Africa Region Human Development Working Paper Series Health Care on the Frontlines Survey Evidence on Public and Private Providers in Uganda Magnus Lindelöw Ritva Reinikka Jakob Svensson Development Research Group The World Bank Copyright © 2003 Human Development Sector Africa Region The World Bank The views expressed herein are those of the authors and do not necessarily reflect the opinions or policies of the World Bank or any of its affiliated organizations Cover design by Word Express Interior design by Word Design, Inc. Cover photo by Das Fotoarchiv ii Contents Foreword . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .v Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .vii Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .viii Abbreviations and Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .ix 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 2. Background: Health Care in Uganda . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 3. The Health Facility Survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5 Purpose of the Survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5 Survey Design and Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6 4. Public and Private Health Care Providers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .8 Oversight, Management, and Competition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .11 Inputs and Costs at Health Facilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .13 User Fees and Financing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .24 Outputs and Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .28 Client Perceptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .32 5. Summary and Recommendations for Further Research . . . . . . . . . . . . . . . . . . . . . . . .35 Ownership and Health Facility Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .35 Human Resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .36 User Fees and Financing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .38 Drug Use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .39 Appendixes A. Methodology and Data Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .41 B. Consistency between Facility and District Records . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .44 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .46 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .50 iii Tables 1. Number of Health Care Facilities by Ownership Category and Region . . . . . . . . . . . . . . . . .8 2. Characteristics of Health Facilities by Ownership Category and Region . . . . . . . . . . . . . . . . .9 3. Range of Health Services Provided at Facility by Ownership Category and Region . . . . . . . .10 4. Management of Health Facilities by Ownership Category and Region . . . . . . . . . . . . . . . . .12 5. The "Competitive Environment" in Health Care by Ownership Category and Region . . . . .13 6. Number of Staff by Ownership Category and Region . . . . . . . . . . . . . . . . . . . . . . . . . . . . .14 7. Health Facility Staff by Position . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .14 8. Staff Mix by Size of Facility, Ownership Category, and Region . . . . . . . . . . . . . . . . . . . . . .15 9. Median Salaries in Government Facilities by Position and Source of Financing . . . . . . . . . . .17 10. Median Salaries and Sources of Financing for Government Facilities by Region . . . . . . . . .17 11. Median Salaries in Private Facilities by Region . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .18 12. Average Daily Supply of Selected Drugs to Facilities by Ownership Category and Region . .19 13. Probit Analysis of Prescription of Antibiotics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .21 14. Average Monthly Supply of Vaccines to Government and Nonprofit Facilities . . . . . . . . . .22 15. Vaccination Stock-Outs and Resupply, Government and Nonprofit Facilities . . . . . . . . . . .23 16. Availability of Equipment by Ownership Category . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .24 17. User Fees by Ownership Category . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .25 18. Proportion of Facilities That Exempt Specific Patient Groups by Ownership Category . . . .26 19. Payment for Services by Ownership Category and Region . . . . . . . . . . . . . . . . . . . . . . . . .26 20. Amount Paid for Services, in Ugandan Shillings, by Ownership Category . . . . . . . . . . . . . .27 21. Number of Outpatients and Deliveries by Ownership Category . . . . . . . . . . . . . . . . . . . . .29 22. Health Worker Productivity by Ownership Category and Region . . . . . . . . . . . . . . . . . . . .31 23. Unit Costs, Labor, by Ownership Category and Region . . . . . . . . . . . . . . . . . . . . . . . . . . .32 24. Clients' Reasons for Coming to the Facility by Ownership Category and Region . . . . . . . .33 25. Client Perceptions Regarding Services by Ownership Category and Region . . . . . . . . . . . .33 26. Clients' Main Reason for Choosing a Specific Facility by Ownership Category . . . . . . . . .34 27. Differences in Health Care Facilities across Ownership Categories . . . . . . . . . . . . . . . . . . .37 Figures 1. Reconciling District and Facility Staff Records . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .15 2. Average Remuneration by Ownership Category . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .19 3. Forms of Chloroquine Supply . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .19 4. Calculating Total Drug Use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .20 5. Amounts of Drugs Prescribed per Patient, All Facilities . . . . . . . . . . . . . . . . . . . . . . . . . . . .20 6. Prescription Patterns for Drugs by Ownership Category . . . . . . . . . . . . . . . . . . . . . . . . . . .21 7. Prescription Patterns for Drugs by Staffing and Ownership Category . . . . . . . . . . . . . . . . . .21 8. Vaccine "Wastage" (Polio and BCG Vaccines), Government and Nonprofit Facilities . . . . . .23 9. Proportion of Government Facilities Charging Fees, by Region . . . . . . . . . . . . . . . . . . . . . .25 10. Spending of User Fee Revenues by Ownership Category . . . . . . . . . . . . . . . . . . . . . . . . . .27 11. Composition of Outpatients by Ownership Category . . . . . . . . . . . . . . . . . . . . . . . . . . . .29 12. Number of Vaccinations per Month by Ownership Category . . . . . . . . . . . . . . . . . . . . . . .30 B.1. Comparing District and Facility Output Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .45 iv Foreword H ealth care is at the center of many To help fill this gap, the Development poverty reduction strategies. Yet too Research Group of the World Bank is carrying often, health care services fail poor out, in collaboration with local institutions people. Budgetary allocations tend to and the Bank's Africa Region, a multicountry favor the better-off, limiting poor people's study of health care provision in Africa. The access to services or preventing improvements research covers Chad, Madagascar, Mozam- in quality. Even when funding for primary bique, and Nigeria, as well as Uganda, the sub- health care is allocated in the budget, it may be ject of this paper. The purpose of the research captured by the politically and economically is to compare and contrast the behavior of powerful. And the ability of medical staff to frontline providers in different institutional offer good care may have suffered severe blows and organizational contexts. The study pilots a as a result of persistent economic hardship or new instrument, the Quantitative Service political conflict. Delivery Survey (QSDS), in which the basic In many poor countries improvements in health facility is the primary unit of observa- health care thus call for institutional--not tion. Beyond its use in analyzing provider merely managerial--reforms. Such reforms behavior and service delivery, the QSDS fits include bottom-up measures to give users a well into the larger goal of impact evaluation. stronger voice and more power over providers. When combined with household surveys, it They also include top-down measures to allows exploration of interactions between ensure better monitoring of providers and frontline providers and users, and by adding introduce effective incentives for improving surveys of local politicians and officials, it can staff performance. Both types of reform also shed light on the political economy of depend on a body of systematic information on service delivery and on interactions between performance, incentives, and other aspects of providers and policymakers. frontline service delivery. This information is In the 1990s Uganda succeeded in reversing indispensable for catalyzing and guiding the the deterioration of the health infrastructure institutional reforms needed to improve health that had occurred during the economic and care and health outcomes--yet little of this political turmoil of the previous two decades. essential data are currently available. Most health indicators, except for life v expectancy related to the HIV/AIDS pandemic, health care that are being undertaken in many have been improving since recovery began in developing countries--including Uganda-- 1986. Infant mortality, although still high, fell with the goal of improving services. They are from 119 to 81 deaths per 1,000 live births also relevant for civil service reform efforts, between 1989 and 1995. But results from the which need to focus on strengthening profes- 2000/01 Uganda Demographic and Health sionalism and to take account of the entire