proponents of CEA defend their position with equal vigor (Musgrove 2000a, 2000b; Rivlin 2000). First, analysts and decisionmakers have correctly noted that resource alloca- tion decisions affecting the entire health sector must also take into account social concerns for the sick, reducing social inequities in health, the well-being of future generations, the insurance effect (spreading the cost of infrequent but expensive care across population groups), and the political economy of the middle classes who pay taxes (Hauck, Smith, and Goddard 2002). Second, current CEA practice often fails to identify existing misallocation of resources by focusing on the eval- uation of new technologies or strategies. Third, for all but the richest societies, the cost and time needed to evaluate the large set of interventions required to use CEA to identify opportunities to enhance allocative efficiency may be prohibi- tive. Fourth, the difficulties of generalizing context-specific CEA studies have been institutionalized by the proliferation of multiple national and subnational guidelines for CEA practice, all using slightly different methods. For example, costs can vary greatly from one country, context, and interven- tion modality to another. A naive generalization of the finding from one study to another can lead to serious mistakes in the planning and implementing of other- Deficit Financing of Health Care for the Poor 371 wise effective interventions. For example, in the management of malaria, it is sug- gested that at low-income levels one would choose case management and prophy- laxis for pregnant women. At middle-income levels, one would add spraying. At higher income levels, one would add bed nets. A single estimate of the cost- effectiveness of malaria treatment could lead to the wrong conclusion that malaria programs are unaffordable if the estimate was based on a multitherapy program in a low-income country. Or it could lead to a serious underestimate of the actual cost of the program if the estimate was based on single therapy calculations but implementation of a multitherapy program in a middle-income country. Factors other than income may also alter a program's actual cost during implementation. Major overlooked factors include availability, mix, and quality of inputs (especially trained personnel, drugs, equipment, and consumables); local prices, especially labor costs; implementation capacity; underlying organi- zational structures; incentives; and supporting institutional framework. Other confounding factors include poor quality of data; confusion between marginal, average, and shared costs; competing risks and synergies; failure to include non- monetary costs such as time and lost income; and miscalculation of discount rates (Hammer 1996; Peabody 1999). The counterargument often used by proponents of cost-effectiveness analysis is that, although international estimates may not fully reflect local circum- stances, there is a risk that excessive contextualized analyses will be too complex and resource-intensive for most low-income countries (Murray and others 2000). A move in this direction could ultimately lead to less use of evidence-based pol- icy dialogue. This school of thought recommends an alternative approach: focusing on a general assessment of the different interventions' costs and health benefits based on general league tables of the cost-effectiveness of interventions for a group of populations with comparable health systems and epidemiological profiles. Such information on generalized cost-effectiveness can then be used alongside consideration of the effects of different resource allocations on other important social goals. Spending on Management Capacity, Organizational Structures, and Institutional Environment Several systemic factors in the health system may also act as drivers of improved outcomes. These include: management capacity, organizational incentives, and institutional environment (see figure 11.5). Many countries currently failing to make progress toward the MDGs are plagued by, for example, weak management capacity, negative organizational incentives, and lack of a strong regulatory environment to ensure quality con- trol and deal with the private sector (see figure 11.6). Often these countries ignore the demand side of utilization of health services. Critical supply chains, such as those involving pharmaceuticals and vaccines, are broken. In addition, top-down centralized control over public services excludes participation by the private sector, communities, and households in the care that they receive. 372 Health Financing for Poor People: Resource Mobilization and Risk Sharing FIGURE 11.5 Three Nonfinancial Determinants of Good Outcomes Capacity Incentives Institutions Outcomes FIGURE 11.6 Lack of Management Capacity, Adverse Incentives, and Weak Institutions Break the Fulcrum Weak policies and corruption Government . gov Local Lack of community Providers involvement Problems with supply Individuals Money alone Lack of demand Outcomes Benefits Deficit Financing of Health Care for the Poor 373 Notably the strategic use of stable and pooled revenue flows remains a critical factor in a country's ability to achieve good overall health outcomes (including those related to the MDGs) and protect its people against the impoverishing effects of illness (Preker and others 2002, 2001; WHO 2000; Carrin and others 2001). Countries and communities that channel health care financing through risk-sharing and collective-purchasing arrangements do significantly better on outcome indicators than countries that rely more heavily on out-of-pocket spending (Jakab and others 2001; Preker and others 2001). The exact contribution of management, organizational, and institutional variables to the MDGs has not yet been quantified. The cost of reforms in this area is highly context sensitive, making it risky to apply estimates from one set- ting to another. An early attempt has been made by WHO (2000) to quantify the benefits of health systems in terms of the level and distribution of health out- comes, financial fairness, and responsiveness to patient expectations in terms of the quality and ethical dimensions of care. Unfortunately, as in the case of cost effectiveness, this initiative has met with great resistance from the international development community and countries themselves (A. Williams 2001; Murray and others 2001; Wagstaff 2001, 2002d; Navarro 2000, 2001; Blendon, Kim, and Benson 2001; Murray, Kawabata, and Valentine 2001). ESTIMATING THE COST OF ACHIEVING THE MDGS These caveats notwithstanding, several traditional approaches have been used to estimate the cost of accelerated progress toward achieving the MDGs. They include: · Estimates of the cost of achieving the MDG goals based on known expendi- ture elasticities of health outcomes (Wagstaff 2002b; Devarajan, Miller, and Swanson 2002) · Estimates of additional public expenditure needed to achieve outcome-based production frontiers · Estimates of the total cost of introducing new programs using bottom-up costing (Evans 1990a; WHO 2001, 2002b; Kumaranayake, Kurowski, and Conteh 2001). · Estimates of the marginal cost of addressing major constraints (Soucat and others 1992) · Specific estimates of scaling up child heath and other interventions and treat- ment for priority diseases (Tulloch 1999; Lambrechts, Bryce, and Orinda 1999; Weissman 2001; Weissman and others 2001; Garrison and Mccall 1990). The Use of Production Frontiers In this chapter, we used production frontiers to estimate the costs of accelerated progress toward achieving the MDGs. Production frontiers have been the subject of a great deal of research on inefficiency (Farrel 1957). The production function 374 Health Financing for Poor People: Resource Mobilization and Risk Sharing shows the maximum output (for example, real observed outcome in terms of health) that can be obtained from a given level of input mix, such as factors of production (for example, capital, human resources, drugs, equipment, and con- sumables), product market (for example, hospital services, ambulatory services, and diagnostic clinics), and prevailing technology. Alternatively, production functions can also be used to describe the minimum amount of inputs required to achieve various specific outputs (MDG targets). The production frontier describes the limits to the outputs (outcome) that can be achieved using differ- ent combinations of inputs. Variations in maximum output can occur either as a result of stochastic effects (for example, random events, unpredictable economic shocks, or bad weather), or it can occur from the fact that firms (health care providers) may operate at various levels of inefficiency due to suboptimal use of existing tech- nology (input mix, throughput processes, outputs), mismanagement, inefficient organizational structures (for example, lack of economy of scope or scale, com- promised decision rights, and adverse incentives), or dysfunctional institutional environment (for example, lack of an appropriate legal framework to deal with market failure or to promote efficient competition). Outputs (such as health out- comes) that fall short of a given target can sometimes be brought up to the pro- duction frontier by changing one or more of these critical variables. Recently, stochastic production frontiers have been widely applied to assess firm inefficiencies in various settings (Aigner, Lovell, and Schmidt 1977; Battese and Coelli 1995). In theoretical models, a wide range of input and output vari- ables can be specified. In real life, observable output levels and available data on the inputs impose constraints. Thus an initial challenge is to construct an empir- ical production function or frontier, based on the observable data (Lewin and Knox 1990; Coelli, Prasada-Rao, and Battese 1998). Past Use of Production Frontiers in the Health Sector The use of production frontiers in the health sector in developing countries is not new. Production frontiers have been used for many years to study efficiency in the heath sector of the Organisation for Economic Co-operation and Develop- ment (OECD). Production frontiers based on cost-effectiveness data were used in the 1993 World Development Report to select from among alternative cost-effective treatments that would be most effective and affordable at low-income levels (World Bank 1993). Production frontiers have been used to examine hospital per- formance issues (Wagstaff 1989; Wagstaff and Lopez 1996). Expenditures linked with inputs (such as beds and staff), intermediate measures (such as average lengths of hospital stays, waiting times), and outcomes (such as infant mortality) can begin to provide some relative notions of health system efficiency (Schieber, Poullier, and Greenwald 1991; OECD 1992, 1994; Anderson and Poullier 1999). They have been used to compare different levels of input mix (Anell and Willis 2000). Production frontiers have been used to examine efficiency issues related to public-private mix in service delivery and financing (Musgrove 1996). Deficit Financing of Health Care for the Poor 375 Production frontiers have also been used to assess the overall country perfor- mance in achieving good health outcomes and more specifically U5MR (Wang and others 1999). WHO (2000) used production frontiers to rank the perfor- mance of health systems in different countries, using the relationship between health expenditure and three outcome indicators--financial fairness, patient responsiveness, and outcomes. The OECD (2002) also made use of production frontiers but employed a different set of indicators. Application of Production Frontier Analysis in Estimating the Global Expenditure Gap As described earlier, the production function for health and the determinants of U5MR, MMR, and UW are complex and multisectoral. A number of different production frontier models could be constructed that would shed some light on various dimensions of the input and product mix needed to attain maximum outcomes in terms of the MDG-related indicators. The production frontier approach was one of the methodologies used to esti- mate the global expenditure gap for the Macroeconomic Commission on Health (WHO 2002c). To establish a production frontier relative to health spending and the MDGs, two assumptions were made (Preker, Langenbrunner, and Suzuki 2001). First, the maximum level of total resources for health care that a country can mobilize (if it could) is likely to be less than, or equal to, the current highest spender at similar income levels. Although there is some variation within income bands, the income versus health spending elasticity is well-known and research documented (Schieber and Maeda 1997). Second, although countries already spending much more than the best performers in outcomes at similar income levels may still benefit from additional spending (based on spending- outcome elasticities), there is probably also considerable scope for improving their efficiency of spending since other countries are able to do much better with fewer resources. Source of Data We examined 135 countries where gross domestic product (GDP) per capita is less than, or equal to, US$7,000. This cut-off point was chosen so that the analysis would correspond to low- and middle-income. The analysis was done for both total and public expenditure on health care. We used the most recent data on health expenditure available in the WHO health expenditure database (usually 1998 data) and GDP per capita from the World Bank SIMA database (matching as closely as possible the 1998 health data). For most of the health indicators, we also matched as closely as possible the date of the health expen- diture (HIV prevalence 1999, incidence of tuberculosis 1999, life expectancy 1997, adult male mortality 1997, adult female mortality 1997, maternal mortal- ity 1995). To assess relative performance in improving U5MR, we used the 1990 to 2000 trend. 376 Health Financing for Poor People: Resource Mobilization and Risk Sharing Establishing the Production Frontier for Health Expenditure per Capita To establish the production frontier, we used the health expenditure (total and public) for the 20 percent of countries that performed the best in absolute terms on several health indicators (most, but not all, related to the MDGs). The health indicators used to select the best performers in absolute terms included: U5MR, MMR, HIV prevalence rate between the ages of 15 and 49, incidence rate of tuberculosis, life expectancy at birth, adult male mortality rate, and adult female mortality rate (World Bank 2002c). For the U5MR, we also created a second pro- duction frontier, based on the health expenditure level for the 20 percent of countries that performed the best in relative improvement between the years 1990 and 2000. An exponential regression on a double log scale was employed to construct the production frontier using the health expenditure (total and public) per GDP per capita data points for the high performers under each of the selected health indicators. Given the small size of the resulting dataset for the production fron- tier countries, the application of more refined statistical techniques, such as sto- chastic analysis, was not relevant. Once the production frontier was established, the gap between the target health expenditure corresponding to the best performers and the observed health expenditure for any given country (adjusting for the population size) could be calculated. Figures 11.7a and 11.7b provide an example of the expendi- ture gap using the absolute level of spending in countries that performed the best on U5M compared with the expenditure for the production frontier coun- tries. The figures illustrate the expenditure gap that would have to be filled in the case of six countries used as case study examples on scaling up for the September 2002 Development Committee Report (World Bank 2002a). For China and India, where over 90 percent of the world's poor live, total expenditure on health care is already higher than in the corresponding produc- tion frontier countries. Public spending on health in India is, however, well below the best performance expenditure frontier, while public spending on health in China is slightly above expenditure in the production frontier coun- tries for U5M. Table 11.1 provides the numerical estimate of the expenditure gap for these six countries, using different outcome indicators to establish the pro- duction frontier. The production frontier trend lines for total expenditure on health care using best performance on various health outcomes--under-five mortality, maternal mortality, life expectancy, adult male mortality, adult female mortality, TB prevalence, HIV/AIDs prevalence--are shown in figure 11.8. Both the countries used to determine the frontier (large circles) and the nonfrontier countries (small circles) are indicated on each graph. The production frontier trend lines for public expenditure on health care, using best performance on various health outcomes, are shown in figure 11.9. The total and public expenditure gap for countries with a per capita income of less than US$7,000 was estimated by summing the gap for each of the countries Deficit Financing of Health Care for the Poor 377 FIGURE 11.7 Expenditure Frontier and Six Countries a. GDP per Capita and Total Health Expenditure (frontiers = under-5 mortality) Log10 (HE per capita US$) 3 y = 0.1951e0.6941x Nonfrontiers 2.5 R2 = 0.9064 Frontiers (under-5 mortality) 2 Uganda India 1.5 Haiti China 1 Dominican Republic Brazil 0.5 Expon. (frontiers [under-5 mortality]) 0 2 2.5 3 3.5 4 Log10 (GDP per capita US$) b. GDP per Capita and Public Health Expenditure (frontiers = under-5 mortality) Log10 (public HE per capita US$) 3 y = 0.0538e1.0273x Nonfrontiers 2.5 R2 = 0.7694 Frontiers (under-5 mortality) 2 Uganda India 1.5 Haiti China 1 Dominican Republic Brazil 0.5 Expon. (frontiers [under-5 mortality]) 0 2 2.5 3 3.5 4 Log10 (GDP per capita US$) total exp. as 5,381,798 47,300,000 29,500,000 39,000,000 1997 Gap public 3,600,000,000 5,780,000,000 2 4 4 0 5 35 expectancy Gap/ capita Life GDP) 6 8 9 13 41 capita 189 nda Needed/ HE for total exp. total exp. as as 76,600,000 62,400,000 37,800,000 87,100,000 log10 Gap public 144,000,000 155,000,000 8,970,000,000 3,820,000,000 1995 Gap public 4,500,000,000 2,140,000,000 1999 10,200,000,000 10,300,000,000 and of mortality 7 9.2 8 3 5 5 2 incidence Gap/ capita 10 19 23 Gap/ capita 11 62 TB Maternal Performers 7 9 11 13 15 21 55 10 15 47 Regressions, 177 216 capita capita Needed/ Needed/ Percent 20 of 0 Best the total exp. total exp. Exponential as as are 75,000,000 46,300,000 Performers 39,600,000 26,400,000 59,400,000 Gap public 177,000,000 1997 5,350,000,000 4,690,000,000 Gap public 3,210,000,000 1999 24,300,000,000 13,300,000,000 Using Percent Frontiers 20 mortality 4 5 6 4 2 3 3 0 7 (US$, 22 80 prevalence Gap/ capita 147 Best Gap/ capita female HIV the are 8 9 Adult 6 7 8 11 17 58 13 43 Frontiers 301 234 capita capita Needed/ Needed/ Frontiers Like 0 More total exp. total exp. as as Be 31,600,000 22,000,000 36,300,000 60,600,000 change 2000 1997 110,000,000 121,000,000 to Gap public 2,740,000,000 Gap public 7,130,000,000 6,850,000,000 4,460,000,000 % 11,600,000,000 and 1990 2 3 3 0 4 mortality 5 7 8 6 70 15 27 Needed mortality Gap/ capita Gap/ capita -5 male between Under 6 7 8 Adult 9 Health 12 40 11 31 19 51 224 181 capita capita on Needed/ Needed/ on on (US$) (US$) Expenditure capita 4 4 5 13 36 capita 4 4 5 13 36 154 154 expenditure per expenditure per Public Public health Public health 11.1 ABLET Republic Republic Country Uganda India Haiti China Dominican Brazil Country Uganda India Haiti China Dominican Brazil 378 Deficit Financing of Health Care for the Poor 379 FIGURE 11.8 Production Frontiers for Total Expenditure on Health Care (Using Best Performance on Various Health Outcomes) Frontiers = under-5 mortality Frontiers = HIV prevalence Log10 (HE per capita US$) Log10 (HE per capita US$) 3 3 y = 0.1951e0.6941x y = 0.1992e0.6973x 2.5 R2 = 0.9064 2.5 R2 = 0.8552 2 2 1.5 1.5 1 1 0.5 0.5 0 0 2 2.5 3 3.5 4 2 2.5 3 3.5 4 Log10 (GDP per capita US$) Log10 (GDP per capita US$) Frontiers = TB incidence Frontiers = life expectancy Log10 (HE per capita US$) Log10 (HE per capita US$) 3 3 y = 0.2946e0.5796x y = 0.3078e0.5637x 2.5 R2 = 0.8715 2.5 R2 = 0.7606 2 2 1.5 1.5 1 1 0.5 0.5 0 0 2 2.5 3 3.5 4 2 2.5 3 3.5 4 Log10 (GDP per capita US$) Log10 (GDP per capita US$) (continued) within several income bands (tables 11.2 and 11.3).3 Countries spending more than the expenditure frontier were assigned a value of zero even though they might still benefit from additional spending in terms of reaching the target out- come indicator. (11.1) Country gap = Y ­ y, where Y = cebx (if Y > y) and gap = 0 where Y < y where: c and b are constants e is the base of the natural logarithm y is observed log10 (HE per capita US$) of all countries Y is the estimated log10 (HE per capita US$) from frontiers x is observed log10 (GDP per capita US$) of frontiers. 