Survey suggest that these improvements have health labor market when making decisions on not been sustained, despite the country's major public sector remuneration. My hope is that by successes in economic growth, poverty reduc- offering a new perspective from the frontlines tion, and education. Part of the problem has to of health care, this paper will make a useful do with household behavior and characteristics contribution to the reform agenda for improv- of individuals, but the poor quality of health ing health services for poor people in Uganda services also plays an important role. and across Africa. The study reported in this paper sheds new light on various dimensions of primary health Ok Pannenborg care delivery in Uganda, using a baseline sur- Senior Advisor for Health, vey of public and private dispensaries--the Nutrition, and Population most common lower-level health facilities in Human Development Department the country. The findings are highly relevant Africa Region for the emerging private-public partnerships in vi Abstract T his report presents findings from a base- The evidence suggests a close link between line survey of 155 primary health care the three types of providers through the labor facilities (dispensaries, with and without market for health workers. Government dis- maternity units) that was carried out in pensaries, for example, pay higher salaries Uganda in the latter part of 2000. By collecting than private facilities, and for-profit facilities data both from the dispensaries and from local appear to pay more than nonprofits for quali- governments, it was possible to validate the col- fied health staff. These salary differences affect lected data and check for discrepancies in the movement of staff between provider organ- reporting. Data from client exit polls provided izations. Several other dimensions of service a qualitative measure of performance. The delivery--mix of services, pricing, quality, use analysis compares service delivery performance of drugs, and cost-efficiency--also vary among in three ownership categories: government, pri- ownership categories. The findings are highly vate for-profit, and private nonprofit. Among relevant for public policy in Uganda and in the topics it explores at the facility level are other countries in Africa that are undertaking staffing, availability of drugs and other inputs, civil service reform and promoting private- remuneration, outputs, and financing. public partnerships in health care. vii Acknowledgments T he authors thank the following for assis- Useful comments were received from the Ugan- tance in the design and implementation dan Ministry of Health and from Shiyan Chao, of the Ugandan health facility survey: Ros Cooper, Lionel Demery, Satu Kähkönen, Delius Asiimwe and the team from the Mary Mulusa, and Peter Okwero. Financial Makerere Institute for Social Research; the support from the Japanese Policy and Human staffs of the Ugandan Ministries of Health and Resources Development Fund (PHRD) is grate- of Finance, Planning and Economic Develop- fully acknowledged. ment; and Charles Byaruhanga and Jan Dehn. viii Abbreviations and Acronyms BCG Bacillus Calmette-Guérin (vaccine for tuberculosis) DDHS district director of health services DHFDS District Health Facility Data Sheet DHTQ District Health Team Questionnaire DPT diphtheria, pertussis, and tetanus (vaccine) GDP gross domestic product FDS Facility Data Sheet HFQ Health Facility Questionnaire HUMC health unit management committee NGO nongovernmental organization ORS oral rehydration salts QSDS Quantitative Service Delivery Survey ix CHAPTER 1 Introduction T his Working Paper on health care in gross domestic product (GDP) tends to be a Uganda reports on findings from a small and statistically insignificant determi- baseline survey, conducted in 2000, of nant. Several studies have argued that the neg- dispensaries--the most common lower- ligible effect of social sector spending on level health facilities in the country--and of human development outcomes is likely to district administrations. The study represents reflect the weak link between spending and an effort to gain a better understanding of public services. More spending does not neces- incentives, motivations, and various aspects of sarily imply more public services (Pritchett service delivery at the level of frontline 1996; Reinikka 2001; Reinikka and Svensson providers. This is particularly important in 2001). Four general explanations of why this Uganda, where steady improvements in budg- should be so have been advanced. First, as etary management and shifts in the composi- numerous incidence studies show, government tion of public spending in favor of health over spending tends to favor nonpoor people and the past decade have not been matched by cor- private goods. Second, transfers of funds from responding improvements in the quality of the center to the frontline provider may suffer health services or in health outcomes (Hutchin- from leakage. Third, even if funds reach the son 2001). frontline provider, the production of goods and More generally, the survey was motivated by services at that level may be low in efficacy findings concerning the link between public because of incentive problems, absenteeism, or spending and health outcomes. Cross-country poor motivation of staff. Finally, even if servic- evidence suggests that, on average, total public es are delivered, household demand may be spending on health has had much less impact lacking (Devarajan and Reinikka 2002). on health status than one might expect. (For a All this suggests that cross-country budget review, see Filmer, Hammer, and Pritchett data are not sufficient for the analysis of serv- 2000, 2002; Musgrove 1996.)1 Socioeconomic ice delivery. Microlevel tools are needed to characteristics, including income and female understand the process by which public spend- education, explain most cross-country varia- ing is translated into services. This paper tion in child mortality and life expectancy. applies a new survey tool, the Quantitative Ser- Public expenditure on health as a share of vice Delivery Survey (QSDS), to frontline 1 2 Health Care on the Frontlines: Survey Evidence on Public and Private Providers in Uganda health service delivery in Uganda.2 In the fees, and the divergence between the amounts QSDS the facility or frontline service provider of drugs supplied to the facility and those actu- is typically the main unit of analysis, in much ally prescribed). the same way as the firm is the unit of obser- Section 2, which follows, provides a brief vation in enterprise surveys and the household overview of health care in Uganda. The subse- in household surveys.3 In the Uganda study, quent sections give a full description of the quantitative data were collected both through data collected and report on key diagnostic staff interviews and directly from the service findings from the first round of analysis. Sec- providers' records. Three types of dispensaries tion 3 describes the survey of dispensaries and were included: those run by the government, the sample. Section 4 presents key findings, by private for-profit providers, and by private covering, among other topics, oversight and nonprofit providers (mainly religious). Facility management, inputs and costs, user fees and data were triangulated by also surveying local financing, outputs, and client perceptions. (district) governments that oversee the health Finally, section 5 discusses key issues for poli- care providers.4 In this way it was possible to cymaking and research. More detailed descrip- verify and cross-check the information of inter- tions of the sample, the methodology, and data est (for example, reported patient numbers, issues are contained in the appendixes. CHAPTER 2 Background: Health Care in Uganda S ince 1986, the health sector in Uganda try began its recovery in the mid-1980s, added has been undergoing a process of another burden to the health system. Overall rebuilding and renovating health HIV prevalence is now in the 6 to 7 percent infrastructure. Most health indicators, range, down from 9 to 12 percent in the early except those related to HIV/AIDS, have been 1990s. (For a discussion of health problems in improving. For example, between 1989 and Uganda, see Hutchinson 2001.) 1995, infant mortality decreased from 119 Despite measurable improvements in some deaths per 1,000 live births to a (still high) 81 areas, the available health services are still deaths. But progress has not been completely inadequate to meet the needs of the popula- smooth. Results from the 2000/01 Uganda tion. The capital investments of the past Demographic and Health Survey show that decade, which have increased the population's by 2000, infant mortality had risen to 88 proximity to health facilities, have not been deaths per 1,000 live births, while under-five matched by improvements in quality. Partly for mortality rates increased from 147 deaths per this reason, use of the public sector for curative 1,000 live births in 1995 to 152 in 2000. care has remained remarkably constant since Although the differences in mortality rates are the late 1980s. According to evidence from not statistically significant, the failure to sus- household surveys, the poor and the nonpoor tain improvements is worrying and can be at alike tend to prefer curative care from non- least partly attributed to poor access to assis- governmental organizations (NGOs) and pri- tance during delivery and to declining immu- vate for-profit providers to the less expensive nization rates (Möller 2002; Republic of government care (Hutchinson 2001). Many Uganda 2002). government health units are faced with a situ- Health spending now represents 7 percent of ation of unused physical capacity, lack of total public expenditure, one of the highest trained staff, and supply shortages (Okello and shares in Africa (Hay 1998; Republic of Ugan- others 1998). Until now, there has been little da 2000), and is currently on the increase. quantitative and representative information on Even so, the health sector has been facing the scope and nature of problems in govern- many obstacles. The AIDS epidemic, which ment facilities or on the differences in per- emerged in Uganda at about the time the coun- formance across ownership categories. 3 4 Health Care on the Frontlines: Survey Evidence on Public and Private Providers in Uganda The government's ongoing health sector the decline in immunization rates between reform has attempted to address these weak- 1995 and 2000 can be at least partly attributed nesses. Since 1993, a process of decentraliza- to the unintended consequences of decentral- tion of responsibility for provision of health ization. In particular, it has proved difficult to services from the central Ministry of Health to incorporate formerly vertical programs into a local governments has been going on. The decentralized system, particularly where local impact of decentralization on health service priorities differ from national ones.5 The lack delivery and health outcomes is not yet clear. of central control has also made it difficult to Although decentralization has reportedly coordinate the response to the withdrawal of increased public participation in the health sec- support for outreach activities by the United tor, new problems have arisen. For example, Nations Children's Fund (UNICEF). CHAPTER 3 The Health Facility Survey A first effort to survey frontline health Purpose of the Survey facilities in Uganda was carried out in 1996, but at the time it was not possi- The survey differs from previous studies in that ble to obtain systematic quantitative it (a) analyses health service delivery from a data on inputs and outputs due to the domi- public expenditure perspective with a view to nance of in-kind transfers and lack of records informing expenditure and budget decisionmak- (Reinikka 2001). A rapid data assessment car- ing, as well as sector policy; (b) collects ried out in 1999, however, indicated that daily microlevel data on service provision, thus mak- patient, user fee, and drug use data could now ing a contribution toward redressing the lack to be compiled from most Ugandan health units, date of systematic examination of incentives and facilitating a detailed analysis of service deliv- other supply-side issues in frontline health care ery performance in primary health care (World delivery in low-income countries; and (c) focus- Bank 1999). In response to these changing cir- es on quantitative information--unlike some cumstances, a QSDS-type sample survey of dis- other performance evaluation techniques, which pensaries with and without maternity units are typically surveys of perceptions. was carried out during October­December In part, the survey was designed to provide 2000. The focus on dispensaries meant that the baseline data for future evaluation of reforms survey captured only part of the national and policies in the health sector and in public ex- health system, but the limited scope also per- penditure. More immediate objectives included: mitted a larger sample, a more in-depth analy- sis, and the inclusion of private health care · Measuring and explaining the variation in providers. This approach was motivated by the cost-efficiency across health units in importance of primary health care for poverty Uganda, with a focus on the flow and use reduction and the prominent role of the private of resources at the facility level; sector in the health care market.6 · Diagnosing problems with facility per- formance, including the extent of drug leakage, as well as staff performance and availability; 5 6 Health Care on the Frontlines: Survey Evidence on Public and Private Providers in Uganda · Providing information on pricing and The survey collected data at three levels: dis- user fee policies and assessing the types of trict administration, health facility, and client. service actually provided In this way it was possible to capture central · Shedding light on the quality of service elements of the relationships between the across the three categories of service provider organization, the frontline facility, provider--government, for-profit, and and the user. In addition, comparison of data nonprofit from different levels (triangulation) permitted · Examining the patterns of remuneration, cross-validation of information. pay structure, and oversight and monitor- At the district level, a District Health Team ing and their effects on health unit per- Questionnaire (DHTQ) was administered to formance the district director of health services (DDHS), · Assessing the private-public partnership, who was interviewed on the role of the DDHS particularly the program of financial aid office in health service delivery. Specifically, the to nonprofits. questionnaire collected data on health infra- structure, staff training, support and supervi- sion arrangements, and sources of financing. Survey Design and Implementation The District Health Facility Data Sheet (DHFDS) was used at the district level to col- The survey was designed and implemented by lect more detailed information on the sampled the World Bank in collaboration with the Mak- health units for fiscal 1999/2000, including erere Institute for Social Research and the data on staffing and the related salary struc- Ugandan Ministries of Health and of Finance, tures, vaccine supplies and immunization activ- Planning and Economic Development.7 The ity, and basic and supplementary supplies of sample design was governed by three princi- drugs to the facilities. In addition, patient data, ples. First, to ensure a degree of homogeneity including monthly returns from facilities on across sampled facilities, attention was restrict- total numbers of outpatients, inpatients, ed to dispensaries, with and without maternity immunizations, and deliveries, were reviewed units (that is, to the health center III level). Sec- for the period April­June 2000. ond, subject to security constraints, the sample At the facility level, the Health Facility was intended to capture regional differences. Questionnaire (HFQ) collected a broad range Finally, the sample had to include facilities in of information relating to the facility and its the main ownership categories: government, activities. The questionnaire, which was private for-profit, and private nonprofit (reli- administered to the in-charge, covered gious organizations and NGOs). The sample of (a) characteristics of the facility (location, type, government and nonprofit facilities was based level, ownership, catchment area, organiza- on the Ministry of Health facility register for tion, and services); (b) inputs (staff, drugs, vac- 1999. Since no nationwide census of for-profit cines, medical and nonmedical consumables, facilities was available, these facilities were and capital inputs); (c) outputs (facility utiliza- chosen by asking sampled government facilities tion and referrals); (d) financing (user charges, to identify the closest private dispensary (see cost of services by category, expenditures, and appendix A). Of the 155 health facilities sur- financial and in-kind support); and (e) institu- veyed, 81 were government facilities, 30 were tional support (supervision, reporting, per- private for-profit facilities, and 44 were non- formance assessment, and procurement). Each profit facilities. An exit poll of clients covered Health Facility Questionnaire was supplement- 1,617 individuals. The fieldwork was carried ed by a Facility Data Sheet (FDS). The FDS was out during October­December 2000. designed to obtain data from the health unit The Health Facility Survey 7 records on staffing and the related salary struc- an Exit Poll was used to interview about 10 ture; daily patient records for fiscal patients per facility on the cost of treatment, 1999/2000; the type of patients using the facil- drugs received, perceived quality of services, ity; vaccinations offered; and drug supply and and reasons for using that unit instead of alter- use at the facility. Finally, at the facility level, native sources of health care. CHAPTER 4 Public and Private Health Care Providers T he final sample consisted of 155 primary maternity care. The facilities vary considerably health care facilities drawn from 10 dis- in size, from units run by a single individual to tricts in the central, eastern, northern, facilities with as many as 19 staff members. and western regions of the country. It Table 2 also sets out some descriptive statis- included government, private for-profit, and tics relating to the facility's infrastructure and private nonprofit facilities. The nonprofit sec- the distance to services and administrative cen- tor includes facilities owned and operated by ters. These variables provide a rough indica- religious organizations and NGOs. Table 1 tion of the structural dimensions of quality and shows the distribution of the sample across the geographic location of the health unit. ownership categories and regions. Table 2 presents general characteristics of INFRASTRUCTURE AND LOCATION. For a large the facilities in the sample. Approximately one- proportion of facilities, boreholes are the pri- third of the surveyed facilities are dispensaries mary source of water, but other water sources, without maternity units; the rest provide including piped water, springs, and collected Table 1. Number of Health Care Facilities by Ownership Category and Region Region Ownership Central Eastern Northern Western Total Government 30 24 12 15 81 Private for-profit 10 9 4 7 30 Private nonprofit 17 12 7 8 44 Catholic Medical Services 8 3 6 8 25 Protestant Medical Bureau 5 5 1 0 11 Muslim Medical Bureau 1 0 0 0 1 Seventh-Day Adventist 2 0 0 0 2 Nongovernmental organizations 1 4 0 0 5 Total 57 45 23 30 155 Source: Health facility survey, 1999/2000. 