380 Health Financing for Poor People: Resource Mobilization and Risk Sharing FIGURE 11.8 Continued Frontiers = adult male mortality Frontiers = adult female mortality Log10 (HE per capita US$) Log10 (HE per capita US$) 3 3 y = 0.2766e0.5991x y = 0.2193e0.6637x 2.5 2.5 R2 = 0.7548 R2 = 0.8075 2 2 1.5 1.5 1 1 0.5 0.5 0 0 2 2.5 3 3.5 4 2 2.5 3 3.5 4 Log10 (GDP per capita US$) Log10 (GDP per capita US$) Frontiers = maternal mortality Log10 (HE per capita US$) 3 y = 0.2592e0.6108x 2.5 R2 = 0.7603 2 Nonfrontiers 1.5 Frontiers 1 Expon. (frontiers) 0.5 0 2 2.5 3 3.5 4 Log10 (GDP per capita US$) The use of production frontiers to estimate the cost of scaling up provides some insights that can be used to supplement other insights gained from work on elas- ticities of health spending and outcomes, bottom-up costing, and marginal con- straints costing. Notably: · For the 135 countries where GDP per capita is less than US$7,000, around $25 billion to $70 billion of additional spending would be needed to bring the low spenders up to the level of the high performers, depending on the outcome indicator used to establish the frontier. · The best performers on health outcome are not always the highest spenders. Both India and China, where the largest share of the world's poor live, already spend more in terms of total expenditure on health than the frontier spend- ing in the best performing countries. Deficit Financing of Health Care for the Poor 381 FIGURE 11.9 Production Frontiers for Public Expenditure on Health Care Using Best Performance on Various Health Outcomes Frontiers = under-5 mortality Frontiers = HIV prevalence Log10 (HE per capita US$) Log10 (HE per capita US$) 3 3 y = 0.0538e1.0273x y = 0.084e0.9192x 2.5 2.5 R2 = 0.7694 R2 = 0.7236 2 2 1.5 1.5 1 1 0.5 0.5 0 0 2 2.5 3 3.5 4 2 2.5 3 3.5 4 Log10 (GDP per capita US$) Log10 (GDP per capita US$) Frontiers = TB incidence Frontiers = life expectancy Log10 (HE per capita US$) Log10 (HE per capita US$) 3 3 y = 0.191e0.6702x y = 0.0654e0.9652x 2.5 2.5 R2 = 0.7962 R2 = 0.7129 2 2 1.5 1.5 1 1 0.5 0.5 0 0 2 2.5 3 3.5 4 2 2.5 3 3.5 4 Log10 (GDP per capita US$) Log10 (GDP per capita US$) (continued) · The best performers are not evenly distributed across income groups. The best performers for HIV prevalence rates are seen across all GDP levels, while the best performers for tuberculosis are seen only at higher GDP levels (log10 [GDP per capita > 2.9] or GDP per capita > $900). Other health outcomes start having best performers around the GDP levels of log10 (GDP per capita < 2.7) or GDP per capita < $500, Armenia, Azerbaijan, Georgia, and so on. More than anything, the use of production frontiers indicates the developing countries' constraints in scaling up spending needed to accelerate progress toward the MDGs. Since the limit is not the "sky" and since developing coun- tries have to live within realistic budget constraints, this analysis points to a 382 Health Financing for Poor People: Resource Mobilization and Risk Sharing FIGURE 11.9 Continued Frontiers = adult male mortality Frontiers = adult female mortality Log10 (HE per capita US$) Log10 (HE per capita US$) 3 3 y = 0.1436e0.7488x y = 0.0557e1.0205x 2.5 2.5 R2 = 0.7599 R2 = 0.7524 2 2 1.5 1.5 1 1 0.5 0.5 0 0 2 2.5 3 3.5 4 2 2.5 3 3.5 4 Log10 (GDP per capita US$) Log10 (GDP per capita US$) Frontiers = maternal mortality Log10 (HE per capita US$) 3 y = 0.0763e0.9297x 2.5 R2 = 0.6958 2 Nonfrontiers 1.5 Frontiers 1 Expon. (frontiers) 0.5 0 2 2.5 3 3.5 4 Log10 (GDP per capita US$) need for significant targeting and selectivity in additional spending. One way to reduce the total gap in needed spending would be to target specific income ground where the needs might be deemed the greatest. For example: · For the 22 countries where GDP per capita is less than $300, around $350 mil- lion to $1.4 billion of additional spending would be needed to bring the low spenders up to the level of the high performers, depending on the outcome indicator used to establish the frontier. For public expenditure on health, the additional amount would be $150 million to $1 billion. · For the 47 countries where GDP per capita is between $300 and $1,000, around $2 billion to $15 billion of additional spending would be needed to Deficit Financing of Health Care for the Poor 383 TABLE 11.2 Total Health Expenditure Gap in US$ (in Millions) to Achieve Frontier Expenditure Levels (Countries with GDP < US$7,000 per Capita) Frontiers are the Best 20 Percent Performers of Under-5 Adult Adult mortality HIV TB Life male female Maternal GDP per % change prevalence incidence expectancy mortality mortality mortality capita (N) 1990­2000 1999 1999 1997 1997 1997 1995 $0­300 22 $345 $505 $1,137 $1,356 $1,133 $654 $916 $300­1,000 47 $1,947 $3,462 $10,875 $14,670 $11,731 $3,992 $5,134 $1,000­3,000 40 $1,472 $3,891 $4,475 $4,322 $5,110 $2,905 $2,562 $3,000­7,000 26 $28,488 $60,986 $20,958 $13,882 $25,731 $29,728 $15,003 Total 135 $32,252 $68,844 $37,445 $34,230 $43,705 $37,279 $23,615 TABLE 11.3 Public Health Expenditure Gap in US$ (in Millions) to Achieve Frontier Expenditure Levels (Countries with GDP < US$7,000 per Capita) Frontiers are the Best 20 Percent Performers of Under-5 Adult Adult mortality HIV TB Life male female Maternal GDP per % change prevalence incidence expectancy mortality mortality mortality capita (N) 1990­2000 1999 1999 1997 1997 1997 1995 $0­300 22 $154 $407 $1,079 $241 $725 $186 $337 $300­1,000 47 $4,606 $14,211 $26,414 $6,125 $19,537 $5,478 $9,954 $1,000­3,000 40 $675 $3,901 $3,617 $823 $2,613 $972 $1,547 $3,000­7,000 26 $47,864 $86,565 $19,280 $28,657 $21,859 $51,968 $40,807 Total 135 $53,299 $105,084 $50,390 $35,846 $44,734 $58,604 $52,645 bring the low spenders up to the level of the high performers, depending on the outcome indicator used to establish the frontier. For public expenditure on health, the additional amount would be around $5 billion to $26 billion. · For the 40 countries where GDP per capita is between $1,000 and $3,000, around $1.5 billion to $5 billion of additional spending would be needed to bring the low spenders up to the level of the high performers, depending on the outcome indicator used to establish the frontier. For public expenditure on health, the additional amount would be $700 million to $4 billion. · For the 26 countries where GDP per capita is between $3,000 and $7,000, around $14 billion to $61 billion of additional spending would be needed to bring the low spenders up to the level of the high performers, depending on 384 Health Financing for Poor People: Resource Mobilization and Risk Sharing the outcome indicator used to establish the frontier. For public expenditure on health, the additional amount would be $19 billion to $87 billion. The use of efficiency frontiers for this type of analysis has several significant con- straints. Unlike the work using elasticities of health spending and outcomes, production frontiers do not provide any insights into the potential impact of additional spending. The lagging countries could spend up to, and even beyond, the efficiency frontier without improving health outcomes. Although the fron- tier approach sheds some light on the limits to spending, like elasticities using aggregate spending levels, it does not inform the policymaker about the con- tents of those reforms. To understand the policy contents and implementation issues, other tools of investigation are needed. FINANCING THE EXPENDITURE GAP Although the various costing methods provide different estimates for the global funding gap, all of them indicate that additional funding will be needed to accel- erate progress toward the health-related MDGs. Without additional funding, many countries are likely to continue tracking along the path of the past few years, significantly missing the 2015 outcome targets. Other management, orga- nizational, and institutional reforms are necessary as well, but they will not be sufficient without additional funding. Prospects for significant increases in health spending, however, are grim for several reasons. Potential Sources of Funding Several potential sources of additional funding can be identified. They include: · Increase in total health expenditure that will accompany economic growth, even if the share of GDP remains constant · Increase in health expenditure as a share of GDP (that is, reallocation of some public funds from other spending programs toward health and an increase in private health care consumption) · Reallocation of part of the health budget toward priority programs · Donor aid from multilaterals, bilaterals, and nongovernmental organizations · Private sector donation and differential pricing. GDP Growth Health care behaves as a superior good in economic terms--the share of expen- diture increases with GDP--and pubic expenditure on health as a share of GDP also increases with income (World Bank 1993; Schieber and Maeda 1997; WHO 2000; OECD 2001). These trends are demonstrated in figure 11.10. Economic growth is therefore good for health spending. Deficit Financing of Health Care for the Poor 385 FIGURE 11.10 Income and Health Spending Total spending as share of GDP Public share of total spending 13 90 11 9 70 7 50 5 30 3 1 10 5,000 10,000 15,000 20,000 5,000 10,000 15,000 20,000 GDP per capita (1998 international dollars) GDP per capita (1998 international dollars) Unfortunately, the sluggish global economic outlook, with slower growth in the next 12 to 18 months than previously anticipated, will impede poverty reduction and health spending in many developing countries (World Bank 2002b). According to the latest forecasts, global GDP is expected to rise by 2.5 percent. This is higher than during the previous two years but significantly below previous long-term growth rates. High-income countries are expected to grow at about 2.1 percent in 2003. On average, developing countries will grow considerably faster, at 3.9 percent. Health spending in low- and middle-income countries is currently around US$280 billion. Since the elasticity of health spending to income is just over one in low-income countries, annual GDP growth of 3.9 percent would yield US$11 billion in additional funding for the health sector. But such averages mask wide regional differences, with East Asia leading the pack at 6.1 percent, followed by South Asia at 5.4 percent. Other regions are expected to grow less than 4 percent, with Latin America managing a mere 1.8 percent. Outside Asia and Eastern Europe, growth rates in most developing countries are too low to generate any significant increase in health expenditure in the immediate future. Even though global spending increased dramatically during the 20th century, many of the world's poor did not benefit from this increase in overall spending. This is reflected in the large differences in the proportion of national GDP spent on health--from under 1 percent in some countries to 15 percent in the United States. Per capita health expenditures (public and private) vary almost 1,000-fold among countries--from around US$3 to $5 per capita per year in some low- income countries such as Mali to $3,600 in the United States (the ratio would be 225 using PPP-adjusted dollars). Once again, Sub-Saharan Africa and South Asia are the hardest hit in terms of both current spending and dismal prospects for increased spending due to economic growth (see figure 11.11). 386 Health Financing for Poor People: Resource Mobilization and Risk Sharing FIGURE 11.11 Only 11 Percent of Global Spending for 90 Percent of the World's Population Asia 3.5% Americas 3.2% Europe 2.4% Middle East and N. Africa 1.5% Africa 0.4% Developed countries 88.9% Increase in Health Expenditure as a Share of GDP Domestic funds can be raised for the health sector through a number of different mechanisms: general revenues, earmarked social health insurance premiums, community financing, private insurance, and direct user charges. Many low- income countries spend less than 2 percent of GDP, and some even less than 1 percent, on health care. The production frontier for health spending at different income levels indicates that, with good public polices and taxation practices, even poor countries can mobilize and allocate enough financial resources through public channels to reach spending levels of up to 7 percent of GDP. A small increase in the share of health spending in GDP would yield huge gains (a 0.5 to 1.0 percent increase would yield US$25 billion to $50 billion in additional resources for the health sector). But once again the poorest countries in Africa and South Asia would benefit the least for several reasons. Weak taxation capacity prevents many low-income countries from mobilizing more than 5 to 10 percent of GDP through the public sector (Dror and Preker 2002; Preker and oth- ers 2002). Private spending is much greater than public spending at low-income levels (for example, in India, Bangladesh, and many other low-income countries, private spending composes more than 80 percent of total spending). As a result of these factors, low-income countries have little protection against the financial burden of illness even when, as in India, they spend a significant part of their income on health care--6 percent of GDP (Peters and others 2002). As seen in fig- ure 11.12, a large share of the population in some of the world's poorest countries do not benefit from insurance protection against the cost of illness. Reallocation of Health Budget toward Priority Programs One frequently hears rhetoric about mobilizing additional resources through efficiency gains or channeling health expenditure from within the existing health budget toward priority health programs such as child and maternal care or the priority infectious diseases. In reality, this is not an easy policy to imple- Deficit Financing of Health Care for the Poor 387 FIGURE 11.12 Low-Income Countries Have Less Pooling of Revenues Pooled health revenues as % of total 100 80 Share of the world's 1.3 billion people living on less than US$1/day Indonesia 60 indicated by size of the bubbles China Mexico 40 Pakistan 20 India 0 ­10,000 0 10,000 20,000 30,000 40,000 50,000 GNP per capita ment. For example, existing budget categories are often rigid, allowing for little fungibility among spending programs. Strong vested stakeholders will resist hav- ing their budgets channeled to other programs. Any resources that might be freed up through efficiency gains are better absorbed by expanding the scope and activities of the programs in which the gains were made: "Why scrimp and save if you cannot keep the results of your frugality?" (Wilson 1989). It is there- fore prudent not to anticipate additional resources mobilized in this manner. Donor Aid from Multilaterals, Bilaterals, and Nongovernmental Organizations Total Overseas Development Assistance (ODA) on health was estimated at US$6 billion to $7 billion by Working Group 6 of the Macroeconomic Commission on Health. Additional funding from the international donor community is unlikely (WHO 2002a). Annual World Bank lending has remained around US$1 billion to $1.5 billion despite efforts to increase this during the past few years. Although there was much posturing about increasing donor aid for the health sector during creation of the Global Trust Fund for HIV/AIDs, Tuberculosis, and Malaria, replenishment of the fund has been slower than expected and disbursement so far almost nonexistent. 388 Health Financing for Poor People: Resource Mobilization and Risk Sharing FIGURE 11.13 Types of Private Financial Flows US$ billion 350 300 250 200 150 100 50 0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 Foreign direct investments Portfolio equity flows Other private flows Hence donor aid has "hit the wall" in terms of both additional commitments and ability by developing countries to absorb such aid under existing proce- dures. Even an unrealistically optimistic scenario of a 15 percent increase would yield only an additional US$1 billion in donor assistance. Direct Private Sector Investment and Differential Pricing Policies The sagging global economy has also reduced private capital flows to developing countries (see figure 11.13). Net commercial bank lending has turned negative, and foreign direct investment (FDI) flows to developing countries have fallen since their peak in 1999. FDI flows to developing countries peaked at $180 billion in 1999 and have fallen back to the $160 billion range (World Bank 2002b). Although only a small amount of these resources benefit the health sector directly, they often contribute significantly to the health-enhancing environment of a country (for example, roads, clean water, and telecommunications systems). Rising global risk premiums have led to a reversal in debt capital flows. The precarious market conditions have also reduced infrastructure investment sharply. Besides the fall in investment in absolute terms, investors are becoming more selective in choosing their investment destinations. As a result, investment is flowing to countries with better domestic investment climates: good gover- nance, sound institutions, and a system of property rights. These are precisely the conditions that are often lacking in some of the poorest countries in Sub- Saharan Africa and South Asia, making future private sector engagement in those regions less attractive. Deficit Financing of Health Care for the Poor 389 Adopting policies that promote competition is central to raising productivity. Policy barriers to competition in developing countries are common--legal restrictions prevent entry of foreign participants, trade barriers limit import competition, state monopolies protect domestic firms from private sector com- petition, and poorly designed regulatory regimes in privatized industries shun competitors--stifling productivity growth. The same is true in the health sector of developing countries (Harding and Preker 2003). Health care providers in Korea, Malaysia, and Thailand are more productive than in India and China partly because of lower trade restrictions and administrative barriers to entry. Introduction of competition in corporatized and privatized industries is help- ful as well. Private restraints on competition can also have adverse effects on prices for consumers and producers in developing countries and impair their ability to compete in global markets. For example, international cartels such as those in the pharmaceutical industry can tax consumers in developing coun- tries. Ending exemptions to antitrust laws can help reduce private anticompeti- tive practices (Preker and Harding 2003). Corporate social responsibility has also been evoked recently as a way to engage the private sector in helping developing countries (European Commis- sion 2002). Assistance ranges from direct donations in cash and kind by the pharmaceutical industry to differential pricing of products that gives low- income countries access to vaccines, medicines, and supplies at a reduced price. Gross annual revenues in the pharmaceutical industry are currently around US$300 billion. Even a modest 1 to 2 percent "Robin Hood" tax on their rev- enues or differential reduction in prices would yield significant marginal addi- tional resources (US$3 billion to $6 billion). Other industries such as the private health insurance industry, medical equipment industry, and health service providers could be called upon to make an equally modest contribution. CONCLUSIONS The Millennium Development Goals have become a quantitative set of targets for poverty reduction and improvements in health, education, gender equality, the environment, and other aspects of human development. The selected health-related goals and targets were limited in number to help focus national and international attention on a few priority areas: reducing the number of chil- dren who die before their fifth birthday and women who die from complications of childbirth and reversing the spread of HIV/AIDS, malaria, tuberculosis, and other major diseases. This chapter has reviewed several approaches that have been used to estimate the cost of accelerating progress toward achieving the MDGs--the elasticity of out- comes with respect to health spending, health spending production frontiers with respect to outcomes, bottom-up costing, and marginal benefit costing. The chapter concludes that improvements in clinical effectiveness, management capacity, 390 Health Financing for Poor People: Resource Mobilization and Risk Sharing organizational incentives, and institutional environment are necessary elements but are not sufficient in themselves to make progress toward achieving the health-related MDGs. In many low- and middle-income countries, additional money is also necessary. The level of aid needed might be more than the US$25 billion to $50 billion estimated by the research done for the Macroeconomic Commission on Health (WHO 2002a, 2002b, 2002c). The expenditure gap increases dramatically when the scenarios include countries with a GDP above US$1,000. The gap could be almost three times previous estimates, especially if middle-income countries and counties that are already high spenders are included in this aid effort. The additional funding could come from growth in income (which would be automatically translated into higher health spending), a shift of some public spending toward the health sector, increased private consumption of effective care, increased international aid (public and private), and direct foreign invest- ment in the health sector of developing countries. Unfortunately, in many of the worst performing countries, the needed additional resources outstrip both potential domestic and international sources of revenues, especially in the case of South Asia and Sub-Saharan Africa. National governments, households, the international development commu- nity, and the private sector could do better in mobilizing these additional resources and provide some estimates of the order of such action. In many low- income countries, however, the targets appear unattainable. Caution should be exercised not to discourage countries that could make modest progress toward better health outcomes through the fervor of the current debate on achieving the MDGs. The argument that health levels are always "lower than they need be" could be said of every country in the world. The diseases would be different, but even in the United States, there are diseases for which technology is available to improve health but whose cost is considered prohibitive for parts of the popula- tion, given competing claims on scarce resources. Donor aid is also unlikely to be a solution by itself but may be important at the margins if it is well targeted to the poor and if it contributes to systemic reforms. This suggests that any analysis should not assume away the problem of governments' inability to allocate resources and deliver services effectively. Indeed, donor support should tie in with government performance in terms of resource allocation and outcomes. Much more work is needed to fully under- stand the variation in performance of different countries and their health sys- tems in achieving maximum value for money. Acknowledgments: The authors are grateful to Flavia Bustreo, Mariam Claeson, Shantayanan Devarajan, Dean Jamison, Margaret Miller, Philip Musgrove, Juan Rovira, Eric V. Swanson, and Adam Wagstaff for their reviews and helpful guidance. The findings, interpretations, and con- clusions expressed in the chapter are entirely those of the authors and do not represent the views of the World Bank, its Executive Directors, or the countries they represent. Deficit Financing of Health Care for the Poor 391 NOTES 1. The proposal to develop such a set of goals was first made by the Ministers of Develop- ment from the Development Assistance Committee (DAC) of the Organisation for Eco- nomic Co-operation and Development (OECD) in 1995 (OECD 1996). The General Assembly of the United Nations incorporated these goals in the Millennium Declara- tion in September 2000, while setting new targets for reducing the proportion of peo- ple suffering from hunger, increasing access to improved water sources, improving the lives of slum dwellers, and reversing the spread of HIV/AIDS, malaria, tuberculosis, and other major diseases (United Nations 2000, 2001). 2. Other outcome measures such as birthrate and population composition (for example, proportion of the population under 15 years) are not significantly associated with health spending (Gbesemete and Gerdtham 1992). 3. We ultimately used exponential regressions in log scale for both health expenditure and GDP, after observing patterns of relative fit for simple linear regressions, power regressions, polynomial regressions, as well as in ordinary scale for both health expen- diture and GDP. Log scale for both GDP and health expenditure was used because con- verting them into log scale gave a higher R-squared than the one in ordinary scale. Exponential regressions were chosen because the fitted line showed higher R-squared (0.91) than R-squared of power regression line (0.87) and polynomial regression line (0.83). They were also chosen because the exponential regressions gave a higher inter- cept than simple linear regressions, where the values of R-squared of these two regres- sions were almost the same (0.91). In addition, we fitted the exponential regression line by controlling for levels of female primary education, but this did not produce a significantly different result. REFERENCES Aigner, D. J., C. A. Lovell, and P. Schmidt. 1977. "Formation and Estimation of Stochastic Frontier Production Function Models." Journal of Econometrics 6: 21­37. Alderman, H., H. Appleton, L. Haddad, L. Song, and Y. Yohannes. 2000. Reducing Child Malnutrition: How Far Does Income Growth Take Us? Washington, D.C.: World Bank. Anderson, G., and J.-P. Poullier. 1999. "Health Spending, Access, and Outcomes: Trends in Industrialized Countries." Health Affairs 18(3): 172­92. Anell, A., and M. Willis. 2000. "International Comparisons of Health Care Systems Using Resource Profiles." Bulletin of the World Health Organization 78(6): 770­78. Barnum, H. 1987. "Evaluating Healthy Days of Life Gained from Health Projects." Social Science and Medicine 24(10): 833­41. Barnum, H. N. 1986. "Cost Savings from Alternative Treatments for Tuberculosis." Social Science and Medicine 23(9): 847­50. Barnum, H. N., D. Tarantola, and I. F. Setiady. 1980. "Cost-Effectiveness of an Immuniza- tion Programme in Indonesia." Bulletin of the World Health Organization 58(3): 499­503. Battese, G. E., and T. J. Coelli. 1995. "A Model of Technical Inefficiency Effects in a Sto- chastic Frontier Production for Panal Data." Empirical Economics 20: 325­32. 392 Health Financing for Poor People: Resource Mobilization and Risk Sharing Blendon, R. J., M. Kim, and J. M. Benson. 2001. "The Public Versus the World Health Orga- nization on Health System Performance." Health Affairs 20(3): 10­20. Carrin, G., R. Zeramdini, P. Musgrove, J-P. Poullier, N. Valentine, and K. Xu. 2001. The Impact of the Degree of Risk-Sharing in Health Financing on Health System Attainment. Report submitted to Working Group 3 of the Commission on Macroeconomics and Health, Jeffrey D. Sachs (Chairman). WHO, Geneva. Claeson, M., and R. J. Waldman. 2000. "The Evolution of Child Health Programmes in Developing Countries: From Targeting Diseases to Targeting People." Bulletin of the World Health Organization 78(10): 1234­45. Claeson, M., T. Mawji, and C. Walker. 2000. Investing in the Best Buys: A Review of the Health, Nutrition, and Population Portfolio, FY1993­99. Washington, D.C.: World Bank. Coelli, T. J., D. S. Prasada-Rao, and G. E. Battese. 1998. An Introduction to Efficiency and Pro- ductivity Analysis. Boston: Kluwer. Devarajan S., M. J. Miller, and E. V. Swanson. 2002. Goals for Development: History, Prospects, and Costs. Washington, D.C.: World Bank. Dror, D., and A. S. Preker, eds. 2002. Social Re-Insurance: A New Approach to Sustainable Community Health Financing. World Bank, HNP Discussion Paper. Washington, D.C. European Commission. 2002. Promoting a European Framework for Corporate Social Responsi- bility: Green Paper. Brussels. Evans, D. B. 1990a. "Principles Involved in Costing." Medical Journal of Australia 153(suppl.): S10­2. ------. 1990b. "What Is Cost-Effectiveness Analysis?" Medical Journal of Australia 153(suppl.): S7­9. Farrel, M. J. 1957. "The Measurement of Production Efficiency." Journal of Royal Society, Series A, 120: 253­90. Filmer, D., and L. Pritchett. 1999. "The Impact of Public Spending on Health: Does Money Matter?" Social Science and Medicine 49(10): 1309­23. Filmer D., J. Hammer, and L. Pritchett. 2000. "Weak Links in the Chain: A Diagnosis of Health Policy in Poor Countries." World Bank Observer 15(2): 199­224. ------. 2002. "Weak Links in the Chain: A Prescription for Health Policy in Poor Coun- tries." World Bank Observer 17(1): 47­66. Garrison, L., and N. Mccall. 1990. Measuring the Impact of Medical Technology on Health Care Costs: Report for the Prospective Payment Assessment Commission. Washington, D.C.: U.S. Congress. Gbesemete, K. P., and U. G. Gerdtham. 1992. "Determinants of Health Care Expenditure in Africa: A Cross-Sectional Study." World Development 20(2): 303­8. Gwatkin, D. R. 2000. "Health Inequalities and the Health of the Poor: What Do We Know? What Can We Do?" Bulletin of the World Health Organization 78(1): 3­18. Gwatkin, D. R., and P. Heuveline. 1997. "Improving the Health of the World's Poor." BMJ (London) 315(7107): 497­98. Hammer, J. 1996. Economic Analysis for Health Projects. World Bank, HNP Discussion Paper. Washington, D.C. Deficit Financing of Health Care for the Poor 393 Harding, A., and A .S. Preker, eds. 2003. Private Participation in Health Services. World Bank, HNP Discussion Paper. Washington, D.C. Hauck, K., P. C. Smith, and M. Goddard. 2002. Priority Setting for Health. Washington, D.C.: World Bank. Jack,W. 2000. "Public Spending on Health Care: How Are Different Criteria Related? A Sec- ond Opinion." Health Policy 53(1): 61­67. Jakab, M., A. S. Preker, C. Krishnan, P. Schneider, F. Diop, J. Jütting, A. Gumber, K. Ranson, and S. Supakankunti. 2001. Social Inclusion and Financial Protection through Community Financing: Initial Results from Five Household Surveys. Report submitted to Working Group 3 of the Commission on Macroeconomics and Health, Jeffrey D. Sachs (Chair- man). WHO, Geneva. Jamison, D. T., W. H. Mosley, A. Measham, and J. L. Bobadilla, eds. 1993. Disease Control Priorities in Developing Countries. Washington, D.C.: World Bank. Kumaranayake, L., C. Kurowski, and L. Conteh. 2001. Costs of Scaling-up Priority Health Interventions in Low and Selected Middle Income Countries: Methodology and Estimates. Geneva: World Health Organization. Lambrechts, T., J. Bryce, and V. Orinda. 1999. "Integrated Management of Childhood Ill- ness: A Summary of First Experiences." Bulletin of the World Health Organization 77(7): 582­94. Lewin, A. Y., and C. A. Knox. 1990. "Frontier Analysis: Parametric and Nonparametric Approaches." Journal of Econometrics (Netherlands) 46: 3­245. Mills, A. 1985a. "Economic Evaluation of Health Programmes: Application of the Princi- ples in Developing Countries." World Health Statistics Quarterly 38(4): 368­82. ------. 1985b. "Survey and Examples of Economic Evaluation of Health Programmes in Developing Countries." World Health Statistics Quarterly 38(4): 402­31. Murray C. J., K. Kawabata, and N. Valentine. 2001. "People's Experience Versus People's Expectations." Health Affairs 20(3): 21­24. Murray, C. J., D. B. Evans, A. Acharya, and R. M. P. M. Baltussen. 2000. "Development of WHO Guidelines on Generalized Cost-Effectiveness Analysis." Health Economics 9(3): 235­51. Murray, C. J., J. Frenk, D. Evans, K. Kawabata, A. Lopez, and O. Adams. 2001. "Science or Marketing at WHO? A Response to Williams." Health Economics 10(4): 277­82; discus- sion 283­85. Musgrove, P. 1996. Public and Private Roles in Health: Theory and Financing Patterns. Wash- ington, D.C.: World Bank. ------. 1999. "Public Spending on Health Care: How Are Different Criteria Related?" Health Policy 47(3): 207­23. ------. 2000a. "Cost-Effectiveness as a Criterion for Public Spending on Health: A Reply to William Jack's `Second Opinion.'" Health Policy 54(3): 229­33. ------. 2000b. "A Critical Review of `A Critical Review': The Methodology of the 1993 World Development Report, `Investing in Health.'" Health Policy Plan 15(1): 110­15. Navarro, V. 2000. "Assessment of the World Health Report 2000." The Lancet 356(9241): 1598­601. 394 Health Financing for Poor People: Resource Mobilization and Risk Sharing ------. 2001. "Science or Ideology? A Response to Murray and Frenk." International Journal of Health Services 31(4): 875­80. Newhouse, J. P., and L. J. Friedlander. 1980. "The Relationship between Medical Resources and Measures of Health: Some Additional Evidence." Journal of Human Resouces 15(2): 200­18. OECD (Organisation for Economic Co-operation and Development). 1992. The Reform of Health Care: A Comparative Analysis of Seven OECD Countries. Paris. ------. 1994. The Reform of Health Care Systems: A Review of Seventeen OECD Countries. Paris. ------. 1996. Shaping the 21st Century: The Contribution of Development Co-operation. Paris. ------. 2001. OECD Health Data, 2000. Paris. ------. 2002. Measuring Up: Improving Health Systems Performance in the OECD. Paris. Peabody, J., ed. 1999. Policy and Health: Implications for Development in Asia. Cambridge: Cambridge University Press. Peters, D. H., A. S. Yazbeck, R. R. Sharma, G. N. V. Ramana, L. Pritchett, and A. Wagstaff, eds. 2002. Better Health Systems for India's Poor: Finding, Analysis, and Options. World Bank, HNP Discussion Paper. Washington, D.C. Preker, A. S., and A. Harding, eds. 2003. Innovations in Health Service Delivery: The Corpora- tization of Public Hospitals. World Bank, HNP Discussion Paper. Washington, D.C. Preker A. S., J Langenbrunner, and E Suzuki. 2001. The Global Expenditure Gap: Securing Financial Protection and Access to Health Care for the Poor. Report submitted to Working Group 3 of the Commission on Macroeconomics and Health, Jeffrey D. Sachs (Chair- man). WHO, Geneva. Preker, A. S., J. Langenbrunner, M. Jakab, and C. Baeza. 2001. Resource Allocation and Pur- chasing (RAP) Arrangements that Benefit the Poor and Excluded Groups. World Bank, HNP Discussion Paper. Washington, D.C. Preker, A. S., G. Carrrin, D. Dror, M. Jakab, W. Hsiao, and D. Arhin-Tenkorang. 2001. Role of Communities in Resource Mobilization and Risk Sharing: A Synthesis Report. Geneva: World Health Organization. ------. 2002. "Effectiveness of Community Health Financing in Meeting the Cost of Ill- ness." Bulletin of the World Health Organization 8(2): 143­50. Prescott, N., and D. De Ferranti. 1985. "The Analysis and Assessment of Health Programs." Social Science and Medicine 20(12): 1235­40. Pritchett, L., and L. H. Summers. 1996. "Wealthier Is Healthier." Journal of Human Resources 31(4): 841­68. Rivlin, M. M. 2000. "Why the Fair Innings Argument Is Not Persuasive." BMC Medical Ethics 1(1): 1. Schieber, G., and A. Maeda. 1997. "A Curmudgeon's Guide to Health Care Financing in Developing Countries." In G. Schieber, ed., Innovations in Health Care Financing. Washing- ton D.C.: World Bank. (Proceedings of a World Bank Conference, March 10­11, 1997.) Schieber G., J.-P. Poullier, and L. Greenwald. 1991. "Health Systems in Twenty-Four Coun- tries." Health Affairs 10(3): 22­38. Deficit Financing of Health Care for the Poor 395 Soucat A., W. Lerberghe, F. Diop, and R. Knippenberg. 1992. Buying Results, Budgeting for Bottlenecks: The New Performance Frontier. World Bank, HNP Discussion Paper. Washing- ton, D.C. Tulloch, J. 1999. "Integrated Approach to Child Health in Developing Countries." The Lancet 354(suppl. 2): 16­20. United Nations. 2000. A Better World for All: Progress toward the International Development Goals. New York: United Nations. ------. 2001. Road Map towards the Implementation of the United Nations Millennium Decla- ration. New York. Van Doorslaer, E., A. Wagstaff, H. Van Der Burg, T. Christiansen, D. De Graeve, I. Duch- esne, U. G. Gerdtham, M. Gerfin, J. Geurts, L. Gross, U. Hakkinen, J. John, J. Klavus, R. E. Leu, B. Nolan, O. O'Donnell, C. Propper, F. Puffer, M. Schellhorn, G. Sundberg, and O. Winkelhake. 2000. "Equity in the Delivery of Health Care in Europe and the U.S." Journal of Health Economics 19(5): 553­83. Wagstaff, A. 1989. "Estimating Efficiency in the Hospital Sector: A Comparison of Three Statistical Cost Frontier Models." Applied Economics 21: 659­72. ------. 2001. "Economics, Health, and Development: Some Ethical Dilemmas Facing the World Bank and the International Community." Journal of Medical Ethics 27(4): 262­67. ------. 2002a. Economic Growth and Government Health Spending: How Far Will They Take Us towards the Health MDGs? Washington, D.C.: World Bank. ------. 2002b. Intersectoral Synergies and the Health MDGs: Preliminary Cross-Country Find- ings, Collaboration, and Policy Simulations. Washington, D.C.: World Bank. ------. 2002c. "Poverty and Health Sector Inequalities." Bulletin of the World Health Orga- nization 80(2): 97­105. ------. 2002d. "Reflections on and Alternatives to WHO's Fairness of Financial Contribu- tion Index." Health Economics 11(2): 103­15. Wagstaff, A., and G. Lopez. 1996. "Hospital Costs in Catalonia: A Stochastic Frontier Analysis." Applied Economics 3: 471­74. Wang, J., D. T. Jamison, E. Bos, A. S. Preker, and J. Peabody. 1999. Measuring Country Per- formance on Health: Selected Indicators for 115 Countries. Washington, D.C.: World Bank. Weissman, E. 2001. IMCI Costing Tool. Geneva: World Health Organization. Weissman, E., F. Bustreo, K. Krasovec, M. Claeson, and H. Troedson 2001. Estimating the Cost of Basic Health Services for Children: The Experience from Nepal and Nigeria. New York and Geneva: UNICEF and World Health Organization. Williams, A. 2001. "Science or Marketing at WHO? A Commentary on `World Health 2000.'" Health Economics 10(2): 93­100. Williams, M. 1997. "Rationing Health Care: Can a `Fair Innings' Ever Be Fair?" BMJ (Lon- don) 314(7079): 514. Wilson, J. Q. 1989. Bureaucracy: What Government Agencies Do and Why They Do It. New York: Basic Books. World Bank. 1993. World Development Report 1993: Investing in Health. New York: Oxford University Press. 396 Health Financing for Poor People: Resource Mobilization and Risk Sharing ------. 1997. Sector Strategy for HNP. Washington, D.C. ------. 2000. World Development Report 2000/2001: Attacking Poverty. New York: Oxford University Press. ------. 2002a. Development Effectiveness and Scaling Up: Lessons and Challenges from Case Studies. Washington, D.C. ------. 2002b. Global Economic Prospects and the Developing Countries 2003. Washington, D.C. ------. 2002c. World Development Indicators. Washington, D.C. WHO (World Health Organization). 2000. World Health Report 2000: Health Systems-- Improving Performance. Geneva. ------. 2001. Macroeconomics and Health: Investing in Health for Economic Development. Geneva. ------. 2002a. Development Assistance and Health. Geneva. ------. 2002b. Improving Health Outcomes of the Poor. Geneva. ------. 2002c. Mobilisation of Domestic Resources for Health. Geneva. CHAPTER 12 Impact of Risk Sharing on the Attainment of Health System Goals Guy Carrin, Riadh Zeramdini, Philip Musgrove, Jean-Pierre Poullier, Nicole Valentine, Ke Xu Abstract: A simple econometric analysis is undertaken concerning the impact of the degree of risk sharing in countries' health-financing organizations on the goals of the health system. Those goals are defined in the World Health Report 2000--the level of health and its distribution across the population, the level of responsiveness and its distribution across the population, and fair financing. The degree of risk sharing varies according to whether countries have a universal coverage system, financed via social health insurance or general taxation, or systems with less-developed coverage. We undertook a classifica- tion of countries according to the degree of risk sharing, based primarily on the health- financing legislation of the World Health Organization's 191 member states and on its database of Health System Profiles. The results obtained give empirical support to the hypothesis that the degree of risk sharing in health-financing organizations positively affects health system attainment, as measured by the five goals indicators. T here are important linkages between what health systems can achieve in terms of preset goals and the functions that they undertake. The World Health Report (WHR) 2000 has designed a coherent framework for analyzing these linkages (WHO 2000; see also Murray and Frenk 2000). In this chapter, we specifically address the health-financing function of pooling of resources and how it influences health systems attainment. One essential question is whether health-financing organizations provide sufficient financial risk protection for the population. People's access to health services depends on this protection. Health-financing organizations that do not include the low-income population groups, for instance, will lead to many individuals' being unable to pay for care. The extent to which these population groups are effectively included in risk- sharing arrangements is therefore likely to affect a goal such as the equality of health status. Health-financing organizations may also be more or less engaged in purchasing an adequate package of health services for the entire population. In this sense, they may affect the average level of access to good care and there- fore indirectly have an impact upon the average level of health. Apart from the level and distribution of health status, other goals may be considered. In the next section, we give an overview of the goals of health systems as proposed by the WHR 2000 and discuss how they relate to the functions of these systems. 398 Health Financing for Poor People: Resource Mobilization and Risk Sharing This chapter's main purpose is to undertake a simple econometric analysis of the impact of the degree of risk sharing in countries' health-financing organiza- tion on the goals of the health system. The degree of risk sharing will vary accord- ing to whether countries have a universal coverage system, financed via social health insurance or general taxation, or systems with less-developed coverage. For the latter, specific population groups may be covered by variants of social health insurance and general taxation. Risk sharing via community health- financing schemes could not be considered for lack of data at the national level. In preparation of the econometric analysis, we turn to the specific linkage between the goals and the health-financing function. Then we classify the health-financing organization of 191 countries by the degree of risk sharing. This classification will help in defining the variables that measure risk sharing and will be used in the econometric analysis. A descriptive data analysis of the endogenous variables is provided further. Finally, the specification of the econo- metric models and estimation results are presented. HEALTH SYSTEM GOALS AND FUNCTIONS IN A NUTSHELL The framework, as presented in the WHR 2000, defines a set of goals or objec- tives and includes ways to measure the achievement toward these goals. Of course, to obtain these achievements, health systems do need to carry out a number of functions. Below we address both goals and functions. The goals considered are good health, responsiveness, and fair financing. Good health is approached in two ways. One is by striving for the best attainable average level for the entire population. The other is by minimizing the differ- ences in health status among individuals and groups. In other words, the distri- bution of health as well as the level of health matters. Note that health is measured via disability-adjusted life expectancy,1 whereby account is taken of time lived with a disability. Responsiveness measures how the health system per- forms relative to nonhealth aspects of provided health services. Responsiveness captures the extent to which the health system is client-oriented and treats peo- ple with respect. Respect for people includes the following aspects: respect for the dignity of the person, confidentiality, and autonomy. Within client orienta- tion, we consider prompt attention, the quality of the amenities, the access to social support networks, and the choice of provider. Note that the distinction between overall level and distribution across the population also applies to responsiveness. Fair financing requires that the health expenditure of house- holds be distributed according to an individual's ability to pay rather than his or her risk of illness. In a fairly financed system, everyone should be financially pro- tected. It is crucial therefore that health systems rely as fully as possible on pre- paid contributions that are unrelated to individual illness or utilization. It is clear that when analyzing fair financing, we are concerned with distributive aspects only. We thus obtain five objectives: the level and distribution of health, Impact of Risk Sharing on the Attainment of Health System Goals 399 the level and distribution of responsiveness, and fair financing. Measurements have been designed so as to quantify the achievement with respect to each of these objectives. (See WHO 2000 for a summary of the methods.)2 We further consider four main functions of the health system: the delivery of health services, the creation of resources for health (investment in people, build- ings, and equipment), health financing (raising, pooling, and allocating the rev- enues to purchase health services), and stewardship. The latter refers to a government's responsibility for the general health of its population. The stew- ardship function is of special importance as it has an impact on the way the other three functions are carried out. Work is currently underway at the World Health Organization (WHO) to define indicators for the various functions so their possible impact on goal achievement can be measured. This chapter can be seen as an element of this par- ticular work in that it focuses on the nature of risk sharing in the world's different health-financing systems and its possible impact on the goals as defined above. THE ORGANIZATIONAL FORM OF HEALTH FINANCING AND ITS LINK TO THE ATTAINMENT OF HEALTH SYSTEM GOALS A crucial concept in health financing is that of pooling. Pooling is defined as the "accumulation and management of revenues in such a way as to ensure that the risk of having to pay for health care is borne by all members of the pool and not by each contributor individually" (WHO 2000, p. 96). The larger the degree of pooling, the less people will have to bear the financial consequences of their own health risks. Health-financing systems encompass various degrees of risk sharing. There are two major ways to ensure financial risk protection for a nation's entire popula- tion. One is a system whereby general taxation (GT) is the main source of financ- ing health services. Services are usually provided by a network of public and contracted private providers, often referred to as a national health service. The second is social health insurance (SHI), whereby workers, enterprises, and gov- ernment pay financial contributions. The base for workers' and enterprises' con- tributions is usually a worker's salary. Social health insurance either owns its own provider networks, works with accredited private providers, or combines the approaches. In principle, both systems pool all of the population's risks, with contributions that are delinked from individual risks. This approach theoretically avoids exposing individuals to insufficient or no access to the health care they need. These systems are often denoted as universal coverage systems, but finan- cial protection may still be judged inadequate in a number of these systems. There are also systems with no explicit reference to overall coverage of the population. These include mixed health-financing systems, in which some part of the population is partially covered via general taxation and another part is covered by health insurance schemes. The latter may address specific groups 400 Health Financing for Poor People: Resource Mobilization and Risk Sharing only. Still, they may practice full pooling among their members and define health insurance contributions according to capacity to pay rather than individ- ual health risks. In other words, these schemes may apply a community rating (as a social health insurance scheme does) but for specific groups only. Such schemes may include voluntary private insurance arrangements, mutual health funds, and enterprise-based and community health insurance. Finally, some countries do finance health services via GT but offer only incomplete coverage. For the purpose of this chapter, we will say that countries that have achieved or are near to universal coverage, and use either general taxation or social health insurance, enjoy systems with advanced risk-sharing. Such schemes allow for more equal access to health services among individuals. In addition, such schemes generally better define an adequate package of health services to which citizens are entitled. Countries with mixed health-financing systems will be associated with medium risk-sharing. The countries with general taxation systems that incompletely cover the population are then associated with low risk-sharing. In this chapter, we will investigate whether larger degrees of risk sharing have a beneficial impact on the five indicators of goal achievement. ORGANIZATION OF HEALTH FINANCING IN THE WORLD In table 12.3 (see appendix), countries are classified according to the criterion of risk sharing as defined above, based on the health care-financing legislation of WHO's 191 member states. Our main source for this revision was the publi- cation Social Security Programs throughout the World, provided by the U.S. Social Security Administration (1999). However, for 52 countries, insufficient infor- mation or none was given. For the latter group, and in order to identify the category of health-financing system, WHO's database of Health System Profiles (http://www.who.int/country_profiles/main.cfm) and other selected publica- tions were used (Nolan and Turbat 1995; see also the Web site of the Center for International Health Information http://www.cihi.com). In table 12.3, approximately 40 percent of the countries are characterized as advanced risk-sharing systems, having either a general taxation system (50 countries) or a social health insurance scheme (30 countries) that covers nearly the entire population. The 61 countries with medium