8 Public and Private Health Care Providers 9 Table 2. Characteristics of Health Facilities by Ownership Category and Region Ownership Region Private Private Government for-profit nonprofit Central Eastern Northern Western Total Type of facility (percentage of total) Dispensary 30.4 38.7 43.2 22.8 17.8 45.5 80.0 35.7 Dispensary with maternity unit 69.6 61.3 56.8 77.2 82.2 54.6 20.0 64.3 Source of water (percentage of total) Piped water 7.6 25.8 15.9 12.3 13.3 0.0 27.6 13.6 Borehole 45.6 29.0 47.7 29.8 60.0 78.3 13.8 42.9 Protected spring 13.9 22.6 9.1 14.0 20.0 8.7 10.3 14.3 Unprotected spring 17.7 9.7 4.6 8.8 6.7 4.4 34.5 12.3 Harvested rainwater 12.7 9.7 22.7 33.3 0.0 0.0 13.8 14.9 Purchased water 1.3 0.0 0.0 1.8 0.0 0.0 0.0 0.7 Other 1.3 3.2 0.0 0.0 0.0 8.7 0.0 1.3 Waste disposal (percentage of total) Public waste disposal 0.0 0.0 2.3 1.8 0.0 0.0 0.0 0.7 Pit (dumping) 41.3 30.0 36.4 37.5 37.8 60.9 20.0 37.7 Pit (burning) 51.3 66.7 54.6 57.1 57.8 39.1 60.0 55.2 Incineration 0.0 0.0 2.3 1.8 0.0 0.0 0.0 0.7 Other 7.5 3.3 4.6 1.8 4.4 0.0 20.0 5.8 Distance to services and institutions (kilometers) To telephone Less than 5 22.5 54.8 47.7 57.9 28.9 21.7 16.7 36.1 6­20 36.3 29.0 25.0 29.8 37.8 17.4 36.7 31.6 21­50 32.5 12.9 20.5 8.8 22.2 52.2 40.0 25.2 51­100 8.8 3.2 6.8 3.5 11.1 8.7 6.7 7.1 To district headquarters Less than 5 3.8 9.7 11.6 7.1 4.4 13.0 6.7 7.1 6­20 25.0 38.7 37.2 33.9 35.6 17.4 30.0 31.2 21­50 51.3 45.2 41.9 48.2 44.4 56.5 43.3 47.4 51­100 20.0 6.5 7.0 10.7 15.6 8.7 20.0 13.6 More than 100 0.0 0.0 2.3 0.0 0.0 4.4 0.0 0.7 To health subdistrict headquarters Less than 5 15.2 42.9 32.6 20.8 25.0 34.8 26.7 25.3 6­20 62.0 50.0 48.8 66.0 56.8 34.8 53.3 56.0 21­50 22.8 7.1 18.6 13.2 18.2 30.4 20.0 18.7 To subcounty center Less than 5 73.8 80.7 65.1 78.6 77.8 65.2 60.0 72.7 6­20 26.3 19.4 34.9 21.4 22.2 34.8 40.0 27.3 To village center Less than 5 98.7 100.0 97.7 98.2 97.8 100.0 100.0 98.7 6­20 0.0 0.0 2.3 1.8 0.0 0.0 0.0 0.7 21­50 1.3 0.0 0.0 0.0 2.2 0.0 0.0 0.7 Source: Health facility survey, 1999/2000 (Health Facility Questionnaire). 10 Health Care on the Frontlines: Survey Evidence on Public and Private Providers in Uganda rainwater, are also important. Piped water is portion of for-profit facilities providing mental more prevalent in private facilities. There are health care, dental care, immunizations, and also noteworthy regional differences. For training of health workers is smaller than for example, facilities in the northern region have the other two types. Similarly, only 60 percent no access to piped water and rely more on of the nonprofit facilities provide family plan- dumping for waste disposal. With respect to ning services, compared with over 90 percent location, the data suggest that government for government and private for-profit facilities. facilities are more remote and isolated than This difference most likely reflects, at least in nonprofit facilities.8 On average, government part, the religious affiliation of some facilities. facilities have more limited access to tele- Laboratory services are an important excep- phones and are farther away from district and tion to the general pattern of broader service health subdistrict headquarters. provision in government facilities: only 16 per- cent of government facilities offer these servic- SERVICES. Although the sample is restricted to es, compared with over 50 percent for private primary-level health care facilities, a broad for-profit and nonprofit providers. This clearly range of services is represented. There are note- has implications for the diagnostic capabilities worthy variations between facilities and across of the facility and hence for the quality of care ownership categories (see table 3). In general, provided. government and nonprofit facilities are more likely than private for-profit facilities to offer a broad range of services. For example, the pro- Table 3. Range of Health Services Provided at Facility by Ownership Category and Region (percentage of facilities) Ownership Region Private Private Service Government for-profit nonprofit Central Eastern Northern Western Total Outreach 96.2 16.1 84.1 82.1 80.0 73.9 63.3 76.6 Outpatient care 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Inpatient care 53.2 63.3 79.6 52.6 77.3 56.5 65.5 62.8 Medical care 90.0 93.6 97.7 100.0 91.1 87.0 86.7 92.9 Eye care 50.0 13.3 40.9 52.6 28.9 60.9 17.2 40.3 Mental health care 20.5 3.3 11.6 15.8 4.6 13.0 29.6 14.6 Dental health 25.0 13.3 18.2 29.8 8.9 39.1 6.9 20.8 Minor surgery 90.0 64.5 84.1 80.7 80.0 73.9 100.0 83.2 Deliveries 85.0 61.3 70.5 89.5 84.4 60.9 50.0 76.1 Laboratory services 16.3 51.6 53.5 36.8 13.3 45.5 50.0 33.8 Preventive care Health education 100.0 77.4 97.7 96.5 93.3 100.0 89.7 94.8 Immunizations 97.5 31.0 81.8 91.2 77.3 65.2 75.0 80.3 Antenatal care 98.7 73.3 88.6 98.3 91.1 73.9 89.3 90.9 Family planning 96.3 90.0 60.5 89.5 86.4 82.6 75.9 85.0 Training Nursing aides 32.9 20.0 34.1 8.9 47.7 60.9 23.3 30.7 Community health workers 21.9 3.2 32.6 17.0 31.8 30.4 3.7 21.1 Source: Health facility survey, 1999/2000 (Health Facility Questionnaire). Public and Private Health Care Providers 11 Oversight, Management, The districts do not exercise functions of and Competition supervision, staff assessment, and audit consis- tently across all facilities. Indeed, the HFQ pro- Both government and private health facilities vides evidence that, from the facility perspective, operate within an institutional context that monitoring and oversight are patchy. According affords at least some degree of oversight, to the responses by dispensary in-charges, 70 supervision, and regulation and that, in the percent of government and private nonprofit case of government and private nonprofit facilities are supervised regularly (monthly or facilities, provides resources for the operation "at any time") by the district, subdistrict, or of the facilities. Supervision and other moni- subcounty. A much lower proportion of for- toring activities are clearly an important profit facilities is regularly supervised by the dis- means by which local and central governments trict administration. In half of the government can seek to improve the performance of front- facilities, staff are assessed frequently (monthly, line health providers. Data on various aspects weekly, daily, or "at any time"). The percentage of external incentives of this nature were col- is much lower in private for-profit and nonprof- lected through the District Health Team Ques- it facilities (10 and 38 percent, respectively). tionnaire (DHTQ) and the Health Facility Half of the government facilities report that Questionnaire (HFQ). audits are regularly carried out, whereas the proportions for for-profit and nonprofit facili- Oversight by Local Government ties are 17 and 52 percent, respectively. The data from the DHTQ indicate that the dis- Management of Facilities trict plays an active role in supporting and super- vising health facilities. Indeed, in all the districts Most government and private nonprofit facili- covered by the survey, respondents report that ties have a health unit management committee staff from the district, the health subdistrict, and (HUMC) or a governing board that is typical- the central Ministry of Health provide support ly made up of seven to eight members drawn and supervise government facilities. In addition, from the facility, the community, and the sub- most district authorities reportedly have a sup- county administration (see table 4). In most portive and supervisory role with respect to pri- cases the HUMC meets three to four times a vate for-profit and nonprofit providers.9 Accord- year. The main topics addressed include remu- ing to the respondents, supervision covers service neration and other staff issues, the physical quality, management, and record-keeping. Other condition of the facility, the utilization of user institutional mechanisms exist whereby the dis- charges levied by the facility, and drug supply. trict administration seeks to hold the facility and its staff accountable. The most important of Competition these are staff assessments and audits. In all dis- tricts, according to the reports, the performance In general, health facilities do not operate in of staff in government facilities is formally isolation but, rather, in a complex health care assessed, sometimes leading to the promotion, provision market. Indeed, the data provide evi- demotion, or firing of staff. In four districts, staff dence that facilities are exposed to consider- members in private nonprofit and for-profit able competition and that they are aware of the facilities are also reportedly assessed. In 8 of the other providers operating in their catchment 10 surveyed districts, the district reports that it area. Table 5 shows that the degree of compe- regularly audits inputs, incomes, and expendi- tition--proxied by the number of other tures for all facilities. providers in the catchment