risk-sharing are further classified into three main variants. In the first variant, health insurance covers all employees and the self-employed, though subject to a number of exclusions.3 The second variant covers only employees. The third covers specific groups only, using, for instance, mutual health funds and enterprise-based health insurance for particular categories of workers. In these three variants, there are 9, 20, and 32 countries, respectively. Finally, 50 countries are classified among those with low risk-sharing. These countries are generally characterized by underfinanced health systems as compared with the health needs of the population. It is admit- ted, however, that for a number of countries, the proposed classification is not Impact of Risk Sharing on the Attainment of Health System Goals 401 final, due to incomplete or absent information on the size and structure of the eligible population effectively covered by the health-financing system. We can also rank countries according to the category of risk sharing and the per- centage share of public health expenditure in total health expenditure.4 The three categories considered are shares between 75 and 100 percent, between 50 and 75 percent, and below 50 percent. We use the latter ratio as a simple quantitative indi- cator of the system's degree of financial risk protection. In fact, of the countries with advanced risk-sharing, 74 out of 80 have a ratio above 50 percent and 41 have a ratio above 75 percent. Of the countries with medium risk-sharing, only 3 out of 61 have a ratio above 75 percent. For countries with low risk-sharing, a tilt toward low ratios would be expected. However, for 9 out of 50 countries with low risk- sharing, ratios above 75 percent are reported, which is surprising. However, it is recognized in the WHR 2000 that quite a number of countries have incomplete data and mixed degrees of reliability,5 which may partly explain this finding. In addition, it is interesting to rank countries according to the category of risk sharing and to the income level, as measured by the 1998 gross domestic prod- uct (GDP) per capita (in U.S. dollars). Of the 80 countries with advanced risk- sharing, 20 belong to the category of upper middle-income countries, and 34 to the high-income category. Most countries with low to medium risk-sharing belong to the low-income and lower middle-income categories. In this set of countries, only Andorra and the United States belong to the upper middle- income or high-income category. MODELING THE IMPACT OF THE ORGANIZATIONAL FORM OF HEALTH FINANCING ON HEALTH SYSTEM ATTAINMENT INDICATORS Descriptive Data Analysis As a prelude to the econometric analysis, descriptive statistics for the five health attainment indexes are computed. The health attainment indexes are the disability-adjusted life expectancy (DALE), the index of level of responsiveness (IR), the index of fairness of financial contribution (IFFC), the index of distribu- tion of responsiveness (IRD), and the index of equality of child survival (IECS). All data used originate from the Statistical Appendix of WHR 2000. In table 12.1, statistics related to all countries that have observations on the indexes are pre- sented. In table 12.2, however, countries whose risk sharing classification is uncertain are removed from the samples. The indexes are classified according to the category of risk sharing of countries' health-financing organizations. We present the mean, coefficient of variation, minimum, and maximum. First, a general tendency is that the means of the indi- cators are larger the greater the degree of risk sharing. One exception is in table 12.1, in which the mean fair-financing index for countries with advanced and medium risk-sharing is smaller than that of the countries with low risk-sharing. 402 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 12.1 Descriptive Statistics (Full Samples) Disability Index of Index of Index of Index of adjusted level of fairness of distribution equality life-expectancy responsiveness financial contribution of responsiveness of child survival Statistics (DALE) (IR) (IFFC) (IRD) (IECS) Total sample Mean 56.8262 0.5165 0.8730 0.8967 0.6659 CVa 0.2165 0.1542 0.1203 0.0969 0.2878 Min 25.9000 0.3740 0.6230 0.7230 0.2450 Max 74.5000 0.6880 0.9920 0.9999 0.9990 Number of observations 191 30 21 33 58 Countries with advanced risk-sharing Mean 66.0725 0.5849 0.8732 0.9772 0.9296 CV 0.0755 0.1272 0.0643 0.0252 0.1490 Min 52.3000 0.4430 0.8020 0.9180 0.6320 Max 74.5000 0.6880 0.9390 0.9999 0.9990 Number of observations 80 8 5 9 7 Of which countries with social health insurance Mean 68.5267 0.5452 0.8945 0.9715 0.9990 CV 0.0552 0.1150 0.0704 0.0290 0 Min 62.2000 0.4430 0.8500 0.9180 0.9990 Max 74.5000 0.6120 0.9390 0.9960 0.9990 Number of observations 30 5 2 6 4 Of which countries with general taxation Mean 64.6000 0.6510 0.8590 0.9886 0.8370 CV 0.0785 0.0492 0.0694 0.0128 0.2237 Min 52.3000 0.6320 0.8020 0.9750 0.6320 Max 73.0000 0.6880 0.9210 0.9999 0.9990 Number of observations 50 3 3 3 3 Countries with medium risk-sharing Mean 52.9033 0.5153 0.8623 0.8846 0.6792 CV 0.2152 0.1109 0.1463 0.0932 0.2320 Min 29.1000 0.4180 0.6230 0.7230 0.2610 Max 72.3000 0.6230 0.9920 0.9860 0.9660 Number of observations 61 16 11 17 34 Countries with low risk-sharing Mean 46.8180 0.4285 0.8962 0.8227 0.5309 CV 0.2411 0.1165 0.1183 0.0847 0.2816 Min 25.9000 0.3740 0.7140 0.7280 0.2450 Max 66.7000 0.4940 0.9610 0.9490 0.7850 Number of observations 50 6 5 7 17 a. CV is the coefficient of variation. Impact of Risk Sharing on the Attainment of Health System Goals 403 However, in table 12.2, the mean IFFC for countries with advanced risk-sharing exceeds that for countries with low risk-sharing. Second, using the restricted samples (table 12.2), the coefficients of variation (CV) indicate that, except in the case of IR, there is a lower relative dispersion around the mean in countries with advanced risk-sharing than in countries with medium risk-sharing. The lat- ter is consistent with the fact that we have defined three subgroups with differ- ent degrees of risk sharing within the set of countries with medium risk-sharing. Notice also that in three cases (fair financing, distribution of responsiveness, and distribution of health), countries in the low risk-sharing category show lower coefficients of variation than those for the countries with medium risk-sharing. It stands to reason that the low risk-sharing category of countries is likely to be more homogeneous than the group of countries with medium risk-sharing. Except for the value related to IR, the coefficients of variation are higher, how- ever, when compared with the CV of countries with advanced risk-sharing. Specification of the Basic Model A basic model is constructed, with the main objective to examine the degree to which risk sharing has a beneficial impact on five indicators of health systems attainment. These are the indexes of DALE, IR, IFFC, IRD, and IECS. Only the observed data for these indicators were included in the analysis (WHO 2000). The independent variables were divided into three groups based on the degree to which they provided risk-sharing arrangements. As discussed above, countries belong either to the advanced risk-sharing category, or to the medium risk-sharing or low-risk-sharing category. This classification allows us to define the three main organizational dummy variables: DARS = 1 when a country belongs to the set of advanced risk-sharing systems and 0 otherwise; DMRS = 1 when a country belongs to the set of medium risk-sharing systems and 0 otherwise; DSHI = 1 when an country belonging to the advanced risk-sharing category has a social health insurance system and 0 otherwise. We used the following basic specification for the impact of risk sharing on the level of health: (12.1) Ln (80 ­ DALE) = a1 + b1 Ln HEC + c1 Ln EDU + d1 DARS. The dependent variable is the logarithm of the difference between the observed DALE and a maximum. HEC refers to the health expenditure per capita (in US$). EDU refers to the educational attainment in society and is measured by the primary enrollment. Both HEC and EDU are expected to raise DALE and so to be negatively related to the distance of DALE from the maximum of 80. In other words, the higher HEC and EDU, the closer one gets to the maximum DALE. The hypothesis is further that advanced risk-sharing (DARS = 1) is associ- ated with a more comprehensive definition of the benefit package of health ser- vices to which citizens are entitled, which translates into increased overall level of health. 404 Health Financing for Poor People: Resource Mobilization and Risk Sharing TABLE 12.2 Descriptive Statistics (Restricted Samples) Disability Index of Index of Index of Index of adjusted level of fairness of distribution equality life-expectancy responsiveness financial contribution of responsiveness of child survival Statistics (DALE) (IR) (IFFC) (IRD) (IECS) Total sample Mean 58.0588 0.5165 0.8721 0.8967 0.6843 CVa 0.20840 0.1542 0.1233 0.0969 0.2636 Min 25.9000 0.3740 0.6230 0.7230 0.2610 Max 74.5000 0.6880 0.9920 0.9999 0.9990 Number of observations 160 30 19 33 52 Countries with advanced risk-sharing Mean 67.1179 0.5849 0.8910 0.9772 0.9378 CV 0.0645 0.1272 0.0513 0.0252 0.1598 Min 56.3000 0.4430 0.8500 0.9180 0.6320 Max 74.5000 0.6880 0.9390 0.9999 0.9990 Number of observations 67 8 4 9 6 Of which countries with social health insurance Mean 68.5267 0.5452 0.8945 0.9715 0.9990 CV 0.0646 0.1150 0.0703 0.0290 0 Min 62.2000 0.4430 0.8500 0.9180 0.9990 Max 74.5000 0.6120 0.9390 0.9960 0.9990 Number of observations 30 5 2 6 4 Of which countries with general taxation Mean 65.9757 0.6510 0.8875 0.9886 0.8155 CV 0.0674 0.0492 0.0534 0.0128 0.3183 Min 56.3000 0.6320 0.8540 0.9750 0.6320 Max 73.0000 0.6880 0.9210 0.9990 0.9990 Number of observations 37 3 2 3 2 Countries with medium risk-sharing Mean 53.7596 0.5153 0.8623 0.8846 0.6849 CV 0.2081 0.1109 0.1464 0.0932 0.2282 Min 29.1000 0.4180 0.6230 0.7230 0.2610 Max 72.3000 0.6230 0.9920 0.9860 0.9660 Number of observations 57 16 11 17 33 Countries with low risk-sharing Mean 48.0056 0.4285 0.8800 0.8227 0.5655 CV 0.2452 0.1165 0.1307 0.0847 0.2258 Min 25.9000 0.3740 0.7140 0.7280 0.3360 Max 66.7000 0.4940 0.9590 0.9490 0.7850 Number of observations 36 6 4 7 13 a. CV is the coefficient of variation. Impact of Risk Sharing on the Attainment of Health System Goals 405 Variants of the basic model are also tested. One tests whether social health insurance has a specific impact on the health level. A dummy variable DSHI, equal to 1 when the country has a social health insurance scheme and 0 other- wise, will be added to the explanatory variables of equation 12.1. If we reason that, on average, general taxation and social health insurance schemes cover similar population groups with similar health interventions, social health insur- ance should not do better or worse than general taxation; hence we expect an effect that is not statistically different from 0. The second alternative model studies the marginal impact of a mixed health- financing scheme. A dummy variable DMRS, equal to 1 when the country has a mixed health-financing system and 0 otherwise, is included next to DARS. Mixed health-financing schemes also include health insurance schemes apply- ing risk sharing and therefore should have a beneficial impact on health-level attainment. Therefore, our hypothesis is that the marginal impact of DMRS on Ln (80 ­ DALE) is negative. In a third alternative model, we test whether certain groups of schemes within the overall set of mixed health-financing systems would have an addi- tional net effect on the level of health. We select the group of mixed systems that encompass health insurance schemes in which only employees are covered (DMRS1 = 1 and 0 otherwise) and health insurance schemes that cover other specific groups only (DMRS2 = 1 and 0 otherwise). As these health insurance schemes offer a lower degree of financial risk protection, as compared with schemes that cover all employees and the self-employed, the hypothesis is that the impact of DMRS1 and DMRS2 on DALE is negative; hence, the expected sign of the impact on the dependent variable is positive. These alternative models are combined further. Namely, we add both DSHI and DMRS to the explanatory variables of equation 12.1. Finally, we bring DSHI, DMRS, DMRS1, and DMRS2 together into the equation. The level of responsiveness is measured by an index (IR) that varies between 0 and 1, with 1 being the maximum. We adopted a logistic specification, implying that predicted values for IR stay within the 0­1 interval: (12.2) Ln [IR/(1 ­ IR)] = a21 + b21 HEC + c21 EDU + d21 DARS. The impact of HEC is presumed to be positive as more resources are likely to facilitate the responsiveness of the health system. EDU can be understood as capturing the positive effect of a literate and more developed society on respon- siveness. The hypothesis to be tested further is that advanced risk-sharing sys- tems are associated with a larger degree of stewardship. The latter in turn is likely to positively influence the mechanisms and incentives that entail a greater responsiveness. As in the case of the health level, alternative models can be estimated. DSHI is expected to be neutral vis-à-vis responsiveness; we therefore expect a coefficient that is not statistically different from 0. Medium risk-sharing schemes are expected to do better than low risk-sharing schemes in terms of responsiveness; 406 Health Financing for Poor People: Resource Mobilization and Risk Sharing hence the marginal impact of DMRS is expected to be positive. However, a nega- tive impact is expected to be associated with DMRS1 and DMRS2. The basic model further comprises the analysis of three measures for distribu- tional impact: an index of IECS, IFFC, and IRD. Equations are developed that examine the impact of the dummy variable (DARS) on these distributional vari- ables. We have adopted the same functional form as in equation 12.2: (12.3) Ln [Ij/(1 ­ Ij)] = a31 + b31 DARS, where Ij (j = 1,...,3) refers to the three above-mentioned indexes, respectively. The effect of DARS on the indicator of fair financing is expected to be positive when using the logit form of the equation. The hypothesis to be tested is that in countries with advanced risk-sharing, more so than in other systems, people pay financial contributions according to their capacity to pay. This would be associated with a higher IFFC. Second, systems with universal coverage generally pay more attention to the objective of equal treatment for equal need. It is therefore assumed that such systems also respond to people's expectations as to the nonmedical aspects of health care in a more equal way. Hence the effect of DARS on the distrib- ution of responsiveness is anticipated to be positive as well. Third, it is assumed that universal coverage systems are more likely to provide people with a similar benefit package than are other systems, irrespective of their socioeconomic background, with a resulting positive impact on the distributional aspects of child health. For alternative models, we first include DSHI as an additional dummy variable in the equation. The sign of the coefficients of DSHI is uncertain, however. Whether social health insurance is inferior or superior to general taxation in terms of fair financing depends on a host of factors. Those factors include the way health insurance contributions are levied (with an earnings ceiling or with- out), the progressivity of income taxes, the level of copayments, user fees, or both, and the types of health services that are excluded from coverage and their prices. In general, when adding DMRS to the explanatory variables, we expect it to have an additional positive effect. The effects of DMRS1 and DMRS2 are antic- ipated to be negative. Specification of Enlarged Models We also studied enlarged models, adding other potentially significant explana- tory variables to the specification of the equations. In one enlarged model, the GINI index measuring the distribution of income is included: (12.4) Ln [Ij/(1 ­ Ij)] = a41 + b41 GINI + c41 DARS, where Ij (j = 1,...,3) refers again to the three indexes, respectively. Income inequality in society, as measured by the GINI, is expected to be mir- rored, at least partially, in the distribution of the health-financing burden on the various households. For instance, it is expected, other things being held con- stant, that the larger the income inequality, the lower the degree of fair financ- Impact of Risk Sharing on the Attainment of Health System Goals 407 ing. The coefficient b41 is therefore expected to be negative. We further antici- pate that countries with advanced risk-sharing are apt to counteract the initial effect of overall income inequality by introducing better financial risk protection for all the population. Hence we expect that the impact of DARS is maintained. Further variants of equation 12.4 are investigated via the inclusion of DSHI, DMRS, DMRS1, and DMRS2. In principle, there should be no change in the sup- posed direction of the effects already commented upon earlier. In addition, the impact of interaction variables, combining the GINI index with the organiza- tional dummy variables, can be studied. The coefficients of the interaction vari- ables are expected to show that the larger the degree of risk sharing, the more the impact of the GINI index is offset. For instance, the coefficient of the inter- action term between GINI and DARS is anticipated to be positive. The various models considered so far measure the "average" impact of the dif- ferent risk-sharing schemes on the attainment indicators. Enlarged models with the inclusion of interaction variables between the ratio of public health expendi- ture to total health expenditure (PHE%) and the organizational dummy variables among the determinants can also be considered. The equations are the following: (12.5) Ln (80 ­ DALE) = a51 + b51 Ln HEC + c51 Ln EDU + d51 DARS + e51 DARS*PHE% Ln [IR/(1 ­ IR)] = a52 + b52 HEC + c52 EDU + d52 DARS + e52 DARS*PHE% Ln [Ij/(1 ­ Ij)] = a53 + b53 DARS + c53 DARS*PHE%, where Ij (j = 1,...,3) refers to the three equality indexes, respectively. We expect that a higher PHE% would reinforce the effect of the organiza- tional variables in the earlier models. This means that coefficient e51 is expected to be negative, whereas we expect e52 to be positive. The more health expendi- ture is managed through the public sector, and thus the higher the degree of risk pooling, the larger the equality of people within the health system is presumed to be. Coefficient c53 is therefore anticipated to have a positive sign. Note that in alternative equations, we also investigate the interaction of PHE% with DSHI, DMRS, DMRS1, and DMRS2. Estimation Results For a comprehensive account and discussion of all estimation results, we refer to Carrin and others 2001. In this chapter, we only present the "best" regressions6 of both the basic and enlarged models in tables 12.4 and 12.5 (see appendix). In addition, only the results using restricted samples or restricted samples with additional deletion of influential data are presented. The estimation results for the basic model are presented in table 12.4. First, concerning the level of health (DALE), the coefficients of DARS, HEC, and EDU are negative as expected and are statistically significant at the 1 percent sig- nificance level. 408 Health Financing for Poor People: Resource Mobilization and Risk Sharing Second, from the equation for the level of responsiveness (IR), we see that HEC and EDU do not have a statistically significant impact. One major reason is likely to be that the index of responsiveness contains both elements of respect for persons and client orientation and that these are influenced differently by HEC and EDU. For instance, HEC may be important in explaining client orienta- tion, whereas it may not be in explaining respect for persons. Therefore, when analyzing the determinants of the overall index of responsiveness, the effect of HEC may disappear. The coefficient of DSHI is also not statistically significant, as expected. Notice, however, that the coefficients of both DARS and DMRS have the expected sign and are statistically significant. Third, the explanatory power of the regression for the index of fair financ- ing (IFFC) is minimal; DARS does not have a statistically significant impact on the IFFC. We submit that the major reason for this unsatisfactory result is the rel- atively small sample size. Moreover, the sample did not include sufficient data on countries with advanced and with low risk-sharing. For instance, the (full sample) data on advanced risk-sharing are those of Bulgaria, Jamaica, Kyrgyz Republic, Romania, and Russia and inadequately reflect the experience of high-income countries with either social health insurance or general taxation financing. Fourth, in the equation for the distribution of responsiveness (IRD) the coefficient of DARS is statistically significant. The impact of DSHI is statistically insignificant. Fifth, the results for the index of equality of child survival (IECS) show that both DARS and DMRS have statistically significant impacts. We next present an overview of the estimation results for the enlarged models with the GINI index as an explanatory variable in the equations for the distribu- tional measures. The results are presented in table 12.5. In the fair-financing equation (IFFC), which has very low explanatory power, the coefficient of the GINI index has the anticipated sign but is not statistically significant. The coeffi- cient of DARS is also not statistically significant. Related to the distribution of responsiveness (IRD), the result shows sig- nificant impacts of both DARS and DMRS, as well as of the GINI index. All coef- ficients have the expected sign. One can conclude that these risk-sharing arrangements are efficient in counterbalancing the overall effect of income inequality. A threshold for the GINI indexes can be computed, indicating the value above which risk sharing is no longer able to counteract the effect of over- all income inequality. In the case of a country with an advanced risk-sharing scheme, the threshold value is 57.9. In the case of medium risk-sharing schemes, the threshold is 26.3. From these estimates, one can infer that advanced risk- sharing schemes are more effective in counteracting the effects of overall income inequality in society. In the regression result related to the inequality of child survival (IECS), the sign of the GINI coefficients is against our expectations. Surprisingly, the coefficient of GINI is also statistically significant at the 10 percent level. The coefficient of DARS has the anticipated sign, however, and is statistically signifi- cant at the 1 percent level. Impact of Risk Sharing on the Attainment of Health System Goals 409 Inclusion of the interaction variables with PHE% in the equations did not result in a general improvement of the estimation results. For instance, in a number of cases, the coefficients of DARS have the correct sign but are statisti- cally insignificant. In other instances, the coefficient of DARS has a negative sign. Further estimations were done with transformed interaction variables. In the case of the interaction between DARS and PHE%, the variable constructed was DARS*(PHE% ­ 0.5). The coefficient associated with this variable reveals the impact of the difference between PHE% and a threshold of 50 percent. The results for IR, IFFC, IRD, and IECS are not satisfactory: the coefficient of the new interaction variable has a wrong sign, is not statistically significant, or both. Only in the case of DALE did we obtain a satisfactory result: both the coefficients of DARS and the interaction variable have the expected sign and are statistically significant. This result is presented in summary table 12.5. In other words, for those advanced risk-sharing systems with a PHE% above 50 percent, the level of PHE% reinforces the "average" effect of DARS. Key Conclusions A first conclusion is that the degree of advanced risk-sharing, as measured by the dummy variable DARS, is significant in the equations for four of the five goal measurements. No impact could be found in the case of the index of fair financ- ing, but we submit this is due to the small sample size. In addition, in at least two of these measurements (level of responsiveness, distribution of health), the variable DMRS also has been shown to have a statistically significant impact. Second, when enlarging the set of explanatory variables in the models for the distributional measures with the GINI index, DARS remains statistically signifi- cant in the equations for IRD and IECS. In addition, DMRS has a statistically sig- nificant impact in the equations for IRD. An additional interpretation emerges from the results, namely that risk sharing corrects for, or may even outweigh, the negative effect of overall income inequality on the fair-financing index and the index of distribution of responsiveness. Third, using interaction terms with PHE% leads to plausible results for DALE only: the level of PHE% reinforces the average positive effect of advanced risk- sharing. An analysis with preliminary updated data was also undertaken; since publi- cation of the WHR 2000, WHO has developed updated estimates for the level (HEC) and share of public health expenditure in total health expenditure (PHE%). When using updated data for HEC in the equations for DALE and IR, similar results (in terms of explanatory power, sign and statistical significance of coefficients) are obtained as those presented here. The use of the updated PHE% does not significantly change the estimates for the equations with the interac- tion terms. Estimates of the index of fair financing were also obtained for an additional 30 countries. Reestimation of the equations using an enlarged sample of 50 now leads to two interesting results: (1) the advanced risk-sharing dummy 410 Health Financing for Poor People: Resource Mobilization and Risk Sharing variable DARS exerts a statistically significant effect on the fair-financing index; and (2) the GINI index has a statistically significant impact on IFFC but is coun- terbalanced by a health-financing system characterized by advanced risk- sharing. These preliminary results prove to be more in line with those obtained for the other distributional measures. COMMUNITY RISK-SHARING ARRANGEMENTS: FURTHER NEED TO MEASURE THEIR IMPACT Community risk-sharing arrangements are increasingly recognized as an inter- mediate response to the constraints many countries experience in rapidly extending financial risk protection to the national population. A body of research exists with respect to community-financing arrangements and their functioning within communities, districts, or regions. Information at the national level is clearly lacking. We have made an attempt to scan the literature and other sources (especially, Atim 1998; Bennett, Creese, and Monasch 1998; Carrin, De Graeve, and Devillé 1999; ILO and PAHO 1999; and Ginneken 1999) to see whether community risk-sharing organizations exist at country level. Only countries with low to medium risk-sharing will be considered, as countries with advanced risk-sharing in principle do not need to be complemented by community risk-sharing schemes. We recorded that in the set of countries with a public health expenditure ratio of 50 to 75 percent, 25 out of 44 countries have community risk-sharing schemes operating. In the countries with a ratio below 50 percent, 42 out of 58 are reported to have such schemes. This is not unexpected, as we would expect community risk-sharing schemes to be established where governments are not able to make sufficient advancement in risk protection. However, these data are insufficient for econometric analysis. Further work concerning the quantitative importance of community risk-sharing arrangements at the country level is needed. This could be measured by the number of risk-sharing schemes and the percentage of population covered by such schemes. Alternatively, one could measure the ratio of the expenditures incurred by such schemes to overall pri- vate health expenditure. The higher this ratio, the greater is the effort to share risks. Current work on National Health Accounts at WHO goes in this direction by attempting to collect data on expenditure by nongovernmental institutions and communities. Further work is needed on identifying the part of this expen- diture that is spent within the framework of risk-sharing arrangements. Impact of Risk Sharing on the Attainment of Health System Goals 411 CONCLUDING REMARKS The results presented here give empirical support for the hypothesis that the degree of risk sharing in health-financing organizations matters for health system attainment, as measured by the five indicators. The categorical variables indicat- ing whether a country has a health-financing organization with advanced or medium risk-sharing categories are especially seen to have significant impact. These effects prove to be quite robust after the GINI index is introduced among the explanatory variables in the models for the distributional measures. We noted that the plausibility of the results improves when using the restricted samples, deleting data for those countries whose classification was con- sidered uncertain. Further information will be necessary to address this uncer- tainty. In general, final data for larger samples of countries are welcome for four of the health system attainment indexes, especially for the index of fair-financing contribution, so that these will better reflect the patterns of risk sharing in the world. In the current samples, some of the risk-sharing schemes are underrepre- sented, which has entailed sensitivity of the results to specific data points. Further work could also be done on designing much more refined quantitative measures for the degree of risk sharing. Indeed, within each of the categories of health-financing organization that we considered, a further variety in the degree of financial protection of different population subgroups may well be present. In addition, more work needs to be undertaken to measure the quantitative importance of risk-sharing schemes for communities and the informal sector at the country level as well as the depth of risk sharing of these schemes. Only then can further econometric analysis be undertaken. In the meantime, given the empirical results obtained so far, one can clearly hypothesize beneficial impacts of these schemes on the health system attainment indicators. See page 415 for acknowledgments, notes, and references. APPENDIX 12A. STATISTICAL DATA TABLE 12.3 Classification of Countries by Degree of Risk Sharing in the Health-Financing System Advanced Risk-Sharing Medium Risk-Sharing Low Risk-Sharing All employees and Specific groups self-employed (with some All employees only covered Social health exclusions) covered covered by by health insurance (SHI) General taxation by health insurance health insurance insurance Australia Albania Colombia Algeria Botswana Afghanistan Austria Antigua and Ecuador Andorra Brazil Angola Belgium Barbuda El Salvador Argentina Burkina Faso Armenia Bulgaria Azerbaijan Equatorial Guinea Bolivia Burundi Bahamas, The Chile Bahrain Libya Cape Verde Cameroon Bangladesh Costa Rica Barbados Mongolia Congo, Rep. of China Benin Croatia Belarus Peru Egypt, Arab Rep. Côte d'Ivoire Bhutan Czech Republic Belize Tunisia Gabon Dominican Cambodia Estonia Bosnia and Uruguay Guinea Republic Central African France Herzegovina Honduras Guatemala Republic Germany Brunei Darussalam Lebanon Guinea-Bissau Chad Greece Canada Mali Haiti Comoros Hungary Cook Islands Mexico India Congo, Dem. Israel Cuba Namibia Indonesia Rep. of Japan Cyprus Panama Iran, Islamic Djibouti Korea, Rep. of Denmark Paraguay Rep. of Eritrea Latvia Dominica Philippines Iraq Ethiopia Lithuania Finland Senegal Jordan Fiji Luxembourg Iceland Turkey Kenya Gambia, The Macedonia, FYR Ireland Venezuela Lesotho Georgia Monaco Italy Madagascar Ghana Netherlands Jamaica Mauritania Grenada Norway Kazakhstan Morocco Guyana Poland Korea, DPR Mozambique Kiribati Romania Kuwait Myanmar Lao PDR San Marino Kyrgyz Republic Nicaragua Liberia Slovak Republic Malaysia Niger Malawi Slovenia Malta Pakistan Maldives Switzerland Mauritius South Africa Marshall Islands Yugoslavia, Moldova Thailand Micronesia Fed. Rep. New Zealand Trinidad and Nauru (Serbia/ Niue Tobago Nepal Montenegro) Oman United States Nigeria Palau Vietnam Papua New Guinea Portugal Yemen, Rep. of Rwanda Qatar São Tomé and Russian Federation Principe St. Kitts and Nevis Sierra Leone St. Lucia Solomon Islands St. Vincent Somalia Samoa Sri Lanka Saudi Arabia Sudan Seychelles Suriname Singapore Swaziland Spain Syrian Arab Rep. Sweden Tanzania Tajikistan Togo Turkmenistan Tonga Ukraine Tuvalu United Arab Uganda Emirates Vanuatu United Kingdom Zambia Uzbekistan Zimbabwe 412 Appendix--Statistical Data 413 TABLE 12.4 Estimation Resultsa for the Basic Model Explanatory variables DALEb IRc IFFCb IRDd IECSe Ln (80- DALE) Constant 4.9423 ­0.4896 2.2874 1.6327 0.2798 (0.3328) (0.2160) (0.2786) (0.4507) (0.2038) (14.8493) (­2.2663) (8.2099) (3.6228) (1.3329) HEC ­0.1919 0.0000 (0.0197) (0.0003) (­9.7498) (0.1150) EDU ­0.2141 0.0032 (0.0834) (0.0026) (­2.5684) (1.2540) DARS ­0.2963 0.7244 ­0.1146 4.2257 6.6269 (0.0654) (0.2244) (0.6072) (0.8228) (0.3868) (­4.5321) (3.2275) (­0.1888) (5.1355) (17.1343) DSHI ­0.2521 ­1.4049 (0.1987) ( 0.9107) (­1.2688) (­1.5427) DMRS 0.2673 0.7217 1.0737 (0.1148) (0.5355) (0.4202) (2.3294) (1.3478) (2.5550) DMRS1 ­0.1079 (0.4607) (­0.2343) DMRS2 ­0.6458 (0.3995) (­1.6165) R-squared 0.7874 0.5678 0.0021 0.5749 0.8778 Adjusted R-squared 0.7821 0.4597 ­0.0566 0.5276 0.8671 S.E. of regression 0.2639 0.2134 1.0791 1.1924 0.7350 Ak. info criterion 0.2049 ­0.0525 3.0894 3.3097 2.3149 Sample size 124 26 19 31 51 a. The first and second coefficients in the parentheses refer to the standard error and t-statistic, respectively. b. Restricted samples. c. Bulgaria excluded from the sample. d. Chile and Poland excluded from the full sample. e. Uzbekistan excluded from the restricted sample. 414 Appendix--Statistical Data TABLE 12.5 Estimation Resultsa for the Enlarged Models (Summary) Explanatory variables IFFCb IRD IECSb DALEb Ln(80- DALE) Constant 2.8260 3.0610 ­0.7471 4.9446 (1.3698) (0.7956) (0.9164) (0.3306) (2.0630) (3.8539) (­0.8153) (14.9580) GINI ­0.0119 ­0.0375 0.0355 (0.0296) (0.0180) (0.0206) (­0.4020) (­2.0853) (1.7240) DARS ­0.2568 2.1713 5.3537 ­0.2088 (0.7162) (0.5222) (0.5531) (0.0843) (­0.3586) (4.1577) (9.6789) (­2.4774) DARS*[PHE%- 0.5] ­0.4556 (0.2798) (­1.6284) DMRS 0.9873 (0.4637) (2.1291) HEC ­0.1897 (0.0196) (­9.6837) EDU ­0.2166 (0.0828) (­2.6155) R-squared 0.0121 0.5191 0.7053 0.7920 Adjusted R-squared ­0.1114 0.4590 0.6906 0.7850 S.E. of regression 1.1067 0.9320 1.1912 0.2621 Ak. info criterion 3.1846 2.8286 3.2550 0.1990 Sample size 19 28 43 124 a. The first and second coefficients in the parentheses refer to the standard error and t-statistic, respectively. b. Restricted samples. Impact of Risk Sharing on the Attainment of Health System Goals 415 Acknowledgments: The authors of this chapter are grateful to the World Health Organiza- tion for having provided an opportunity to contribute to the work of the Commission on Macroeconomics and Health and to the World Bank for having published the material in this chapter as an HNP Discussion Paper. Discussions with and suggestions from Alex Preker, Melitta Jakab, David Evans, Kei Kawabata, and Ajay Tandon are gratefully acknowledged. Thanks also to Chris James for a thorough review of an earlier draft of this chapter. During the writing, Philip Musgrove was on secondment to the World Health Organization. The views expressed in this document are solely the responsibility of the authors. NOTES 1. This summary-measure of population health adjusts life expectancy at birth for the burden of disability. Disability weights are used to convert years lived in disability into equivalent years lived in good health. (See further Mathers and others 2000.) 2. For further details, we refer to http://www3.who.int/whosis/discussion_papers/ discussion_papers.cfm?path=whosis,evidence,discussion_papers&language=english. 3. For instance, the agricultural self-employed population may not be covered. Workers in small enterprises with fewer than 10 workers may not be insured either. 4. Note that social insurance expenditure is included in public health expenditure. 5. For these nine countries, the data are either incomplete with low reliability (two coun- tries out of nine) or are incomplete with high to medium reliability. 6. "Best" according to the adjusted R-square, the Akaike criterion, or both, as well as the theoretical consistency of the model. REFERENCES Atim, C. 1998. Contribution of Mutual Health Organizations to Financing, Delivery, and Access to Health Care: Synthesis of Research in Nine West and Central African Countries. Technical Report No. 18. Partnerships for Health Reform Project. Abt Associates Inc., Bethesda, Md. Bennett S., A. Creese, and R. Monasch. 1998. Health Insurance Schemes for People Outside Formal Sector Employment. ARA Paper 16. World Health Organization (WHO), Geneva. Carrin, G., D. De Graeve, and L. Devillé, eds. 1999. "The Economics of Health Insurance in Low- and Middle-Income Countries." Social Science and Medicine 48(special issue): 859­64. Carrin G., R. Zeramdini, P. Musgrove, J-P. Pouillier, N. Valentine, and K. Xu. 2001. The Impact of the Degree of Risk-Sharing in Health Financing on Health System Attainment. World Bank, HNP Discussion Paper. Washington, D.C. ILO (International Labour Office) and PAHO (Pan American Health Organization). 1999. Synthesis of Case Studies of Micro-Insurance and Other Forms of Extending Social Protection in Health in Latin America and the Caribbean. Paper presented at the meeting on exten- sion of social protection in health to excluded groups in Latin America and the Caribbean. Mexico, November 29­December 1; http://oitopsmexico99.org.pe. 416 Health Financing for Poor People: Resource Mobilization and Risk Sharing Mathers, C., R. Sadana, J. Salomon, C. J. L. Murray, and A. D. Lopez. 2000. Estimates of DALE for 191 Countries: Methods and Results. GPE Discussion Paper 16. WHO, Geneva. Murray, C. J. L., and J. Frenk. 2000. "A Framework for Assessing the Performance of Health Systems." Bulletin of the World Health Organization 78 (6): 717­31. Nolan, B., and V. Turbat. 1995. Cost Recovery in Public Health Services in Sub-Saharan Africa. Washington, D.C.: Economic Development Institute of the World Bank. SSA (Social Security Administration). 1999. Social Security Programs throughout the World, 1999. Washington, D.C.: U.S. Government Printing Office. van Ginneken, W., ed. 1999. Social Security for the Excluded Majority: Case Studies of Devel- oping Countries. Geneva: International Labour Office. WHO (World Health Organization). 2000. The World Health Report 2000. Health Systems-- Improving Performance. Geneva. About the Coeditors and Contributors THE COEDITORS Alexander S. Preker, lead economist and editor of the Health, Nutrition, and Population (HNP) Publication Series, is responsible for overseeing the World Bank's analytical work on public policy in the health sector, market dynamics, health financing, service delivery, pharmaceuticals, and health labor markets, focusing on ways to help developing countries accelerate progress toward achieving the U.N. Millennium Development Goals by 2015. He is a member of a team of researchers currently undertaking a major review of disease con- trol priorities in developing countries, with support from the Gates Founda- tion and the Fogarthy International Center of the U.S. National Institutes of Health. Recently, he has worked closely on a number of projects with the World Health Organization (WHO), the International Labour Organization (ILO), the Rockefeller Foundation, the International Federation of Pharmaceu- tical Manufacturers Associations, the International Hospital Federation, the International Federation of Health Plans, and several leading academic centers involved in health and financial protection in developing countries. Preker has published extensively and is a frequent speaker at major interna- tional events. He is on the editorial committee of the Bank's Publications Department, is the chief editor of its HNP publications, and is a member of the editorial committees of several international journals. Preker also is on the External Advisory Board of the London School of Economics Health Group and is a member of the teaching faculty for the Harvard/World Bank Institute Flagship Course on Health Sector Reform and Sustainable Financing. He has an appointment as adjunct associate professor at the George Washington Uni- versity School of Public Health. His training includes a Ph.D. in economics from the London School of Eco- nomics and Political Science, a fellowship in medicine from University College London, a diploma in medical law and ethics from King's College London, and an M.D. from the University of British Columbia/McGill. Guy Carrin is senior health economist in the Department of Health Financing and Stewardship at the WHO in Geneva, Switzerland. His current work deals with issues of community and social health insurance. He has published exten- sively in the areas of social security, macroeconomic modeling and simulation, and health economics. In the field of health economics, he is the author of the books Economic Eval- uation of Health Care in Developing Countries and Strategies for Health Care Finance in Developing Countries; and coeditor of Macroeconomic Environment and Health 418 Health Financing for Poor People: Resource Mobilization and Risk Sharing and a special issue of Social Science and Medicine on "The Economics of Health Insurance in Low- and Middle-Income Countries." Carrin was a member of Working Group 3 of the WHO Commission on Macroeconomics and Health and is currently a member of the editorial commit- tee of the Bulletin of the World Health Organization. He is associated with the Uni- versity of Antwerp (Belgium) and with Boston University. He holds an M.A. and a Ph.D. in economics from the University of New Hamp- shire (United States) and the University of Leuven (Belgium), respectively. He held a Canada Council Fellowship at the University of Toronto and was visiting research fellow at the University of Michigan and Brandeis University. He was econometrician at the Ministry of Planning in Tunis (Tunisia), researcher at the Center for Operations Research and Econometrics in Leuven, fellow in interna- tional health at the Harvard School of Public Health, and economic adviser to the Minister of Public Health of the federal government of Belgium in 1987­88. OTHER CONTRIBUTING AUTHORS Dyna Arhin-Tenkorang is a public health physician and health economist and holds a clinical lectureship position in health economics at the London School of Hygiene and Tropical Medicine (LSHTM). The London School is Britain's national school of public health and a leading postgraduate institution in Europe for public health and tropical medicine. In 2000­02, Arhin-Tenkorang held the position of senior economist of the WHO Commission on Macroeconomics and Health (CMH), which was established to assess the place of health in global economic development. At LSHTM, Arhin-Tenkorang has managed several health economics research projects focusing on the equity of access to health care and the role of fees and risk-sharing mechanisms and has published extensively on the findings of her empirical work. She was also a contributor to the World Health Report 2000: Health Systems--Improving Performance and the editorial manager of the report of the WHO Commission on Macroeconomics and Health, released in December 2001. Her recent publications include: Health Insurance for the Informal Sector in Africa: Design Features, Risk Protection, and Resource Mobilization, a World Bank HNP Discussion Paper; "Mobilizing Resources for Health: The Case for User Fees Revisited," CMH Working Paper Series WG3; and "Beyond Communicable Dis- ease Control: Health in the Age of Globalization," a chapter in a United Nations Development Programme (UNDP) publication on global public goods (2002). Arhin-Tenkorang's qualifications in health economics and health planning include a Ph.D. from the London School of Economics and Political Science, an M.Sc. from the University of York, and a postgraduate diploma from the Nuffield Center, University of Leeds. She is a member of the West African College of Physicians and has a medical degree from the University of Ghana Medical School. About the Coeditors and Contributors 419 François Diop is a senior health economist who has extensive experience in health financing, demography, and survey research. Currently, he serves as one of the Partnerships for Health Reform (PHR) advisers in Senegal and PRIME II in Rwanda. Diop has extensive experience in health sector reform and health care financing, with a focus on Africa. He specializes in conducting applied research on alternative health-financing mechanisms for primary health care, including community-based financing schemes, demand surveys, facilities surveys (such as cost studies), analyses of means testing policies, and studies on quality of health services. Under the Health Finance and Sustainability Project, he served as the resident adviser to the Ministry of Health in Niger. He played a major role in the design, implementation, and analysis of large-scale pilot tests in three districts in which cost-recovery mechanisms and quality improvements were introduced in public primary health facilities, which led to the development of a national policy of cost recovery for primary health care. Diop placed special emphasis on building the institutional capabilities of the Ministry of Health in formulating and assess- ing policy and in conducting applied research. Under the PHR project, he played a key role in the design and evaluation of the Rwanda prepayment schemes and community-based health insurance experiment; he is currently providing tech- nical support to Rwanda for scaling up community-based health insurance nationwide. Diop designed and initiated the experimentation of a multilevel integrated health-planning and financing scheme to support the coordination of health interventions within the devolved health sector in Senegal. While employed at the World Bank, Diop worked on the implementation of country status reports on poverty and health in African countries to support the elabora- tion of the health component of the poverty reduction strategy papers. He has also worked with the Ministry of Finance of Senegal and has conducted research and provided technical assistance in demography and family planning in Côte d'Ivoire, Liberia, Nigeria, and Senegal. He has worked as a consultant to many other organizations, including the Population Council, Centers for Disease Con- trol, the World Bank, and the WHO. Diop received his B.A. in geography from the University of Dakar in Senegal, an M.S. in demography from the Institute for Demographic Training and Research in Cameroon, and a Ph.D. in demography and economics from the Johns Hopkins University. He is fluent in English and French. David Dror is associate director of research at the Laboratoire d'Analyse de Sys- tèmes de Santé (UMR 5823, CNRS), University of Lyon 1, France (ongoing, since 2002). From January 1998 to April 2003, he served as senior health insurance specialist within the International Labour Organization's Social Protection Sec- tor, with special interest in development and stabilization of microinsurance schemes. During this period, he was the project leader for an ILO­World