area and the dis- 12 Health Care on the Frontlines: Survey Evidence on Public and Private Providers in Uganda Table 4. Management of Health Facilities by Ownership Category and Region Ownershipa Region Private Government nonprofit Central Eastern Northern Western Total Number of facilities in subsample 81 44 47 36 19 23 125 Facility has HUMC (percent) 97.5 100.0 100.0 100.0 94.7 95.7 98.4 Number of HUMC members Mean 7.5 8.7 8.0 8.4 8.2 7.3 8.0 Median 7.0 8.0 7.0 7.0 7.0 7.0 7.0 Number of HUMC meetings per year Mean 3.9 4.0 4.6 4.1 3.3 2.8 3.9 Median 3.0 3.0 4.0 3.0 2.0 2.0 3.0 Issues addressed by HUMC (percent) Drug supply 58.4 58.0 58.7 69.7 38.9 54.6 58.0 Allowances/remuneration 24.7 30.8 37.0 39.4 27.8 8.7 30.8 Transport 11.7 14.2 19.6 12.1 11.1 8.7 14.2 Staff issues 80.5 78.3 71.7 87.9 94.4 65.2 78.3 Physical condition of facility 74.0 72.5 80.4 81.8 61.1 52.2 72.5 Relations with district 10.4 10.8 6.5 21.2 11.1 4.4 10.8 Mobilization of donor and other support 19.5 25.8 28.3 15.2 38.9 26.1 25.8 Utilization of user charges 72.7 65.8 58.7 81.8 55.6 65.2 65.8 Representation on HUMC/board (percent) In-charge 100.0 90.7 93.6 100.0 100.0 95.5 96.7 Other facility staff 96.2 65.9 85.1 88.6 88.9 78.3 85.4 District officials 2.6 11.4 6.4 2.9 5.6 9.1 5.7 District politicians 2.5 11.4 6.4 5.7 5.6 4.4 5.7 Health subdistrict officials 2.5 9.1 6.4 2.9 11.1 0.0 4.9 County officials 1.3 2.3 2.1 0.0 5.6 0.0 1.6 Subcounty official 48.1 20.5 38.3 28.6 27.8 60.9 38.2 Parish official 42.3 34.1 51.1 23.5 55.6 26.1 39.3 Village official 98.7 59.1 76.6 94.3 88.9 82.6 84.6 Community representative 87.2 75.0 73.9 88.6 94.4 82.6 82.8 Religious leaders 59.0 93.0 73.3 65.7 44.4 95.7 71.1 Teacher representative 71.8 61.4 50.0 68.6 94.4 82.6 68.0 Mode of selection of HUMC/ board members (percent) Appointment by district 3.9 2.4 2.3 2.9 11.1 0.0 3.4 Appointment by subcounty 84.6 0.0 61.4 48.6 44.4 63.6 55.5 Appointment by village 10.3 9.8 9.1 2.9 33.3 4.6 10.1 Locally elected 23.1 29.3 2.3 51.4 61.1 0.0 25.2 Volunteer 2.6 14.6 2.3 2.9 22.2 9.1 6.7 By virtue of employment 71.8 61.9 60.0 51.4 94.4 90.9 68.3 Note: HUMC, health unit management committee. a. According to the data, only two private for-profit facilities have governing boards. Private for-profit facilities are therefore not considered in this table. Source: Health facility survey, 1999/2000 (Health Facility Questionnaire). Public and Private Health Care Providers 13 tance to the closest provider--varies substan- ture of the link between public spending in the tially across regions and ownership categories. health sector and the availability of inputs at the facility level. This survey collected extensive information Inputs and Costs at Health Facilities on key inputs in the production of health serv- ices at the facility level. The information was Efficient delivery of primary health care servic- collected at both district and facility levels. es depends on the availability of capital and The existence of comparable data from differ- recurrent inputs at the facility level. The range ent sources made it possible to assess the of capital inputs depends on the type of servic- validity of the collected data and the opera- es provided but typically includes buildings, tions of the internal management information vehicles, refrigerators, sterilizers, and other systems. This section discusses the evidence equipment. Important recurrent inputs include from these data, focusing primarily on personnel; supplies (food, drugs, vaccines, staffing, drugs, and vaccines, but with some syringes, and so on); and requirements for attention to other inputs. operation and maintenance of vehicles (fuel, lubricants, spare parts, and insurance) and Staffing buildings (electricity, water, fuel, telephone, cleaning, and repairs). Data on facility staffing were collected prima- In the case of government facilities, the rily through the Facility Data Sheet (FDS) and inputs into the production of health services the District Health Facility Data Sheet are not generally procured directly by the facil- (DHFDS). Some additional information con- ity. Rather, financial resources are allocated to cerning training and staff assessments was col- different cost centers upstream, which are lected through the complementary instru- responsible for the procurement, payment, and ments. In general, one would expect the most distribution of the inputs. The complex institu- detailed information about the number of tional structure that governs budget execution frontline staff, their positions, and their remu- makes it difficult to get a comprehensive pic- neration to be available at the facility level.10 Table 5. The "Competitive Environment" in Health Care by Ownership Category and Region Mean number of other providers in catchment area Distance to closest Provider type Aa Provider type Bb other provider (kilometers)c Ownership Government 7.8 3.6 2.3 Private for-profit 10.9 5.7 0.8 Private nonprofit 11.1 5.4 1.4 Region Central 11.5 5.3 1.2 Eastern 9.4 4.9 1.1 Northern 5.2 2.7 2.4 Western 8.5 3.9 3.3 All regions 9.4 4.5 1.7 a. Includes aide post/subdispensary, dispensary, health center/hospital, clinic, and drug shop/pharmacy. b. As in note a, but excludes drug shop/pharmacy. c. Excludes drug shop/pharmacy. Source: Health facility survey, 1999/2000 (Health Facility Questionnaire). 14 Health Care on the Frontlines: Survey Evidence on Public and Private Providers in Uganda At that level, the enumerators were also able to nursing aides and "other" staff make up more verify the presence of the staff members. For than half of facility staff (see table 7).12 these reasons, the staff data from the FDS form According to government policy, a health the primary basis for the analysis. Data col- center III is not expected to have a medical lected from facilities are also compared with doctor, but it is supposed to be managed by a those obtained from the district headquarters. clinical officer. Yet, as table 8 shows, many In total, data were collected on 1,087 health facilities operate without either a doctor or a facility staff.11 clinical officer. Almost 10 percent of the facili- ties in the sample have only nonmedical staff. STAFF NUMBERS AND STAFF MIX. The number of Not surprisingly, staff composition is closely staff varies considerably among the sampled related to the size of the facility; facilities with facilities, ranging from 1 to 19. The sample more staff are also likely to have qualified staff mean and median are both 7. As can be seen (a doctor, nurse, or clinical officer). Govern- from table 6, government and private nonprof- ment facilities are less likely to operate with it facilities have more staff, on average, than only nonmedical staff or without a doctor or private for-profit facilities. Indeed, most pri- clinical officer. Substantial regional differences vate facilities tend to be small; only a small in staff composition can be seen--in particular, number of them have more than five staff the greater likelihood that facilities in the cen- members. It is more difficult to generalize tral region will have doctors or clinical officers. about staffing patterns for private nonprofit Almost all (92 percent) of the facility staff in facilities, where the distribution is considerably the sample reportedly work full time. The use flatter. The largest facilities in this group are of part-time workers is more common in pri- run by Catholic Medical Services. vate facilities. In government facilities the use The survey distinguishes nine categories of of part-time staff is largely restricted to inci- staff. On the basis of the nature of their work dental "other" staff; in private facilities doc- and their average salaries, these categories can tors, clinical officers, nurses, and lab assistants be consolidated into five broad groups: are often contracted part time. (a) medical doctor, (b) qualified nurse or clini- cal officer, (c) basic nurse, midwife, or lab assis- tant, (d) nursing aide, and (e) other. Together, Table 7. Health Facility Staff by Position Percentage Position Frequency of totala Table 6. Number of Staff by Ownership Category Medical doctor 20 1.8 and Region Qualified nurse or Minimum Mean Median Maximum clinical officer 95 8.7 Clinical officer 71 6.5 Ownership Comprehensive nurse 2 0.2 Government 2 7.8 7 19 Registered nurse 22 2.0 Private for-profit 1 4.2 3 12 Basic nurse/midwife 272 25.0 Private nonprofit 2 7.7 6 19 Enrolled nurse 122 11.2 Region Enrolled midwife 121 11.1 Central 1 8.0 7 18 Lab assistant 29 2.7 Eastern 1 5.8 5 14 Nursing aide 298 27.4 Northern 3 8.2 8 19 Other 402 37.0 Western 1 6.1 5 19 Total 1,087 100.0 All regions 1 7.0 7 19 a. Numbers may not sum to total because of rounding. Source: Health facility survey, 1999/2000 (Facility Data Sheet). Source: Health facility survey, 1999/2000 (Facility Data Sheet). Public and Private Health Care Providers 15 Table 8. Staff Mix by Size of Facility, Ownership Category, and Region No doctor or clinical officer Assistant or nonmedical Frequency (percent) staff only (percent) Facility size (number of staff) Small (1­5) 57 79 19 Medium (6­8) 56 46 5 Large (9­19) 42 21 0 Ownership Government 81 38 6 Private for-profit 30 58 16 Private nonprofit 44 73 9 Region Central 57 30 4 Eastern 45 73 13 Northern 23 52 9 Western 30 60 13 Total (all facilities) or average 155 52 9 Source: Health facility survey, 1999/2000 (Facility Data Sheet). COMPARISON OF FACILITY AND DISTRICT DATA ON working in the government facility subsample, STAFFING IN GOVERNMENT FACILITIES. To com- but only 356 of these (56 percent) are to be plement the staffing data collected at the facil- found in the district records.14 In other words, ity level, information on staff in the sampled there are large numbers of staff working in facilities was also collected at the district level. government facilities about whom the district The district administration does not, in gener- authorities appear to have no information. al, keep records of staffing in private for-profit These staff members can be considered "ghost or nonprofit facilities, but it does keep detailed workers" from the district's perspective. There records of staff in government facilities.13 In are also 109 individuals for which the reverse principle, district and facility data on staffing situation holds: they appear in the district should correspond. In reality, the situation is records but not in the corresponding facility far more complicated. The staffing data col- records. These are "ghost workers" in the lected at the facility level sum to 631 staff more traditional sense (see figure 1). Figure 1. Reconciling District and Facility Staff Records Staff outside the formal 275 system Ghost workers 109 Staff who appear Staff records in facility records not updated Staff who appear (631) in district records 356 Staff in facility records who 356 (465) also appear in district records Source: Health facility survey, 1999/2000. 