Bank project on social reinsurance for health care. This project developed the concep- tual framework for, and piloted the reinsurance of, community-funded health 420 Health Financing for Poor People: Resource Mobilization and Risk Sharing schemes in developing countries. From 1989 to 1997, he served as chief of staff, Health Insurance Fund of the ILO-ITU (International Telecommunication Union), an in-house health insurance plan providing coverage in more than 100 countries and operating a full multicurrency financial system. He was an official of the ILO from 1982 to 2003. Prior to working at the international level, he held several posts at the national level, notably with the Israeli Employers' Federation (1977­81), where he was responsible for a national occupational pension agree- ment for the private sector (1979); negotiated innovative solutions for COLA wage indexation, which were applied nationwide during Israel's years of hyper- inflation (1975­81); served as lead representative on the Council of the National Insurance Institute of Israel; and was appointed delegate to the International Labour Conference (1977­81) and a member of the ILO Committee of Experts on Social Security Questions (1980­82). Dror has a Ph.D., with highest distinction (University of Lyon 1, France), a D.B.A., magna cum laude (St. George University), an M.A. (Hebrew University of Jerusalem), and degrees in liberal arts, business administration, health services, and economics (as well as a teaching diploma). He earned a Fulbright Scholar- ship (1965). Anil Gumber is a senior research fellow at the Centre for Health Services Stud- ies, Warwick Business School (United Kingdom). Gumber was previously senior economist at the National Council of Applied Economic Research, New Delhi. He was awarded a Ph.D. in economics from Gujarat University, Ahmedabad. He has also held the Takemi Fellowship at the Harvard School of Public Health, Boston. Gumber, trained in international health, has specifically addressed health issues of the poor, disadvantaged, and vulnerable. He has undertaken several con- sultation projects funded by the World Bank, the UNDP, UNICEF, the United Nations Population Fund, the WHO, the U.K. international development agency (DFID), and the Ford Foundation. He is a specialist in econometrics, manage- ment, and analysis of large data sets as well as in designing survey instruments and sampling frames for primary data collection. Gumber is currently undertak- ing research activities sponsored by the Economic and Social Research Council in the areas of ethnicity, health, and population diversity in the United Kingdom. His research interests are ethnicity and health; equity and international health; health care financing and insurance in developing countries; cost and efficiency issues in health; child health and nutrition; injury prevention and control; migration and resettlement; and poverty, welfare, and social policy issues. Gumber's most recent papers include: "Economic Reforms and Health Equity"; "External Assistance to the Health Sector"; and "Health Insurance for the Poor." He is the author of Displaced by Development: Oustees of an Irrigation Project; coauthor of Inter-Village Differences in Social Overheads and Their Correlates in Matar Taluka; coauthor of Health Insurance for Workers in the Informal Sector: Detailed Results from a Pilot Study; and coauthor of Who Benefits from Public Health Spending in India: Results of a Benefit Incidence Analysis for India. About the Coeditors and Contributors 421 William C. Hsiao is the K. T. Li Professor of Economics and director of the pro- gram in health care financing at the Harvard School of Public Health. He received his Ph.D. in economics from Harvard University and is a qualified actu- ary with extensive experience in insurance. His current research focuses on developing a theory of health system economics that could provide an analytical framework in diagnosing the causes of the successes or failures of a system and on advising governments on their health sector reforms, including China, Colombia, Cyprus, Hong Kong, Poland, South Africa, Sweden, Taiwan, Uganda, and Vietnam. His research has also concentrated on payment for hospital and physician services, social and private insurance, and competition in managed care markets. Hsiao developed a rational fee schedule for physician services, the resource-based relative value scale (RBRVS) adopted by Australia, Canada, France, and the United States. He was named the Man of the Year in Medicine for his work on the RBRVS. Hsiao has published more than a hundred papers and several books. He was elected to the Institute of Medicine, National Academy of Sciences, and National Academy of Social Insurance. He has served on the boards of directors of a num- ber of professional organizations, including the National Academy of Social Insurance and the Society of Actuaries. He has also served as an adviser to three U.S. presidents and to the U.S. Congress, the World Bank, the International Monetary Fund, and the WHO. He is a recipient of honorary professorships from several leading Chinese universities. Melitta Jakab, a Hungarian national, is completing her Ph.D. in health eco- nomics at Harvard University. Previously, she worked at the World Bank's Health, Nutrition, and Population Network. Her research interests include assessing alter- native health-financing options for the poor, the operation of hospital markets, and health system reform in transition economies. Jakab is a member of the teaching faculty of the Harvard/World Bank Institute Flagship Course on Health Sector Reform and Sustainable Financing. She is codirector of international train- ing programs at the Health Services Management Training Centre in Budapest. Johannes Paul Jütting is affiliated with the OECD Development Centre in Paris, where he works as a senior economist. He is in charge of the research pro- ject on Social Institutions and Development, a main activity of the center's work program in 2003­04. He was previously research fellow at the Center for Devel- opment Research (ZEF) in Bonn, where he directed the Research Group on Poverty. In that position, he coordinated ZEF's research on Innovative Health Insurance Schemes for the Poor. He is also a scientific adviser of a joint program of the World Bank, the WHO, and the ILO--Resource Mobilization and Risk Pooling in Low-Income Environments. Jütting holds a Ph.D. in agricultural economics awarded by Humboldt Uni- versity of Berlin and received his postgraduate habilitation degree from the Uni- versity of Bonn. 422 Health Financing for Poor People: Resource Mobilization and Risk Sharing Chitra Krishnan has a master's degree in public health from the Johns Hop- kins School of Public Health. At the time of this research, she was working as a junior researcher at the World Bank's Health, Nutrition, and Population Network. John C. (Jack) Langenbrunner is a senior health economist with both research and operations experience. He is currently working in the World Bank's Europe and Central Asia region. He has worked on health-financing issues in Azerbaijan, Croatia, Kazakhstan, Kyrgyz Republic, Poland, Russia, and Uzbek- istan as well as in several countries in the Middle East and North Africa region. In 2001, Langenbrunner led a Bank study on introducing health insurance in the Kingdom of Bahrain. Before joining the Bank, he worked in several countries in the South Asia and East Asia and Pacific regions. He is currently based in Moscow. Langenbrunner is leading the World Bank's work in Resource Alloca- tion and Purchasing (RAP) and has authored or coauthored a number of papers related to this initiative. He is also leading the Bank's work on a manual for National Health Accounts for low- and middle-income countries, the "NHA Pro- ducers Guide," which is now in draft and is available upon request. Before working at the Bank, Langenbrunner was with the U.S. Health Care Financing Administration, a health insurance program for more than 80 million Americans. There he worked on case-mix payment systems for hospitals and physician relative value scales. Later, in the U.S. Office of Management and Bud- get, he served on the Clinton Health Care Reform Task Force. Langenbrunner holds master's and doctoral degrees in economics and public health from the University of Michigan (United States). He is an occasional lec- turer on health-financing topics at both the Johns Hopkins University and the George Washington University. Philip Musgrove is chief economist of the Disease Control Priorities Project at the Fogarty International Center of NIH. He is also adjunct professor in the Department of Health Policy, School of Public Health and Health Services, George Washington University, having previously taught at American University and the University of Florida. Until 2002, he was a principal economist at the World Bank. In 1982­90, he was an adviser in health economics, Pan American Health Organization. He served from 1996 to 1998 at the Bank's Resident Mis- sion in Brasilia, Brazil, and in 1999­2001 was seconded by the Bank to the World Health Organization in Geneva. Musgrove has worked on health reform projects in Argentina, Brazil, Chile, and Colombia, in addition to dealing with a variety of issues in health econom- ics, financing, equity, and nutrition. His publications include more than fifty articles in economics and health journals; he has also authored, edited, or con- tributed chapters to more than twenty books. Jean-Pierre Poullier is a former research associate at the Brookings Institution, Washington, D.C., for Why Growth Rates Differ (E. F. Denison with J-P. Poullier, About the Coeditors and Contributors 423 1967). He was a consultant for the European Commission (on the competitive capacity of the European economy), the European Commission Statistical Office (on purchasing power parities), the Peruvian and the Algerian governments (on productivity and on industrial development), and the International Institute of the Management of Technology. He worked for the Organisation for Economic Co-operation and Development between 1972 and 1998, where he created and directed the Health Policy Unit; between 1999 and 2003, he headed the National Health Accounts Unit of the World Health Organization. He is the author of more than a hundred publications and articles on public finance and health systems. He has been a consultant to the World Bank on the construction of National Health Accounts in Iran, Saudi Arabia, and other countries in the Middle East. M. Kent Ranson is a lecturer in the Health Policy Unit at the London School of Hygiene and Tropical Medicine, and a research coordinator for the Self-Employed Women's Association Insurance (Vimo SEWA). Ranson has degrees in clinical medicine (M.D., McMaster University), public health epidemiology and biostatis- tics (M.P.D., Harvard University), and health economics and financing (Ph.D., University of London). He currently heads a three-year research project that will assess the impact of several interventions to optimize the equity impact of Vimo SEWA. His other research interests include the cost-effectiveness of public health interventions (antitobacco, antimalaria, and antitrachoma-blindness) and the identification and measurement of characteristics of health systems that prevent implementation of the (theoretically) most cost-effective interventions. Pia Schneider worked for Abt Associates Inc. in Kigali, Rwanda, for two years, until September 2000. She was the head of the Rwanda office of the Partnerships for Health Reform, financed by the U.S. Agency for International Development and provided technical assistance to the Rwandan Ministry of Health. The main activi- ties included the development, implementation, and evaluation of a community- based prepayment-insurance scheme for primary care in three health districts; and the implementation and analysis of the first Rwandan National Health Accounts with a disease-specific HIV/AIDS component. In addition, she managed and imple- mented a series of country activities to improve cost recovery and assess impacts on equity. She trained local staff in health care financing and health sector reform and responded to health sector reform needs as they arose during the transition period at the Ministry of Health. As a faculty member of the Abt Training Institute, she has designed and delivered training in health economics and in policy analysis and for- mulation to local counterparts from various countries. Schneider has an extensive background in health economics, international economics, and nursing. Prior to joining Abt Associates Inc., she served as an applied health research analyst­health economist for HMO Oregon at Blue- Cross/BlueShield of Oregon. In this position, Schneider designed clinical quality improvement studies and analyzed access, quality, effectiveness, costs, outcome, health behavior, satisfaction, and treatment measures. She was also involved 424 Health Financing for Poor People: Resource Mobilization and Risk Sharing with financial analysis of HMO provider groups. Schneider has worked as a dele- gate for the International Committee of the Red Cross, where she collaborated with public health projects and local governmental and nongovernmental orga- nizations in Dem. Rep. of Congo (former eastern Zaire) and South Africa. Schnei- der holds an R.N. degree from Lindenhof Hospital in Bern, Switzerland. She received her B.A. and M.A. in economics at the University of Basel in Switzer- land. She is now working as a part-time employee for Abt on issues related to health financing and economics as she completes her Ph.D. in health economics at the London School of Hygiene and Tropical Medicine. Siripen Supakankunti, Ph.D., is associate professor in economics and cur- rently director of the WHO Collaborating Centre for Health Economics at the Faculty of Economics, Chulalongkorn University, Bangkok, Thailand. Since 1979, the Centre for Health Economics (CHE) has developed expertise in the field of health economics through a research, training, and teaching program. In late 1993, the center was designated a WHO Collaborating Centre for Health Economics, with regional and global responsibilities. The center's objectives are to develop expertise and commitment to the application of health economics in formulating health care policies, planning resource allocation, and refining health care delivery processes in the Southeast Asia region and Thailand; to strengthen health economics research capacity in Thailand and Southeast Asia, particularly with respect to economic analysis and evaluation in the control of tropical diseases; to encourage research in economic analysis and evaluation of disease control, with special attention to tropical diseases; and to provide advi- sory and information services, particularly in Southeast Asia and other regions, on health economics research. As a WHO Collaborating Centre for Health Eco- nomics, CHE has made some progress in accordance with the proposed strategy for strengthening health economics capability in the region and going well beyond that geographic area to form wider international linkages such as collab- oration with the World Bank Institute at the World Bank, for organizing the Asia Flagship Training Program in Health Sector Reform and Sustainable Financing. Supakankunti is responsible for overall activities of the center and for coordi- nating and organizing the Asia Flagship Training Program in Health Sector Reform and Sustainable Financing in Bangkok. She has published on various topics related to health economics. Her area of specialized work and interests are specific components of health sector reform such as health care financing, health insurance, health policy and planning; impacts of structural change on the health sector; economic model- ing; the pharmaceutical industry; and international trade in health services, eco- nomic analysis, and evaluation (malaria, HIV/AIDS, water pollution). Emi Suzuki, a Japanese national, is a consultant to the World Bank's Health, Nutrition, and Population unit. With her strong statistical and demographic skills, she is working on compiling, updating, and analyzing Millennium Devel- About the Coeditors and Contributors 425 opment Goals health indicators and time-series data on health expenditure and health service and utilization. She is also working with UNICEF and the WHO for data harmonization. Suzuki received her Ph.D. in public health from the Johns Hopkins Bloomberg School of Public Health. Prior to her work at the Bank, Suzuki worked in Japan for several NGOs that are implementing health, education, and environmental projects in developing countries. Nicole Valentine is a health economist in the Global Program on Evidence for Health Policy at the WHO in Geneva, Switzerland. For the past three and a half years, she has worked at the WHO on interpersonal aspects of the quality of health systems (being treated with dignity, having personal information kept confidential, having clear communication, being attended to promptly). Prior to this appointment, she worked as a lecturer at the University of Cape Town and as a researcher, whose topics included patient satisfaction, national health accounts, and private, national, and local government health expenditure in South Africa. In the field of interpersonal aspects of quality of care, Valentine has con- tributed several chapters to the book Health Systems Performance Assessment: Debates, Methods, and Empiricism (2003). She holds an M.A. degree in economics from the University of Cape Town (South Africa) and an M.P.H. degree in health services from the University of Washington (Seattle, United States), where she studied as a Rotary Ambassador- ial Scholar. She is currently a Ph.D. candidate at the Amsterdam Medical Centre, University of Amsterdam (Netherlands). Ke Xu joined the WHO as a Global Health Leadership Fellow in 1999. Currently, she works as a health economist in the Global Program on Evidence for Health Policy at WHO headquarters in Geneva, Switzerland. Her focus is health pay- ments distribution and the relation of catastrophic health costs to impoverish- ment. For the World Health Report 2000: Health Systems--Improving Performance, she worked on development of the concept and methodology of the index of fairness in financial contributions. In the field of inequality of health financing, she contributed several chapters to the book Health Systems Performance Assess- ment: Debates, Methods, and Empiricism. She is also associate editor of the elec- tronic journal Evidence for Health Policy. Xu holds an M.D. degree and a Ph.D. in health economics from Fudan Uni- versity (formerly Shanghai Medical University). She has been a faculty member in the Department of Health Economics, Fudan University, teaching health eco- nomics, health insurance, survey design, and other economics courses. Her main research areas include: health financing and service originating; health financ- ing for the elder population in Shanghai; improvement in access to health ser- vices for the urban poor; and sustainable development of a rural cooperative medical scheme in China. She was a visiting scholar at the RAND Corporation and the University of Southern California before joining the WHO. 426 Health Financing for Poor People: Resource Mobilization and Risk Sharing Riadh Zeramdini worked as a health economist at the WHO headquarters in Geneva and as a trainee in the Consulting Cabinet (Sousse). He holds a Ph.D. in health economics and an M.S.E from the University of Lausanne (Switzerland), where he also worked as a research assistant in 1999­2000. He was awarded a license in management from the Faculty of Economic Sciences and Law of Sousse (Tunisia). Index Note: f indicates figures, n indicates notes (nn more than one note), and t indicates tables. ability-to-pay, 34, 35 in SEWA study, 304 Abota Village Insurance Scheme, Guinea- Thai health card program, 344, 351 Bissau, 35, 102­4t2.11, 160, 350 Africa, 121 description, 172­73t4.3, 176­78, health insurance schemes, 162­64, 184t4.4 184t4 Abt Associates, 252 literature review case studies, 54, access to health care, 265­66, 267t7.10, 56­57t2.2 273nn6­7, 343 prepayment schemes, 72­73, 77 Abota scheme, 178 regional reviews of community-based Bwamanda scheme, 181 financing in, 18­19, 34­35 CHF scheme, 176 urban and rural Sub-Saharan Africa, and copayments, 163­64 233­34 and health card program, 336t10.11, See also country listings; Sub-Saharan 337t10.12, 338t10.13, 339t10.14 Africa India, 293­94 After Payment program, Bangladesh, 152 and mutual health insurance schemes, agricultural production, and CMS 232, 244­46 funding, 138 Nkoranza scheme, 179 allocation of resources, 347 relationship to income, 163­64 ambulatory care, 304­05, 308­09t9.7, 311 Rwanda, 256­57 Amul Union, 296 Senegal, 205 Andorra, 401 Thailand, 344 Apairatr, S., 324, 325 activity status Arhin, D.C., 77, 80, 82 as determinant of expenditures Arhin-Tenkorang, Dyna, 3­51, on treatment by type of care, 157­97, 418 312t9.9 Asia as determinant of use of private literature review, 54, 56­57t2.2 facilities, 310, 310­11t9.8 regional reviews of CF in, 18­19, 29, and enrollment in SEWA scheme, 31­34 306t9.6 success and failures of specific CF Adeyi, O., 323 schemes, 128, 130­31 admissions, Nkoranza scheme, 179 asset index, 285 adult female mortality, 378t11.1 Atim, C., 63, 73, 88, 122, 163 adult male mortality, 378t11.1 Atiman Insurance Scheme, Tanzania, adverse selection 104­07t2.11 Bwamanda scheme, 181 CAM households, 174, 175 Bamako Initiative, 64, 69, 71­72, 78, and household illness ratio, 241 100t2.11, 350 as limitation of methodology, 215 Bangladesh, 76, 80 and primary care coverage, 132 See also Dhaka Community Hospital and relationship between elderly and Health Insurance Program, frequency of illness, 288 Bangladesh; Grameen Health and risk pooling, 87, 166­67 Program, Bangladesh Rwanda, 223, 254 beneficiary group, 62 428 Index benefits Card programs. See also Family Health in Abota scheme, 178 Card; health card schemes; School affordability of, 143­45 Children's Card; Sports Card; and CF in China villages, 139, 139t3.5, Worker Health Card 140t3.6, 144t3.9 Carte d'Assurance Maladie, Burundi, 35, claims for, 163 73, 77, 82, 87, 160, 323­24 health card program, 322 characteristics of CF schemes, knowledge of, 324 112­14t2.11 MHO scheme, 87 description, 171­75, 184t4.4 and predicted premiums, participation, case-based reimbursement, 352 and revenues, 187, 187f4.3 cash-earning periods, 209 relationship to premiums, 174­75 and revenue collections, 85­86, 86b2.4 Rwanda's PPS, 207, 228n3, 253, 273n2 See also payment schedules structure of, 164, 165 casualty insurance, 11 Thailand, 208 catastrophic health care, 68, 78­82 and type of health care covered by Catholic Church, health posts in insurance schemes, 295t9.1 Senegal, 205 voluntary health insurance cattle ownership, 218, 264, 269 plans, 294 CBHI. See community-based health Benin, 71­72, 78, 88, 100t2.11 insurance (CBHI) Bennett, S., 26b1.1, 63, 70­71b2.2, 73, 86, Center for Development Research, Bonn, 122, 275, 288 210t5.3, 211, 235 Berman, P., 278 Center for Integrated Development, bias, 215­16, 287­88, 294­95 DRC, 180 birth attendants, 176 Central Africa, 69, 85 Boal, M., 350 Central America. See Honduras Boboye District Scheme, Niger, 64 CF. See community-based financing Bolivia, in-kind contributions to health schemes (CF) care, 86, 86b2.4 Chabot, J., 177, 350 budgets childbirth, 219 reallocation of toward priority child health, 138, 365f11.2, 367­68, 367f11.3 programs, 386­87 as MDG, 363f11.1c, 364­65 Thailand, 317, 318t10.4 child mortality, 363f11.1c, 364, 366, Burundi. See Carte d'Assurance Maladie, 367f11.3 Burundi China, 121, 125, 126, 367 Bwamanda Health Zone, 180 child health care study, 138 Bwamanda Hospital Insurance Scheme, health expenditures, 376, 377f11.7, Democratic Republic of Congo, 35, 378t11.1 63, 64, 107­10t2.11, 276 impact of CBHI schemes on crop-selling season, 85 hospitalizations and out-of-pocket description, 172­73t4.3, 180­81, expenditures, 289t8.7 184t4.4 medical expenditures, 129 hospital utilization, 81 Rand experiment, 130 institutional characteristics, 91 total expenditure on health, 380 participation in, 161 See also Cooperative Medical System, bypass rates, 324 China China Network­Harvard Study, 138, 139 Canada, 319 Chogoria Hospital Insurance Scheme, capitation provider payment, 254 Kenya, 87, 109­10t2.11 Car Accident Insurance, 320 Chowdhury, A. Q., 76, 80, 85, 88 Caribbean, 54, 57t2.2 church-based providers, Senegal, 205 Carrin, Guy, 3­51, 80­81, 397­416, 417 Civil Service Medical Benefit Scheme, 319 Index 429 CLAISS study, 85 resorting to private providers, 6 CMS. See Cooperative Medical System, and resource mobilization, 123­28 China Rwanda, 206­07, 228nn1­4 coefficients of variation, 403 Senegal, 207­08 coffee-harvesting period, 85 strengths of, 27, 28­29b1.2 coinsurance, effect on demand for care, 143 success and failures of specific schemes, collective action, role in community 128, 130­31 financing, 9 summary of coverage, 75, 75t2.7, communes, 174, 194n3 76­77 community, 122, 153n7 terminology concepts, 59­67, 121 definition, 9, 60, 135, 153n7, 154n13 Thailand, 208­09 gains valued by members of, 125, types of, 25t1.5 125t3.1, 154n9 typology of characteristics, 128, involvement in Dana Sehat programs, 129t3.2 146­47 value added by types of, 31, 32t1.9, 33, powers delegated to, 233 133, 134t3.3, 135, 136t3.4 role in health care resources, 60 weakness of, 27, 30­31b1.3 community-based financing schemes (CF) See also performance Asian experience, 132 community-based health insurance assessment of institutional (CBHI), Rwanda, 61b2.1, determinants of, 18, 19t1.3 104­07t2.11, 207, 251, 275­76 classification of, 128­31 analysis of research results, 258­71 conclusions and recommendations, background, 253­54, 273nn1­2 41­43, 92 conclusions, 272­73 core characteristics of, 19t1.3, data sources and methodology 100­114t2.11 analysis, 254­58, 273n3 differentiation among schemes, 123 demand for, 262­65 economic gains and quality of care, first year performance, 252, 252t7.1 125­26, 127f3.2 impact on utilization and features of, 8­9 out-of-pocket expenditures, 289t8.7 Hsiao's classification of, 29, 31 independent variable used to impact of, 24­27 determine probability of, 256t7.2 India, 208, 295­301, 304­13 See also mutual health insurance links to community-level social capital, schemes 10t1.1, 12­13 community characteristics links to mainstream public economics, as determinants of inclusion in CF 10t1.1, 13­14 schemes, 213, 217t5.4, 219, 224 links to microfinance organizations, as determinants of membership in 10­11, 10t1.1 Senegal's health insurance scheme, literature survey, 16­18, 24­29 239t6.3, 241, 241t6.4, 243t6.5, macrolevel cross-country analysis, 245t6.6 22­24, 37­41, 44­45t1.13, 46n2 in SEWA's household survey, methodology for assessing, 14­16 303t9.5, 304 methods of literature review, 54­59 and social inclusion, 36t1.10, 37 microlevel household survey analysis, community cooperatives, 132 10t1.4, 20­22, 35­37 community cost-sharing, 24 objectives of, 14 community financing, 61b2.1 overview, 53­54 background, 122­23, 153nn7­8 and population enrollment, 123­28 terminology use, 121 products valued by members, 125, Community Health Fund, Tanzania, 35, 125t3.1, 154n9 71b2.2, 112­14t2.11, 160 regional review of, 18­19, 29­35 description, 172­73t4.3, 175­76, 184t4.4 430 Index community health funds, 75t2.7, 76­77 DALE. See disability-adjusted life community health worker, 86 expectancy (DALE) community hospitals, Thailand, 348 Dana Sehat, Indonesia, 122, 126, 128, community prepayments, 24 146­47 community provider-based health Dangme West Health Insurance insurance, 24 Scheme, Ghana, 35, 161, competition, 91, 389 172­73t4.3, 181­83, compulsory insurance, 320 184t4.4, 194n4 Congo. See Democratic Rep. of Congo da Silva, A., 350 consumer cooperatives, 131, 132 data control group, and self-selection, 236 CBHI scheme, 254­58, 273n3 cooperative health care, success and health card programs, 327­33 weakness of, 128, 130 Medical Insurance Fund, Cooperative Health Care Scheme, China, 277­78, 291n1 80­81, 110­11t2.11 reliability of, 401, 415n5 Cooperative Medical System, China, 122, sources of health expenditures, 375 126, 128 sources of in household surveys, control and management of, 142 209­10 desire for, 141­42 Thai Health Card Program, 327­28, overview, 135, 137­38 328­33 reasons for lack of CF, 145, 145t3.11 WHO's database, 400 survival of, 138­40 Dave, P., 69, 86, 350 copayment schemes, 269 Democratic Republic of Congo, 35, affordability of in CHF schemes, 176 289t8.7 African experience, 34 See also Bwamanda Hospital Insurance Bwamanda scheme, 180­81 Scheme GK, 151 Demographic and Health Surveys, 20, and health care accessibility, 163­64 226, 227t5.6 Rwanda, 207 Rwanda, 210, 228n5, 254, 274n3 cost-effectiveness analysis, 370 demographic characteristics cost recovery, 71b2.2, 159, 324 of new card purchase and nonpurchase from prepaid premiums, 73, 74t2.6 of, 329t10.7 cost sharing, 24, 64, 65t2.4, 66­67t2.5 and social inclusion determinants, and characteristics of CF scheme, 35­37 100­114t2.11 See also community characteristics; performance variables reported in individual characteristics reviewed studies, 93­99t2.10 design characteristics, health insurance credit and credit programs, 11, 131 schemes, 167­71 Creese, A., 26b1.1, 63, 70b2.2, 73, 86, 122, Desmet, A., 80 275, 288 Destitute Card, 130, 152 Criel, B., 63, 73, 81, 88, 91, 160­61, 276 developing countries, and capital flows to, critical income, 164 388f11.13 crop insurance, 11 Dhaka Community Hospital Health cross-country analysis of community- Insurance Program, Bangladesh, based financing schemes, 22­24, 122, 130, 151­52 37­41, 44­45t1.13, 46n2 Diop, François, 201­30, cross-subsidies, 269 251­74, 419 curative health care, 158, 159, 254, disability-adjusted life expectancy (DALE), 273n7, 294, 300 14, 22, 23, 37, 39, 398, 401, Burundi, 174 409, 415n1 equity in accessibility of, 265­66, disease-induced poverty, 80 267t7.10, 273nn6­7 distance gradient, 81­82 Index 431 distance to hospitals and health facilities, Egypt, 121 76, 77, 79, 245 elderly as determinant of inclusion, 222 as determinant of participation in and enrollment in Nkoranza scheme, insurance scheme, 242, 243t6.5 179, 179f4.2 frequency of illness, 288 and enrollment in PPS, 263 emergency treatments, 165 Ghana scheme, 186 Employees' State Insurance Scheme, and probability of health-seeking India, 80, 211, 295t9.1, 301 behavior, 266 employment Rwanda, 258­59 and health card program, 329t10.7, district-based health insurance 330t10.8, 331t10.9, 332t10.10, schemes, 161 336t10.11, 337t10.12, 338t10.13, district hospitals, 316 339t10.14 district insurance management teams, 182 occupational group relationship to donor aid, 390 health service utilization, 316 donor policy, 190, 191f4.5 endogenity, 236, 238, 247n3 Dror, David, 3­51, 419 enrollment drug markets, 78 in Africa scheme, 187, 187f4.3 drugs community characteristics as in Abota scheme, 178 determinant of, 224 availability of, 82 factors important in, 264­65 cost and CAM cards, 174 in Nkoranza scheme, 179, 179f4.2 distribution at village level, 127­28 one-time enrollment fee, 223 mark-up of prices, 141 in PPS scheme, 207 Niger, 78 in Rwanda's scheme, 253, 259, payment for, 142 272, 273n1 in SEWA's scheme, 302­04, 306­07t9.6 East Asia, 385 and travel distance, 186 economic growth, relationship to in voluntary health insurance achieving MDGs, 366­73 programs, 202 economics Epidemic Prevention Service, China, 141 gains valued by members, 125­26, expenditures, 376, 378t11.1, 379, 127f3.2 379­82ff11.8­11.9, 383t11.3, 391n3 and willingness to prepay, 124­25 Bwamanda scheme, 181 economic status index (ESI), 279t8.1, and CAM cards, 82 280, 285 Dangme West scheme, 183 Ecuador, 82 data sources of, 375 education, 403 distribution of, 22 as determinant of out-of-pocket funding sources, 384 expenditures on treatment by type and GDP growth, 384­85, 386f11.11 of care, 312t9.9 global, 386, 386f11.11 as determinant of using private health care as percent of, 6 facilities, 310, 310­11t9.8 health care relationship to poverty, 142 and enrollment in health insurance and health outcomes, 209t5.2 scheme, 304 on management, organizational and enrollment in SEWA scheme, structures, and institutional 306t9.6 environment, 371­73 and health card program, 329t10.7, monthly, 261, 262f7.1 330t10.8, 331t10.9, 332t10.10, in Nkoranza scheme, 180 336t10.11, 337t10.12, 338t10.13, per episode of illness, 267­71, 339t10.14, 341­42t10.15 270t7.12, 273n9 as MDG, 363f11.1b, 364 as predictor of financial protection, 214 432 Index expenditures--Cont. Free Medical Care for the Elderly Scheme, statistics on distribution of, 260, 319, 320, 340 261t7.7 Free Medical Care for the Low Income total and public spending on health Household Scheme, 319 care, 368, 391n2 fundraising activities, India, 296 trends in Thailand, 316­19 See also out-of-pocket expenditures Gandaho, T., 70­71b2.2, 71 gender bias, 294­95 family cards, 148 gender equality, as MDG, 363f11.1b, 364 Family Health Card, 130, 152 general taxation, 399, 400 farmers' groups, 316 genocide, Rwanda, 204 FDI. See foreign direct investment (FDI) geographic barriers, 81­82 federations, 207, 253, 254 Ghana fee-for-service (FFS), 79 health insurance policy, 193b4.1 barriers to, 141 impact of CBHI schemes on China, 138 hospitalizations and out-of-pocket DCH, 152 expenditures, 289t8.7 field interviews, 327 WTP projections, 188 field workers, 178, 179 See also Dangme West Health Insurance Fifth National Health Development Plan, Scheme; Nkoranza Community Thailand, 316 Financing Health Insurance Scheme financial protection, 41, 53, 75 Gilson, L., 88 assessment of, 26­27 Global Trust Fund for HIV/AIDS, 387­88 as CF performance variable, 59t2.3, Goalpur, 299t9.3 92­99t2.10 Gonosasthya Kendra Health Care System, conclusions from literature review, 68, Bangladesh, 76, 101t2.11 78­82 Gonoshasthaya Kendra Health Care determinants of, 14, 21­22, 31, 46n1, System, Bangladesh, 150­51 214­15, 220, 225, 228n7, 304­13 government insurance, 24, 75t2.7, 77, and health policy, 190, 191f4.5 78t2.8, 82 impact of MHOs on, 237 performance variables reported in level of, 302 reviewed studies, 93­99t2.10 origins of rich-poor differences support schemes, 64, 65t2.5, 66­67t2.5 in, 4­8 Grameen Bank, 80, 149­50 relationship to type of coverage, 313 Grameen Health Program, Bangladesh, stages of, 42f1.8 80, 101t2.11, 122, 131, 296 success of, 84­92 description of, 64, 149­50 financial resources, pooling of, 6­7 establishment of, 131 financial risk, exposure to, 10­11, 10t1.1 as microfinance organization, 76 financial risk protection, 159, 164­65, 397 gross domestic product, 375 performance in selected schemes, 183, and expenditures, 380­81, 384­85, 184t4.4, 185 386f11.11 foreign direct investment (FDI), 388, health expenditures as share of, 4, 316, 388f11.13 317t10.2, 386, 387f11.12, 391n3 formal sector and public health expenditures, and health care schemes in Africa, 377f11.7 157­58 taxation capacity as percent of, 6 links to informal sector, 159 and total health expenditure, 377f11.7 Rwanda, 204 gross national product Sub-Saharan Africa, 232­33 health expenditures as percent of, 316, Thailand, 319 317t10.2 for-profit schemes, India, 296 of low-income countries, 203 Index 433 gross provincial product, 326 drug funds, 77 Guinea, 71­72, 78, 100t2.11 factors that determine success of, 325­26 Guinea-Bissau. See Abota Village Indonesia, 146­47 Insurance Scheme; prepayment Tanzania, 160 schemes; village health workers Thailand, 77, 90, 147­49 Gumber, Anil, 201­30, 293­314, 420 utilization of, 342, 343t10.16 variables to predict purchase vs. Harvard University, China nonpurchase of, 336t10.11 Network­Harvard Study, See also Carte d'Assurance Maladie, 138, 139 Burundi Hausman test, 216, 236 health care health attainment indexes, 401­03 determinants of use of, 14­16, 21­22 health card funds financing sources, 317, 318t10.3 Tanzania, 160 impact of membership in insurance Thailand, 148­49, 348­49 scheme on, 244­46 Health Card Program, Thailand, 77, 90, increased spending on priority 100­111t2.11, 122, 128, 322­23 populations, 368­71 analytic methods of study, 328 India, 205­06, 293­95 assessments and studies of, 323­26 low- and middle-income countries, attitudes among cardholders and 119­21, 153n1 noncardholders, 345­47 probability of professional visits, 257, background, 315, 320, 321t10.5 257t7.3 characteristics of new purchase and Rwanda, 204­05 nonpurchase of, 329t10.7 Senegal, 205 conclusions and recommendations, Thailand, 206 352­54 type of burden on households, 295t9.1 continuity of card purchase, 340­44 use of microinsurance in, 11 data sources, 327­28 utilization patterns, 316t10.1 description of data in study, 328­33 See also access to health care; discussion of study, 335­40 expenditures; specific health care dropout and continued card purchase, programs 330t10.8 health care facilities, 131 health card nonpurchase and dropout Ecuador, 82 groups, 331t10.9 percent of operational costs financed health care seekers by purchase or by CF scheme, 71b2.2 nonpurchase of health card, perception of, 351 341­42t10.15 Rwanda, 253 nonpurchase and dropout groups vs. health care providers, 389 continued and new card purchase, payment for, 153n4 337t10.12 villages, 176, 177 program sustainability, 349­52 health care systems regression analysis, 333­35, 344, flow of funds through, 4, 5f1.2 344t10.18, 345t10.19 goals and functions of, 398­99 results of study, 328 health financing and link to study area, 326­27 attainment of goals, 399­400 unit cost at health service unit, 324, health characteristics, in SEWA model, 325t10.6 303t9.5, 305 variables that predict purchase or health conditions nonpurchase of, 336t10.11 alternative interventions for, 369, health card schemes 369f11.4, 370 behavior of card users, 324 as determinant of inclusion in CF, DCH, 130, 152 218­19 434 Index health delivery systems, Thailand, 316 insurance based, 190­92 health education, 300 private sector investment and health-enhancing activities, 4 differential pricing, 388­89 health expenditure per capita, 23, relationship to institutional factors, 39­41, 403 189­90 establishing production frontier for, health-seeking behavior, 266, 341­42, 376­84, 391n1 343, 343t10.17, 344 impact of, 405 and probability of professional care health financing visit, 267t7.10 conclusions, 409­10 Rwanda, 256­57 estimation results, 407­09, 413t12.4, HIV/AIDS, 365, 378t11.1, 387­88 414t12.5, 415n6 home remedies, 119 global organization of, 400­01, Honduras, 63 412t12.3, 415nn3­5 Hongvivatana, T., 324, 330 modeling impact of organizational hospitalization, 33, 33f1.7 form of, 401­09 Bwamanda scheme, 181 organization form of and its link to costs of, 141, 282t8.3, 290 health system goals, 399­400 as determinant of inclusion in CF, 219 specification of basic model, 403, fees at St. Jean de Dieu Hospital, 234, 405­06 234t6.1 specification of enlarged models, 406­07 frequency of, 288, 289, 289t8.7 See also community-based financing and membership in insurance schemes; community-based health fund, 285 insurance and membership in mutual health health-financing instruments, 4, 5f1.2, 6 insurance schemes, 245­46, Health Insurance Card Scheme, Thailand, 248nn8­10 208­09 in SEWA model, 278, 279t8.1, health insurance schemes, 157 280, 305 DCH system, 152 hospitals demand for, 255­56, 262­65 admissions, 81 design and institutionalization of, China, 137 167­71 Rwanda, 82 determinants of membership in, utilization of, 276 238­44, 248nn6­7 See also distance to hospitals and enrollment in as determinant of health facilities expenditures on treatment by type hospital stays, 87, 228n3 of care, 312t9.9 household budgets, 20 enrollment in as determinant of use of household budget surveys, 226, 227t5.6 private facilities, 310, 310­11t9.8 household characteristics, 251, 259, exclusions from, 400, 415n3 261t7.6 factors influencing demand for, 242 and CF, 217t5.4, 218­19, 221t5.5 and FRP, 165 and health card program, 329t10.7, Ghana, 193b4.1 330t10.8, 331t10.9, 332t10.10, India, 206, 294 336t10.11, 337t10.12, 338t10.13, Thailand, 319­21 339t10.14, 341­42t10.15 urban and rural areas in Sub-Saharan and hospitalizations, 245, 246 Africa, 233­34, 233f6.1 impact of on enrollment in health health maintenance organizations insurance schemes, 262­64 (HMOs), 154n10 by income quartile, 259, 261t7.6 health policies and out-of-pockets expenditure and financial protection, 225 patterns, 38t1.11 Ghana, 193b4.1 Rwanda, 260t7.5, 261t7.6 Index 435 and Senegal's health insurance scheme, relationship to health card purchase, 238­41, 242, 243t6.5, 244t6.6 330­31 in SEWA study, 279t8.1, 280, 303t9.5, as reported by households in SEWA 306t9.6 scheme, 305 and social inclusion, 36­37, 36t1.10 illness-related financial risks, 158­59 Thai health card program, 327­28 income and use of private facilities, 310, and child mortality, 366, 367f11.3 310­11t9.8 China, 135 and viability of health insurance as determinant of inclusion in CF schemes, 246 schemes, 213, 216­18, 222­24 household data, microlevel, 42 and economic growth relationship to household surveys MDGs, 366­73 characteristics of instruments, 210t5.3 and enrollment in SEWA's scheme, criteria for mutual to be included, 306t9.6 235t6.2 and expenditures, 7, 159, 312t9.9, 376, data sources, 209­10 379, 385f11.10 determinants of financial protection by and financial protection, 164­65, CF schemes, 214­15, 220, 228n7 215, 220 determinants of social inclusion in CF and health card program, 329t10.7, schemes, 212­14, 216­24 330t10.8, 331t10.9, 332t10.10, India, 208, 210t5.3, 211­12, 301­02 336t10.11, 337t10.12, 338t10.13, limitations of methodology, 215­16 339t10.14, 341­42t10.15 list of reviewed instruments, 226, households in Rwanda, 259, 261t7.6 227t5.6 and insurance premiums, 246 and management of CF schemes, 126 and membership in Senegal's health method for analysis of, 10t1.4, 20­22, insurance scheme, 238­42, 243t6.5, 35­37 244t6.6, 248nn6­8 prepayment schemes, 254­55 percent spent on curative care, 294 questions omitted, 286­87 and probability of health-seeking Rwanda, 206­07, 210­11, 210t5.3, behavior, 266 228nn1­4 income poverty indicator, 362 Senegal, 207­08, 210t5.3, 211 income redistribution, 165 Thailand, 208­09, 210t5.3, 212, index of distribution of responsiveness 323, 327 (IRD), 22, 23, 401, 408, 409, Hsiao, William C., 3­51, 63, 80, 413t12.4, 414t12.5 119­35, 421 index of equality of child survival (IECS), hunger, 362, 363f11.1a 22, 23, 40, 401, 408, 409, 413t12.4, 414t12.5 illness index of fairness of financial contribution as determinant of inclusion in (IFFC), 22, 23, 24, 39, 401­03, 408, CF, 219 413t12.4, 414t12.5 as determinant of using private index of level of responsiveness (IR), 22, facilities, 310, 310­11t9.8 23, 30­41, 401­03, 404t12.2, 408, and health card program, 329t10.7, 413t12.4 330t10.8, 331t10.9, 332t10.10, India, 36­37, 69, 121, 208, 217t5.4, 336t10.11, 337t10.12, 338t10.13, 218, 222 339t10.14 background of health care service, and health services utilization patterns, 293­95 316t10.1 determinants of utilization and out-of- and out-of-pocket expenditures, pocket expenditure patterns, 221t5.5 267­71, 273n9 and financial protection, 220 rates of, 175 health care, 205­06 436 Index India--Cont. health insurance schemes for, 162­64 health expenditures, 209t5.2, 376, industrial nations, 153n3 377f11.7, 378t11.1, 380 and insurance scheme design factors, health outcomes, 209t5.2 169­70 health status, 219 links to formal sector, 159 household surveys, 210t5.3, 211­12 and SEWA, 300 impact of CBHI schemes on and willingness-to-pay, 160 hospitalizations and out-of-pocket information expenditures, 289t8.7 about health cards, 352 in-kind contributions, 86 and social inclusion, 223 insurance use, 220 transfer of, 170 socioeconomics, 203, 204t5.1 in-kind contributions, 86b2.4, 146, 296 survey instrument inpatient care, determinants of use of, characteristics, 20t1.4 310, 310­11t9.8 See also Medical Insurance Fund, SEWA inpatient costs to premium ratio, 180 individual characteristics in-service training, 182 and choice of ambulatory treatment, 305 Institute for Health and Development, as determinant of inclusion in CF, Dakar, 210t5.3, 211, 235 217t5.4, 218­19, 221t5.5 institutional characteristics, 18, 19t1.3, 160 as determinant of membership in Africa, 34, 167­71, 185­89 Senegal's health insurance scheme, and analytical framework for literature 238, 239t6.3, 240t6.4, 242, 243t6.5, review, 58f2.1 244t6.6 and CF schemes in reviewed literature, as determinant of using private 100­114t2.11 facilities, 310, 310­11t9.8 and forms of community financing, as determinants of fund membership, 65t2.4 282, 283t8.4, 284t8.5 and resource mobilization, 84t2.9, 90­92 determinants of out-of-pocket strengths of, 29b1.2 expenditures on treatment by type weaknesses of, 31b1.3 of care, 312t9.9 institutional environment, spending on, determinants of utilization and 371­73 out-of-pockets expenditure insurance patterns, 38t1.11 benefits of, 153n8 and health card program, 329t10.7, as determinant of financial 330t10.8, 331t10.9, 332t10.10, protection, 220 336t10.11, 337t10.12, 338t10.13, GK system, 151 339t10.14 Grameen Health Program, 150 and SEWA scheme, 278, 279t8.1, 280, lack of, 34 281t8.2, 303t9.5, 304, 306t9.6 as microfinance instrument available and social inclusion, 36t1.10, 36­37 to poor, 10t11.1, 11­12 Thai health card program, 327­28 Tanzania, 175 and viability of health insurance third-party, 130 schemes, 246 willingness to pay for, 188, 188f4.4 individual savings accounts, 63 See also health insurance Indonesia, 63, 126, 146­47 insurance advisory boards, 178 industrial countries, 4, 153n3 insurance benefit packages, 164, 165 industrial workers, 301 insurance companies, 206, 294 infant mortality, China, 138 insurance management teams, Nkoranza informal sector, 153n1 scheme, 178 Africa, 159 insurance schemes features of large population schemes Africa, 162­64, 184t4 for, 172t4.3 conclusions, 192­93 Index 437 Democratic Rep. of Congo, 172­73t4.3, Langenbrunner, John C. (Jack), 180­81, 184t4.4 361­96, 422 features of large population schemes, Latin America, 54, 57t2.2, 73, 385 172t4.3 Levy-Bruhl, D., 70­71b2.2 goals matched to design options, life expectancy, 378t11.1 169t4.2 life insurance, 11 institutional factors and policy Life Insurance Corporation of India, 300, environment, 189­90 301t9.3 institutional influences on design of, literature review 185­89 analytical framework, 55, 58f2.1, performance overview, 183­85 93­99t2.10 policy implications for, 190­92 based on nature of study and by reluctance to join, 162 region, 54, 56­57t2.2 Tanzania, 35, 71b2.2, 112­14t2.11, 160, community-based financing schemes, 172­73t4.3, 175­76, 184t4.4 16­18, 24­29, 41 See also Abota Village Insurance financial protection, 68, 78­82 Scheme, Guinea-Bissau; Carte papers selection criteria, 54­59 d'Assurance Maladie, Burundi; by publication type, 55t2.1 Nkoranza Community Financing resource mobilization capacity, 68, Health Insurance Scheme, Ghana 69­71 Integrated Social Security Scheme, SEWA, selection criteria to assess 276, 288 performance of CF, 55, 59t2.3, interaction variables, 40 93­99t2.10 internally generated funds, 188 social inclusion, 68, 73, 75­77 International Development Goals, 14 summary of case studies by modalities, interviewer bias, 287 66­67t2.5 investments, private sector, 388, 390 Liu, Y., 81, 125 IR. See index of level of responsiveness Living Condition Monitoring Survey, (IR) 273n3 IRD. See index of distribution of Living Standard Measurement Surveys, responsiveness 20, 226, 227t5.6 Islam, K., 76, 80, 85, 88 local initiatives, 264 local providers, 69 Jakab, Melitta, 3­51, 53­117, London School of Hygiene and Tropical 201­30, 421 Medicine, 210t5.3 