16 Health Care on the Frontlines: Survey Evidence on Public and Private Providers in Uganda Of the 275 health workers who appear only able number of staff listed in district records in the facility records, 60 percent fall into the who do not appear to be working in the cor- category "other." Another 20 percent are nurs- responding facility. It is plausible, however, ing aides, and 15 percent are enrolled nurses or that these staff members have simply been midwives. It is not surprising that in most cases replaced by others and that the new staff are the salary of these staff members is low, with a now receiving the salary of the apparent median of only 40,000 Ugandan shillings "ghosts." Similarly, there are many more staff (USh).15 According to the information collect- working at the facility level than would ed at the facility level, the district finances 38 appear to be the case from the aggregated percent of these staff. The rest are financed by (district-level) payroll data. A considerable the facility (27 percent), the subcounty (26 per- proportion of these staff is paid from user fee cent), or another source (9 percent). revenues. These issues clearly merit further How can the lack of consistency between attention and will be addressed in future facility and district records be explained? As analysis of the survey data. noted above, many of the staff who appear only in the facility records are actually STAFF REMUNERATION AND FINANCING: GOVERN- financed by the district or subcounty. In fact, MENT FACILITIES. This section takes a closer in many cases the presence in a facility of a look at remuneration for different categories of "district-level ghost" (an individual who staff, how remuneration differs across facilities appears only in the district record) is matched according to ownership and geographic loca- by the existence of a "facility-level ghost" (an tion, and staff financing. In principle, the con- individual who is reportedly financed by the ditions of employment of staff in government district but appears only in facility records). facilities are clearly established in the form of a This suggests that part of the answer probably salary scale.16 In addition, in fiscal 1999/2000 lies in poorly updated district records: a staff staff in government facilities were entitled to a member may have been replaced, but the dis- lunch allowance. All established staff in hospi- trict-level records were not updated. Given the tals and lower-level units should have received discrepancies in numbers, however (109 staff USh 66,000 per month as a lunch allowance, appear only in district records, whereas 275 while most support staff should have received appear only in facility records), there seems to USh 44,000.17 Formally, these conditions be more to the story. Most likely, there is a apply only to staff in established positions large number of elementary health workers, (that is, on the payroll). financed by the subcounty, for whom the dis- As table 9 shows, the salary scale is not nec- trict does not keep records. There is also a essarily a good guide to salaries as reported by considerable group of staff financed either by the facilities, in particular for staff not financed the facility or by some other source (such as an by the central government or the district. Staff NGO or a donor), and these staff may not be financed by the subcounty receive less than recorded at the district level. All in all, it staff in equivalent positions financed by the appears that approximately 20­25 percent district. This is particularly true for staff in the more staff are working at the facility level heterogeneous "other" category. Staff financed than would be expected on the basis of district from facility revenues--lab assistants, nursing records. aides, and others--receive only a proportion of To sum up, it is difficult to establish from the salary that they would receive if they were the data whether there are many "ghost on the district payroll. workers" on the district payroll. At first Seventy-seven percent of the staff financed glance, it looks as though there is a consider- by the subcounty or the facility do not appear Public and Private Health Care Providers 17 Table 9. Median Salaries in Government Facilities by Position and Source of Financing (Ugandan shillings) Government Source of financing salary District Subcounty Facility Other Government for grade Position n Salary n Salary n Salary n Salary salary scale (midpoint) Medical doctor 2 125,489 -- -- -- -- -- -- U3­U5A 354,063 Clinical officer 48 143,881 -- -- -- -- 1 360,000 U6 126,688 Comprehensive nurse 1 79,776 -- -- -- -- -- -- Registered nurse 7 125,549 -- -- -- -- -- -- Enrolled nurse 62 113,255 -- -- -- -- 5 113,824 U7 105,605 Enrolled midwife 74 113,255 -- -- -- -- 6 143,412 Lab assistant 3 84,040 -- -- 3 35,000 1 113,824 Nursing aide 137 59,902 39 56,930 10 30,000 2 85,546 U8 82,601 Other 89 102,225 54 27,500 54 15,000 14 37,500 n.a. n.a. -- Not available. n.a. Not applicable. Note: n denotes number of staff. Source: Health facility survey, 1999/2000 (Facility Data Sheet). to receive any lunch allowance. By contrast, · Median salaries in the central region are staff financed from "other" sources appear to higher than in the other regions. have conditions broadly similar to staff · User fee revenues are most important as a financed by the district. Thus, differences in source of financing of staff salaries in the sources of financing can lead to inefficient and central and eastern regions. inequitable staffing patterns across facilities and to inequities within facilities. STAFF REMUNERATION AND FINANCING: PRIVATE Finally, two dimensions of regional variation FACILITIES. Staff in for-profit facilities appear to in government facilities are of particular inter- be exclusively financed by the facility, through est (see table 10): funds raised from user charges. The situation is Table 10. Median Salaries and Sources of Financing for Government Facilities by Region Central Eastern Northern Western Number of staff 269 145 104 104 Median salary (Ugandan shillings) Doctor/clinical officer 193,000 140,209 135,161 135,660 Nurse 170,000 108,251 113,824 113,000 Assistant/other 71,000 58,941 56,930 59,530 Source of financing (percentage of staff) District/government 75.6 68.1 53.9 68.3 Subcounty 8.0 14.6 33.7 15.4 Facility (user fees) 13.0 17.4 1.0 7.7 Other 3.4 0.0 11.5 8.7 Source: Health facility survey, 1999/2000 (Facility Data Sheet). 18 Health Care on the Frontlines: Survey Evidence on Public and Private Providers in Uganda similar in private nonprofit facilities, although the salary spectrum. If the lunch allowance is some of them benefit from financial support added, the difference is even more pronounced. (contributions or public subsidy) and certain But although salaries are higher in government staff members in nonprofit facilities receive facilities, there is also evidence that govern- their salaries from the district. ment staff are more likely to experience delays Table 11 suggests that there are regional dif- in payment of salaries. The facility in-charges ferences in salaries in both for-profit and non- were asked about the "average length of delays profit private facilities. For-profit facilities (in weeks) in staff salaries (excluding salaries show a bias in favor of the central region, in paid from user fees)." It is clear from the particular for more qualified staff. In nonprof- responses that delays in payment are a consid- it facilities there appears to be a slight bias in erable problem in some facilities. In only 28 favor of the northern and western regions. percent of government facilities are salaries Preliminary analysis indicates that staff in normally received on time, as against 72 per- for-profit and nonprofit facilities tend to be cent in for-profit and 39 percent in nonprofit paid less than staff in government facilities and facilities. In 20 percent of government facilities that there are also differences within the pri- the delay is reportedly more than 16 weeks. vate sector. Nonprofit facilities pay significant- ly less for staff at the highest qualification lev- STAFF ATTENDANCE. Finally, at the time of the els than for-profit facilities do. Finally, most visit by the