Janjaroen, W. S., 319 low-income countries Japan, 319 capacity to raise revenues, 6, 6f1.3 Jütting, Johannes Paul, 76, 80, 201­30, GNP, 203 231­49, 290, 421 health care, 4, 119­21, 153n1 microfinance instruments available to, Kanage Community Financed Scheme, 10t1.1, 11­12 Rwanda, 73, 82, 85, 107­10t2.11 pro-rich bias of public subsidies Kanage Cooperative Scheme, in, 8f1.5 Rwanda, 88 public services in, 6 Kegels, G., 276 qualified practitioners in, 154n10 Kenya, 73, 87, 88, 109­10t2.11 revenue pooling, 4, 5f1.1, 386, Kiranandana, T., 323, 324, 325 387f11.12 Kisiizi Hospital Health Society, Uganda, 109­10t2.11 Macroeconomic Commission on Health, knowledge, about health cards, 352 375, 387, 390 Krishnan, Chitra, 53­117, 201­30, 422 Macro International, 273n3 Kulkarni, V., 295 malaria, 80, 365, 371 438 Index management results of study, 280 African insurance schemes, 170­71 strengths and limitations of study, of CF schemes, 139, 140t3.7, 142, 285­90, 291n3 146­47, 175, 223 variables in study, 278­80 of PPS, 264 medical savings schemes, 166 spending on, 371­73 Mediclaim, 211, 295t9.1, 300, 305 Thailand's health card scheme, 148 membership cards, Senegal, 208 villages, 153n7 methodology management characteristics, 19t1.3 for assessing impact, strengths, and and analytical framework for literature weaknesses of community-based review, 58f2.1 financing, 14­16 and CF schemes in reviewed literature, limitations of, 215­16 100­114t2.11 for mutual health insurance scheme as concern of CF schemes, 125­26, models, 236­38, 247n3, 247n5 127f3.2 MHOs. See mutual health organizations and forms of community financing, microcredit schemes, 296­301 65t2.4 microfinancing, 10t1.1, 10­11, 159 motivation and competency of, 123 microinsurance, 10t1.1, 11, 12, 61b2.1 and resource mobilization, 84t2.9, microlevel household data, 42 88­89 micropurchasing, 166, 169 strengths of, 28b1.2 middle-income countries, health care, weaknesses of, 30b1.3 119­21, 153n1 Manopimoke, S., 324 Millennium Development Goals (MDGs) marital status background, 361, 362b11.1, 391n1 as determinant of out-of-pocket cost of accelerating progress of, 389­90 expenditures on treatment by type cost of achieving, 373­84 of care, 312t9.9 and economic growth, 366­73 and enrollment in SEWA scheme, financing expenditure gap, 384­89 306t9.6 global aggregate, 363f11.1 and health card program, 329t10.7, and production frontiers, 373­75 330t10.8, 331t10.9, 332t10.10, progress toward achieving, 361­66 336t10.11, 337t10.12, 338t10.13, Ministry of Health, Rwanda, 204, 251 339t10.14, 341­42t10.15 Ministry of Public Health, Thailand, 316, market-based organizations, 13, 201 317, 322 markets, 91, 120, 153n5 modalities maternal health, 363f11.1c, 364­65 and characteristics of CF financing in maternal mortality, 368, 378t11.1 reviewed studies, 100­114t2.11 relationship to economic growth, community financing, 64, 65t2.4, 366­67 66­67t2.5, 69­71, 75, 75t2.7, 76­77 and use of production frontiers, 375 and financial protection, 68, 78­82 Mburahati Health Trust Fund, Tanzania, performance variables reported in 104­7t2.11 reviewed studies, 93­99t2.10 MDGs. See Millennium Development summary of case studies by, 66­67t2.5 Goals Monasch, R., 26b1.1, 63, 70b2.2, 73, 86, Medical Insurance Fund, SEWA, 275, 276 122, 275, 288 data collection and analysis, 277­78, moral hazard, 175, 181, 276 291n1 mortality rates, 364, 376 models of membership, 278, 281t8.2, and expenditures on health care, 282t8.3, 291n2 378t11.1, 379­80f11.8, 381­82f11.9, recommendations and conclusions of 383t11.2, 383t11.3 study, 290­91 low-income countries, 204 regression analyses, 282­85 and MDGs, 363f11.1c Index 439 Muller, C., 261 description of, 172­73t4.3, 178­80, Municipality Health Office, 184t4.4 Indonesia, 147 management of, 89b2.5 Musau, S., 71b2.2 organizational characteristics, 90 Musgrove, Philip, 370, 397­416, 422 participation in, 161 mutual health insurance schemes, 61b2.1 nongovernmental organizations (NGOs), determinants of participation in, 33, 35, 294 240­41t6.4 and GK health care system, 26b1.1, features of in Senegal, 234 150­51 impact of membership on health care India, 205, 206 services, 244­46 and managed health insurance methodology of research, 236­38, schemes, 297­98t9.2, 299t9.3 247n3, 247n5 nonoccupational illness, 320 overview, 231­32 nonprofits, 294 research design, 235­38 nonprofit sector, Senegal, 205 Senegal, 207, 211 nonurgent care, 81 variables used in research, 239t6.3 not-for-profit schemes, 64 See also community-based health insurance observation bias, 287 Mutual Health Organization, Senegal, occupational health services, 300 76­77, 87, 102­4t2.11, 211 one-time enrollment fee, 223 mutual health organizations (MHOs), 24, operational characteristics, and 64, 65t2.4, 73, 75t2.7 improvements in efficiency and case studies by modalities, 66­67t2.5 quality, 127­28, 129t3.2, definition of, 61b2.1, 163 154n10 determinants of participation in, 302 operational costs, household expenditures impact on financial protection, 237 as percent of, 71­72 make-up of, 253 operational revenues, contribution of CF management of, 88 schemes to, 70­71b2.2 performance variables reported in Organisation for Economic Co-operation studies reviews, 93­99t2.10 and Development, 372 Senegal, 76­77, 87, 102­04t2.11, 211 organizational characteristics, 19t1.3 mutual-provider partnership African insurance scheme, 171 models, 163 and analytical framework for literature CHF, 172­73t4.3, 175­76 review, 58f2.1 Dangme scheme, 35, 161, 172­73t4.3, and CF schemes, 65t2.4, 100­114t2.11 181­83, 184t4.4, 194n4 GB credit program, 150 Mutuelle Famille Babouantou de and resource mobilization, 84t2.9, Yaounde, Cameroon, 102­04t2.11 89­90 strengths of, 29b1.2 National Council of Applied Economic weaknesses of, 30b1.3 Research, 210t5.3, 211 organizational structures, spending on, National Health Card Insurance Scheme, 371­73 Burundi, 73 outcomes, 119 national health policy, 190­92 classification and assessment of, National Sample Survey (NSS), 294 132­33, 134t3.3 New India Assurance Company, 300, and expenditures, 209t5.2 301t9.4 improvement of, 371, 372f11.5, 373 Niger, 72, 78, 289t8.7 measures of, 391n2 Nkoranza Community Financing Health out-of-pocket expenditures, 24 Insurance Scheme, Ghana, 35, 64, Africa, 34, 158 77, 85, 107­10t2.11, 276 analysis of, 78, 79­81, 79t2.8 440 Index out-of-pocket expenditures--Cont. Permpoonwatanasuk, C., 324 Burundi, 172­73 Pesos for Health, 86 as determinant of financial pharmaceutical industry, 389 protection, 220 Piyaratn, P., 319 determinants of, 37, 38t1.11, 221t5.5, policies, and community-based financing 311, 312t9.9, 313 schemes, 10t1.1, 13 factors that influence, 269, 270t7.12 Poullier, Jean-Pierre, 397­416, 422 financial impact of, 258, 259t7.4 poverty for hospital care, 220 Bangladesh, 76 and households below poverty line, 7 China, 138 impact of CBHI schemes on, 289t8.7 determinants of, 369 per episode of illness, 267­71, disease-induced, 80­81 271t7.13, 273n9 global nature of, 361­62, 363f11.1a and poverty without risk sharing, hard core poor, 77 7, 7f1.4 and health care, 120, 142, 293 as source of health care and health financing sources, 142­43, financing, 317 143t3.8, 145t3.10 outpatient benefit package, 161 impact of CF on, 202 outpatient visits, 81 and microcredit schemes, 296, and card usage rates, 325 300, 301t9.4 charge for, 141 national poverty line, 203 Thailand, 348 origins of rich-poor differences in Overseas Development Assistance, 387 financial protection, 4­8 ownership, 91, 223 and out-of-pocket expenditures, 7, 7f1.4 prevention of, 27t1.7 paramedics, 151 rates of, 285 participation rates, 160, 194n1 relationship to hospitalization, 33, Partnerships for Health Reform (PHR), 33f1.7 206­07, 210t5.3, 251, 252 rural women, 131 payment for health care, potatoes as, See also low-income countries 86b2.4 poverty alleviation payment schedules, 85­86, 209 Grameen Bank, 149­50 in Abota scheme, 178 and microfinance organizations, Bwamanda scheme, 181 10­11, 10t1.1 Dangme West scheme, 183 practitioners, competency of, 127, 154n10 Nkoranza scheme, 179 pregnancy, 218, 300 performance, 55, 59t2.3, 201­03 impact on enrollment in insurance assessment of, 11­12 plan, 263, 264 discussion of results of review, and probability of health-seeking 82­83, 114n1 behavior, 266 impact of financial protection on, 68, Preker, Alexander S., 3­51, 86, 201­30, 78­82, 84­92 361­96, 417 impact of resource mobilization premiums capacity on, 68, 69­73, 84­92 Abota scheme, 178 impact of social inclusion on, 68, 73, affordability of, 176, 178 75­77, 84­92 Africa scheme, 187, 187f4.3 of insurance schemes, 183­85 Bwamanda scheme, 180­81 overview of, 68 Dangme West scheme, 183 selection criteria, 55, 59t2.3, GK system, 151 93­99t2.10 Nkoranza scheme, 178, 180 variables reported in reviewed studies, rates in Rwanda, 228n2 93­99t2.10 relationship to benefits, 174­75 Index 441 prepayment schemes, 122, 153n5, 153n8 professional care, 257, 257t7.3, 265­66, Asia, 77 267t7.10, 273n6 and average per illness episode, expenditures per visit, 270, 271t7.13 267­71, 273n9 probability of using, 266t7.9 CMS sources, 137 PROSALUD, 153n5 cost recovery from, 26b1.1, 73, 74t2.6 provider-based health insurance, 64, determinants of willingness to pay, 65t2.4, 66­67t2.5, 75t2.7, 77, 79t2.8, 123­25 81­82 features of, 264 models, 163 and financial protection, 26­27, 225 Nkoranza scheme, 172­73t4.3, 178­80, Guinea-Bissau, 35, 102­4t2.11, 160, 184t4.4 172­73t4.3, 176­78, 184t4.4 performance variables reported in impacts of on outcome indicators, 31, studies reviews, 93­99t2.10 32t1.9, 33 provider behavior, 90 income as determinant of inclusion in provider-payment mechanisms, 254 CF schemes, 216­18, 222­24 provinces, 322, 326 India, 296 public facilities, 305, 308­09t9.7 and MHOs, 64, 65t2.4, 66­67t2.5 public finances, links to, 10t1.1, 13­14 Niger, 72 public funding, for health care, 120, percent of recurrent costs from, 70­71b2.2 153nn3­4 provider-sponsored, 130 public health care, 204, 206, 369, as revenue source, 85 369f11.4 Rwanda, 76, 207, 228n3, 251, 253, public health expenditure, 40, 41, 401, 407 254­55 public-private funding of health care, 4, and types of health expenses, 135, 5f1.2, 6 154n14 purchasing functions, 87 voluntary, 73, 74t2.6 preventive health care, 141, 158, quality of care, 78, 82 254­55, 300 disintegration of, 142 equity in accessibility of, 265­66, and operational characteristics, 267t7.10, 273nn6­7 127­28, 129t3.2, 154n10 primary care, affordability of, 125, as valued by members, 125­26, 127f3.2 125t3.1, 154n9 questionnaires primary sampling units, 277 Thailand study, 327, 328 principal-agent problems, 13 See also household surveys private health facilities, 205, 305, 308­09t9.7, 310, 310t9.8 radio ownership, 218­19, 263, 264 Private Health Insurance, 320 RAHA scheme, 69, 86, 299t9.3 private sector Rand Corporation, 130 and equity of health care, 13 Rand Health Insurance Experiment, 46n1, investments and differential pricing 143, 228n7, 278, 302 policies, 388­89, 390 Ranson, M. Kent, 201­30, 275­92, 423 Senegal, 205 ratios of insurance protection, 80­81 types of private financial flows, 388f11.13 referral systems producer cooperatives, 131 and health cards, 324, 332­33 production frontier implications of, 347 application of analysis in global inadequacy of, 169 expenditure gap, 375 rates, 82 for health expenditure per capita, regional reviews 376­84, 391n1 Africa experience, 18­19, 34­35 past use of, 374­75 Asia experience, 18­19, 29, 31­34, use of, 373­75 42­43 442 Index registration periods villages, 153nn7­8 Dangme scheme, 182 See also risk-sharing arrangements Nkoranza scheme, 179 risk protection, 186 religion, 241, 242, 281t8.2 Abota scheme, 178 religious organizations, 204 Africa, 34, 164 resource allocation, 87, 370 Bwamanda scheme, 180­81 resource-generating instruments, 73 CAM, 174 resource mobilization, 25­26, 159 CHF, 176 Abota scheme, 178 Dangme West scheme, 183 Africa, 34, 166­67, 186­87 Nkoranza scheme, 179 Bwamanda scheme, 181 rich-poor differences, 11 CAM, 174­75 risk-sharing arrangements, 4, 5f1.2, 6, 13, capacity of schemes, 53 22­24, 162, 319, 397 and CF scheme, 123­28, 135 categories of, 400­01, 402t12.1, CHF scheme, 176 403­06, 412t12.3 conclusions about from literature conclusions, 409­10 review, 68, 69­73 estimation results, 407­09, 413t12.4, Dangme West scheme, 183 414t12.5, 415n6 determinants of success of, 84­92 impact of, 410 and management characteristics, modeling of, 401­03, 404t12.2 84t2.9, 88­89 and revenues, 37­41 Niger, 72, 78 specification of basic model, 403, Nkoranza scheme, 180 405­06 as performance variable, 55, 59t2.3, specification of enlarged models, 93­99t2.10 406­07 responsiveness, 398­99 Thai health card program, 322, 344 See also index of distribution of Roenen, C., 73, 88 responsiveness (IRD) rural areas revenues health insurance schemes in Sub- in Abota system, 177 Saharan Africa, 233­34, 233f6.1 in Africa, 187, 187f4.3 health systems, 4, 316, 316t10.1 Bwamanda scheme, 181 public services in, 6 CAM card sales in Burundi, 174 Rural Cooperative Medical System, China, capacity of low-income countries to 80­81, 91, 110­11t2.11 raise, 6, 6f1.3 rural health facilities, Ecuador, 82 Dangme West scheme, 183 Rural Health Workers Development MHO resources, 73 Project, China, 139 Nkoranza scheme, 180 Rwanda, 76, 222 operational, 70­71b2.2 CF schemes, 206­07, 217t5.4, 218, 219, pharmaceutical industry, 389 222­23, 224, 228nn1­4 pooling of in low-income countries, 4, community characteristics, 219, 224 5f1.1, 386, 387f11.12 demand for health insurance, 255­56 sources of, 295­96 determinants of inclusion in CF and technical design characteristics, schemes, 36­37, 217t5.4, 218 84t2.9, 85­87 determinants of utilization, 221t5.5 revenue to expenditure ratio, 180 and financial protection, 220 risk-management capacity, 231 health care, 204­05 risk pooling, 86­87, 120­21, 132, 154n11 health outcomes and expenditures, and financial protection, 225 209t5.2 impact of CF on, 133 household surveys, 206­07, 210­11, and prepayment schemes, 124 210t5.3, 228nn1­4 Thai health card program, 149 impact of CBHI schemes, 289t8.7 Index 443 income impact, 218 and financial protection, 220 and out-of-pocket expenditures, health care, 205 221t5.5, 258, 259t7.4, 289t8.7 health outcomes and expenditures, social inclusion determinants, 36­37 209t5.2 socioeconomic determinants, 203, household surveys, 210t5.3, 211 204t5.1, 220 impact of CBHI schemes, 289t8.7 statistics on distribution of impact of membership on access to expenditures, 260, 261t7.7 health care, 244­46 survey instrument characteristics, income as determinant for inclusion, 20t1.4 217­18 See also community-based health mutual health insurance research insurance, Rwanda design, 235­38 Rwandan National Population Office, 210, organizational characteristics, 89­90 210t5.3, 254, 273n3 overview of health insurance scheme in, 234­35 sanitation systems, 364 socioeconomics, 203, 204t5.1, 220 savings, 11 survey instrument characteristics, Schneider, Pia, 201­30, 251­74, 423 20t1.4 School Children's Card, 130, 152 service delivery, 120, 143, 153n4 School Health Insurance Scheme (SHI), Sevagram scheme, 299t9.3 319, 320, 340 SEWA. See Self Employed Women's Seguridad Social Campesino (SSC), Association, India Ecuador, 82, 112­14t2.11 SEWA Bank, 300 selection bias, 236­37, 238 Sewagram scheme, 69, 86 self-employed, 400, 415n3 SHI. See School Health Insurance Scheme; Self-Employed Women's Association social health insurance (SEWA), India, 80, 101t2.11, 122, Singapore, 367 210t5.3, 211­12, 218, 223, 228n6 social capital, 202 coverage under, 301t9.4 community level links to, 10t1.1, description, 300 12­13 determinants of being enrolled in, as determinant for willingness to 306­07t9.6 prepay, 124­25 and health care burden on households, downside of, 12­13 295t9.1 social exclusion, combating of, 27, 27t1.8 methodology of study, 302, 303t9.5 social health insurance, 399, 400 program description, 295­301 social inclusion, 53 research design, 301­02 as CF performance variable, 59t2.3, results of research, 302­13 93­99t2.10 See also Medical Insurance Fund, SEWA and community financing, 35­37, self-selection, 236, 238, 247n5 212­14 self-treatment, 316, 343 conclusions from literature review, 68, Senegal, 36, 73, 76­77, 80, 87, 290 73, 75­77 background, 231­32 determinants of, 14­16, 20­21, 35­37, CF schemes, 19, 207­08, 217­18, 212­14, 216­24 221t5.5, 224 and methodology limitations, 216 community characteristics, 219, 224 success of, 84­92 determinants of inclusion, 218, 221t5.5 social insurance, 24, 75t2.7, 77, 78t2.8, determinants of membership in health 82, 313, 415n4 insurance scheme, 238­44, performance variables reported in 248nn6­7 reviewed studies, 93­99t2.10 determinants of utilization and social insurance-support schemes, 64, out-of-pocket patterns, 221t5.5 65t2.5, 66­67t2.5 444 Index social policy, and community-based Tanzania, 175 financing schemes, 10t1.1, 13 Thailand, 208­09 Social Protection Sector Development Supakankunti, Siripen, 77, 88, 201­30, Program, Indonesia, 147 315­57, 424 Social Research Institute, Chiang Mai Suzuki, Emi, 361­96, 424 University, 323 social responsibility, 389 Tangcharoensathien, V., 319 Social Security Scheme, 223, 320 Tanzania. See Community Health Fund, social values and principles, 62 Tanzania social welfare-oriented health insurance taxation, low-income countries capacity scheme, 163­64 of, 6 social welfare programs, Thailand, 148 tax revenues, 208 socioeconomics technical design characteristics, as characteristic of new card purchase 19t1.3, 160 and nonpurchase of, 329t10.7 and African insurance schemes, 170, as determinant of financial 185­89 protection, 220 and analytical framework for literature as determinant of inclusion in CF review, 58f2.1 schemes, 216­18, 222­24 and CF schemes in reviewed literature, India, 203, 204t5.1 100­114t2.11 Rwanda, 203, 204t5.1 and forms of community financing, 65t2.4 Senegal, 203, 204t5.1 and resource mobilization, 84t2.9, 85­87 Thailand, 203, 204t5.1 strengths of, 28b1.2 solidarity association, 162, 182, 272 weaknesses of, 30b1.3 Soucat, A., 70­71b2.2, 71, 78 Thai-German Technical Cooperation for South Asia, 385 Health, 325 Sports Card, 130, 152 Thailand Sri Lanka, 121 background, 147 stakeholders CF schemes, 208­09, 217t5.4, 222 participation in design of insurance health care, 206 schemes, 168­69, 168t4.1 health delivery system, 316 relationship to insurance schemes, health insurance development, 319­22 167­68 health outcomes and expenditures, state domestic product, 285 209t5.2 stewardship, 90­91 health status, 219 Stinson, W., 122 household surveys, 208­09, 210t5.3, 212 St. Jean de Dieu Hospital, Senegal, social inclusion determinants, 36­37 89­90, 205, 207, 208, 234, socioeconomics, 203, 204t5.1 235t6.1, 246 survey instrument characteristics, 20t1.4 St. Theresa's Hospital, Ghana, 178 trends in health expenditures, 316­19 Students' Health Home, 299t9.3 See also Health Card Program, Thailand Study of Thirty Poor Countries, 139, 143, Thai Medical and Health Company 143t3.8 Limited, 320 subnational health insurance third-party insurance, 130, 163, 194n2 schemes, 159 tontine, 264 Sub-Saharan Africa, 385, 386f11.11 Toonen, J., 86b2.4 health insurance in, 232­35 township health centers, China, 137 poverty in, 362 townships, 120, 147 subsidies Tribhuvandas Foundation, 228n6, 291n1, to individuals by government, 128 296, 299t9.3 in low-income countries, 7, 8f1.5 trust, as factor in PPS participation, 264 and prepayments, 123­24 tuberculosis, 365, 378t11.1 Index 445 U5MR. See under-five mortality rates village health workers, Guinea-Bissau, Uganda, 73, 109­10t2.11 176, 177 under-five mortality rates (U5MR), 368, villages 375, 376, 378t11.1 characteristics of, 239t6.3, 241, relationship to economic growth, 241t6.4, 243t6.5, 245t6.6 366­67 and economic gains, 127, 154n10 and use of production frontiers, 375 health care provision for, 120 undernourished, 362 and health-seeking behavior, 245, 246 underweight among children, 366­67, health stations in China, 137, 142­43, 368, 375 143t3.8 United Nations, 361 managing of health fund, 326 United States, 401 prevalence and benefits of CF in China, unity ratios, 166 139, 139t3.5, 140t3.6 universal coverage, 159 as PSU in SEWA study, 277 urban areas, health insurance schemes in, qualified practitioners in, 154n10 233­34, 233f6.1 in Thailand's Health Card USAID, 252, 273n3 Program, 148 user fees, 34, 85 See also Abota Village Insurance Africa, 158 Scheme, Guinea-Bissau Burundi, 162 voluntary community-based health China, 81 programs, 146­47 impact of, 233 Voluntary Health Insurance Scheme, 320, India, 296 321t10.5 institutionalization of, 189 voluntary insurance, India, 206 Niger, 72 voluntary medical insurance, 294 Tanzania, 175 voluntary private health insurance, 153n8 utilization patterns analysis of, 78, 79­81, 79t2.8 Wagstaff, A., 366, 368 Bwamanda scheme, 181 water, 363f11.1a, 364 determinants of, 37, 38t1.11, Waters, H., 238 220, 237 wealth in DRC, 81 measure of, 285­86 and financial protection, 220 origins of rich-poor differences in of health cards, 348, 348t10.22 financial protection, 4­8 impact of CBHI schemes on, relationship to health, 365, 365f11.2 289t8.7 and SEWA study, 280 India, 304­05 West Africa, 69, 85 Nkoranza scheme, 179 Wiesmann, D., 232­33 prepayment schemes, 76 willingness to pay, 34, 35, 124­25 rural health facilities, 82 for adult insurance, 188, 188f4.4 SEWA model, 278, 282t8.3, 305 Africa's informal sector, 160 significant determinants of, 221t5.5 for improved benefits, 175 need for information concerning, vaccination strategies, 4 161­62 Valentine, Nicole, 397­416, 425 use of in design of schemes, 185­88 value added, by CF type, 133, 134t3.3, women 135, 136t3.4 CAM cards, 174 van der Stuyft, P., 63, 81, 91 as determinant of participation in van Lerberghe, W., 63, 81, 91 insurance scheme, 242, 243t6.5 Veeravongs, S., 323, 330, 340, 347 eligibility for membership in SEWA's Vietnam, out-of-pocket expenditures, insurance fund, 276 7f1.4 illness and poverty relationship, 300 446 Index Worker Health Card, 130, 152 Xu, Ke, 397­416, 425 Workmen's Compensation Scheme, 320 World Bank, 14, 138, 387­88 Yip, W., 278 World Development Report 1993, 190, 369, 374 Zaire (Former). See Democratic Rep. of World Health Organization (WHO), Congo 142­43, 204, 375, 397, 399 Zambia, 88 database, 400 Zeramdini, Riadh, 397­416, 426 World Health Report 2000, 14­16, 40­41, zero copayment rates, 164 397, 401, 409 zonal coordinators, 178 "M illions of impoverished people die every year of conditions that can be readily prevented or treated with existing technologies. This book provides valuable les- sons on what communities can do to improve financing of health care for such conditions and interventions at low income levels, as part of a global strategy for increased invest- ments in health." --JEFFREY SACHS Professor of Economics and Director of the Earth Institute, Columbia University, New York, N. Y., Former Chair of the Commission on Macroeconomics and Health, World Health Organization, Geneva "Health is important to development, and development has an important impact on health. Yet in many low-income countries, governments have fallen behind in financing and delivery of health services. This book contributes to our understanding of the limits to government spending on health care at low income levels and the important role that households, communities, and the private sector can play in this respect." --PETER HELLER Deputy Director, Fiscal Affairs Department, International Monetary Fund, Washington, D.C. "Funding and providing health care for the two billion peasants and ghetto dwellers in developing countries remain urgent and vexing problems for the world. This book reviews problems and solutions." --WILLIAM C. HSIAO K.T. Li Professor of Economics, Department of Health Policy and Management, Harvard University, Cambridge, Massachusetts International Labour Office World Health Organization ISBN 0-8213-5525-2