survey enumerator, most facility staff in private facilities do not receive a lunch staff were reported to be present. Of those who allowance, and if they do, it tends to be lower were not present, the majority were on leave or than the allowance in government facilities.18 off duty. Unauthorized leave (absenteeism) in the sample appears to be low, at 3.1 percent; it SUMMARY OF PAY CONDITIONS. As is clear from is slightly higher (4.4 percent) in government figure 2, even without the lunch allowance, facilities. Although these findings are encour- staff in government facilities are, on average, aging, they should be interpreted with care. better paid than staff in private for-profit and First, it is possible that the respondent was cov- nonprofit facilities. This is true at all levels of ering for staff members who were absent with- Table 11. Median Salaries in Private Facilities by Region (Ugandan shillings) Ownership and staff position Central Eastern Northern Western Private for-profit Number of staff 53 23 23 22 Position Doctor/clinical officer 205,000 100,000 80,000 100,000 Nurse 70,000 127,500 50,000 70,000 Assistant/other 50,000 30,000 25,000 35,000 Private nonprofit Number of staff 133 83 62 57 Position Doctor/clinical officer 150,000 140,000 176,952 -- Nurse 85,000 113,000 70,000 100,000 Assistant/other 30,000 30,000 29,000 50,000 -- Not available. Source: Health facility survey, 1999/2000 (Facility Data Sheet). Public and Private Health Care Providers 19 Figure 2. Average Remuneration Table 12. Average Daily Supply of Selected Drugs to by Ownership Category Facilities by Ownership Category and Region (number of tablets) 200,000 Chloroquine Septrin 180,000 Provide nonprofit 160,000 Provide for-profit All facilities Government (excluding 140,000 Government 88.2 42.9 lunch allowance) Government (including 120,000 Private nonprofit 45.3 40.1 lunch allowance) shillings 100,000 Government facilities 80,000 by region Ugandan 60,000 Central 73.3 35.1 40,000 Eastern 72.3 54.8 20,000 Northern 119.4 44.3 Western 116.2 41.3 0 25th percentile Median 75th percentile Note: The average daily supply is calculated by taking total supply to the facility over an extended period (237 days, on average) and Source: Health facility survey, 1999/2000. dividing that total (excluding supply on the final day) by the number of days covered. Source: Health facility survey, 1999/2000 (Facility Data Sheet). out leave and that absenteeism was therefore its rely in large part on purchased drugs. These underestimated. Also, the fact that facility staff differences are mirrored in the source of sup- had prior warning of the survey visit casts ply; government facilities are almost exclusive- some doubt on the validity of the estimates. ly supplied by the district administration, whereas private nonprofit facilities rely largely Supply and Use of Drugs on both district supplies and private sources, including the joint medical store operated by DRUG SUPPLY TO FACILITIES. Detailed informa- the religious provider organizations. tion on the supply of drugs to facilities was col- lected at both district and facility levels. The DRUG USE AT THE FACILITY LEVEL. Detailed district data are based on stock cards from the information was collected at the facility level district medical store and cover both kit sup- on the use and distribution of drugs. Again, the plies and supplementary supplies of specific drugs. Data were collected on six drugs: chloroquine, Septrin (a combination of antibi- Figure 3. Forms of Chloroquine Supply otics), procaine penicillin fortified, Paraceta- 100 mol (acetaminophen), ergometrine, and oral rehydration salts (ORS). Corresponding data 80 were collected from the facility stock cards. Table 12, based on facility stock cards, totalof Other 60 Purchase shows the average daily supply of chloroquine Kit B and Septrin to government and nonprofit Kit A 40 facilities.19 Percentage Kit E The stock cards also provide information on 20 the form and source of supply of each drug. Taking chloroquine as an example, figure 3 0 Government Private Central Eastern Northern Western demonstrates that kit supply is the most impor- nonprofit Government facilities only tant form of supply for government facilities (except in the western region), while nonprof- Source: Health facility survey, 1999/2000. 20 Health Care on the Frontlines: Survey Evidence on Public and Private Providers in Uganda survey focused on chloroquine, Septrin, pro- Figure 5. Amounts of Drugs Prescribed caine penicillin fortified, Paracetamol, ergome- per Patient, All Facilities trine, and oral rehydration salts. The survey (number of tablets, adult equivalents) tools were designed to capture stock move- 40 Septrin, per patient ments for a period of approximately one month in early 2000 during which no new sup- plies were received and the facility experienced Chloroquine, average dose no stock-outs. Total removals of drugs from 30 stocks for such a period can be easily calculat- Paracetamol, per patient ed from stock levels at the beginning and end of the period. For each type of drug, the analy- Septrin, average dose 20 sis focuses on two main variables: the amount Paracetamol, average dose of drugs removed from stocks, and the number of patients. The amount of drugs removed from stocks in period t to t is calculated as 10 Total drugs removed = removed(t) + stock(t) ­ stock(t). 0 0 10 20 30 40 50 60 70 80 The logic of this calculation is illustrated in Chloroquine tablets per patient with malaria figure 4. Patient registers were reviewed for the same period, and the number of patients was Source: Health facility survey, 1999/2000. counted, distinguishing between adults and children and counting separately the number of patients diagnosed with malaria. trin to be prescribed for all patients, so the The simplest way of assessing how drug use average number of tablets per patient (aver- relates to patient numbers is to look at the aged over all patients, whether they receive a number of tablets given to patients.20 Figure 5 prescription or not) should be lower than the plots the numbers of Paracetamol and Septrin recommended dose. As expected, Paracetamol tablets used (removed from stock) per patient appears to be prescribed more liberally than against the number of chloroquine tablets used Septrin. Drug use per patient is very high in per patient with malaria for each facility. some facilities, and high use per patient of one Clearly, we do not expect Paracetamol or Sep- drug appears to be related to high use of other drugs. The reasons for high drug use per Figure 4. Calculating Total Drug Use patient may include high need, overprescrip- tion, and leakage.21 Stock of drugs EVIDENCE FROM THE EXIT POLL. On the basis of interviews with clients leaving the facility, it appears that most clients received some drugs following the consultation--in many cases, Removed (t) more than one type of drug (see figure 6). The Stock (t) Total drugs removed differences across ownership categories are Stock (t') surprisingly small. On average, more than 60 percent of patients receive aspirin or Paraceta- First removal New from stock supply mol, and 40 percent receive chloroquine. What Public and Private Health Care Providers 21 Figure 6. Prescription Patterns for Drugs Figure 7. Prescription Patterns for Drugs by by Ownership Category Staffing and Ownership Category Aspirin 70 Chloroquine 60 Septrin 50 Other antibiotics patients Nutritional supplements 40 of Mebendazole 30 ORS 20 Quinine Percentage Piriton 10 Fansidar 0 Other Doctor/clinical officer Nurse Unqualified 0 10 2 0 30 4 0 50 6 0 70 Government Private for-profit Private nonprofit Percentage of patients Government Private for-profit Private nonprofit Source: Health facility survey, 1999/2000. Note: ORS, oral rehydration salts. Source: Health facility survey, 1999/2000. government facilities in different regions. This issue will be explored further in future analysis is perhaps more surprising is the high propor- of the survey data. tion of clients who receive Septrin or another All but one of the sampled government facil- form of antibiotic. More than 46 percent of ities report carrying out immunizations, and so patients report receiving an antibiotic, and in 5 do 36 of the 44 private nonprofit facilities. By percent of the cases they receive both Septrin contrast, only 31 percent of the for-profit facil- and another type of antibiotic. Of seven specif- ities perform immunizations. Both government ic drugs--chloroquine, Fansidar, quinine, Sep- and nonprofit facilities rely primarily on the trin (or other antibiotics), mebendazole, and district or health subdistrict for vaccine sup- Piriton--25 percent of patients receive two plies. Indeed, since the supply of vaccines is a types of drugs, and 6 percent receive three or more types. Table 13. Probit Analysis of Prescription of Antibiotics Government facilities exhibit systematic dif- Received ferences in prescription practice, depending on Variable antibiotic z statistic the staffing of the facility. The proportion of Total number of staff (staf_num) 0.017 ­1.82 patients receiving some form of antibiotic is Only unqualified staff over 60 percent in government facilities with- (unqual) (0/1) ­0.193 ­0.96 out a qualified nurse, clinical officer, or doctor Ownership/staffing interaction but less than 50 percent in facilities with qual- (gov_facility*unqual) 0.791 (2.87)** ified staff (see figure 7). The results from a pro- Ownership/staffing interaction (pri_facility*unqual) ­0.104 ­0.37 bit analysis (table 13) suggest that the effect of Eastern region (region 2) 0.165 (1.96)* the interaction between staffing pattern and Northern region (region 3) 0.129 ­1.3 ownership is statistically significant in a multi- Western region (region 4) ­0.840 ­0.87 variate framework. Constant ­0.259 (2.78)** There is some regional variation in the pat- Number of observations 1,516 tern of drugs received by patients in govern- * Significant at the 5 percent level. ment facilities, which may reflect differences in ** Significant at the 1 percent level. Note: Figures in parentheses are the absolute values of z statistics. prescription practices or in access to drugs in Source: Health facility survey, 1999/2000 (Exit Poll). 22 Health Care on the Frontlines: Survey Evidence on Public and Private Providers in Uganda vertical (national) program, the facilities that however, varies considerably from month to receive vaccines from the government are the month, and some facilities reportedly received ones that carry out immunizations. no supplies during the six-month period. The Data were collected at the district level on number of doses of vaccines supplied to gov- the total amount of vaccines supplied to the ernment and nonprofit facilities is very similar sampled facilities in the last six months of fis- for most vaccine types (see table 14). cal 1999/2000 (January­June 2000). Informa- At the facility level, data were collected on tion collected from the records of the district vaccine stock-outs. Approximately 40 percent medical store focused on five vaccines: Bacillus of government and nonprofit facilities report Calmette-Guérin (BCG), for tuberculosis; having run out of some or all vaccines in fiscal polio; measles; tetanus toxoid; and diphtheria, 1999/2000. Most facilities are reportedly pertussis, and tetanus (DPT). The records dis- resupplied without much delay (see table 15), tinguish between regular supplies and vaccines but in some cases stock-outs last for a consid- supplied for national and district immuniza- erable period, ranging from 1 to 12 weeks. tion days. Comparison of the supply of vaccines over As is shown in table 14, facilities receive six months with the number of vaccinations only a limited supply of vaccines for immu- carried out during the same period is possible, nization days. With the exception of measles although it is difficult because of a lead in sup- vaccine, the average monthly supply is less ply and a lag in actual vaccinations. The rela- than five doses. In fact, there are no supplies tionship in any given period will always be for immunization days in the central and tenuous. Subject to these limitations, figure 8 northern regions, and only one facility in the shows that there is frequently a big discrepan- western region receives these supplies.22 In cy between the two numbers.23 In some facili- general, the regular supply is a more important ties the supply is three to four times greater source of vaccines than special supplies for than the actual number of vaccinations carried immunization days. The "regular" supply, out. This could reflect a national policy of Table 14. Average Monthly Supply of Vaccines to Government and Nonprofit Facilities For immunization days (doses) Regular supplies (doses) Facility ownership and vaccine Mean Minimum Median Maximum Mean Minimum Median Maximum Government (n=63) BCG 1 0 0 67 131 0 117 467 Polio 4 0 0 167 172 0 140 1,033 Measles 306 0 0 1,933 60 0 50 210 Tetanus toxoid 0 0 0 0 111 0 100 533 DPT 1 0 0 67 147 0 143 340 Nonprofit (n=28) BCG 2 0 0 67 139 0 125 425 Polio 2 0 0 67 146 0 133 400 Measles 46 0 0 983 61 0 67 200 Tetanus toxoid 4 0 0 67 105 0 100 250 DPT 2 0 0 67 142 0 137 383 Note: BCG, Bacillus Calmette-Guérin vaccine; DPT, diphtheria, pertussis, and tetanus vaccine. For facilities with missing data for one or more months, the average supply is calculated as a mean of the months for which data are available. Source: Health facility survey, 1999/2000 (Facility Data Sheet). Public and Private Health Care Providers 23 Table 15. Vaccination Stock-Outs and Resupply, Government and Nonprofit Facilities Percentage of Average number of facilities with immediate supply weeks to resupply other facilities Government Private nonprofit Government Private nonprofit BCG 45 36 4.3 4.1 Polio 54 73 3.5 4.7 Measles 74 82 2.3 3.0 Tetanus 88 -- 1.7 -- DPT 40 73 2.5 2.3 -- Not available. Note: BCG, Bacillus Calmette-Guérin vaccine; DPT, diphtheria, pertussis, and tetanus vaccine. Source: Health facility survey, 1999/2000. opening vaccine vials even for small numbers MEDICAL CONSUMABLES. The data suggest that of patients, but other factors may also be at not only do all government facilities receive work, and the issue deserves further attention. free supplies but that so too do large propor- tions of private for-profit and nonprofit facili- Other Inputs ties (14 and 65 percent, respectively). More detailed data on quantities and sources of sup- Data were collected at the facility level on the plies for private facilities, especially for-profit supply and availability of other inputs. Ques- facilities, are incomplete. More than 40 per- tions focused on four areas: medical consum- cent of the government facilities report having ables (bandages, cotton wool, syringes, gloves, run out of supplies in the course of fiscal and the like); contraceptives, by type; nonmed- 1999/2000. In many cases the facilities were ical consumables (fuel, kerosene, utilities, uni- restocked within a week, but some facilities forms, detergents, and so on); and capital inputs report waiting several weeks and in some cases (furniture, equipment, and means of transport). up to 20 weeks. A reflection of this situation is Figure 8. Vaccine "Wastage" (Polio and BCG Vaccine), Government and Nonprofit Facilities Polio BCG 450 450 400 400 350 350 doses doses 300 300 250 vaccine 250 vaccine of of 200 200 150 150 supply supply 100 100 Mean Mean 50 50 0 0 0 40 80 120 160 200 240 280 320 360 400 0 40 80 120 160 Vaccinations performed (monthly average) Vaccinations performed (monthly average) Source: Health facility survey, 1999/2000. 24 Health Care on the Frontlines: Survey Evidence on Public and Private Providers in Uganda that almost 40 percent of government facilities percent of government facilities and 15­20 report occasionally buying their own medical percent of private facilities have a truck, consumables. The supply of syringes appears minibus, or car. For government facilities, to be particularly problematic. motorcycles are more common (33 percent), and almost 70 percent of facilities report hav- CONTRACEPTIVES. As with medical consumables, ing a bicycle. The differences in the availability almost all government facilities receive free con- of equipment across ownership categories (see traceptive supplies from the district or subdis- table 16) reflect in large part differences in the trict, primarily in the form of pills, injectable range of services. For example, many for-profit contraceptives, and condoms. Many private for- facilites do not offer immunizations and hence profit and nonprofit facilities also report receiv- do not need refrigerators. ing free supplies. More than half of the govern- ment facilities experienced stock-outs in fiscal 1999/2000, sometimes for considerable periods. User Fees and Financing Almost 30 percent of government facilities pur- chased their own supplies at some point. This section discusses the size, scope, and uti- lization of user fees. It also looks at other NONMEDICAL CONSUMABLES. Many government sources of funds and at facility budget prac- facilities (65 percent) receive free supplies of tices and expenditure patterns. nonmedical consumables, although there appear to be some regional imbalances. (The User Fees share is 90 percent in the central region but only 33 percent in the western region.) In gen- In fiscal 1999/2000 all facilities charged user eral, for-profit facilities do not receive free sup- fees for some services. In 2001 user fees were plies, while approximately 50 percent of pri- abolished in the public sector. Before that, fee vate nonprofit facilities do. Kerosene, structures for government facilities were set by detergents, and, to some extent, fuel appear to the district, the health unit management com- be particularly important items in this catego- mittee (HUMC), or both. By contrast, in most ry. Many government facilities report long private for-profit facilities, the fee structure periods of stock-outs, particularly of kerosene. appears to be the responsibility of the in-charge, although staff and the HUMC are sometimes CAPITAL INPUTS. A majority of facilities in all involved. In private nonprofit facilities, fees are categories report having some means of trans- set either by the in-charge or by the HUMC. port, but almost 30 percent do not. Only 5 With the exception of some private for-profit Table 16. Availability of Equipment by Ownership Category (percentage of facilities) Type of equipment Government Private for-profit Private nonprofit Sterilization equipment 100 83 100 Refrigeration equipment 90 13 66 Weighing scales 94 77 93 Height measurement equipment 41 13 16 Blood pressure machine 90 100 86 Microscope 44 50 61 Sets of protective clothing 43 43 45 Source: Health facility survey, 1999/2000 (Health Facility Questionnaire). Public and Private Health Care Providers 25 Table 17. User Fees by Ownership Category (Ugandan shillings) Government Private for-profit Private nonprofit Low Median High Low Median High Low Median High OPD (new patient) 500 500 600 1,000 2,500 5,000 300 1,000 3,000 OPD (reattendance) 0 200 500 0 0 0 0 0 0 Bed per day 0 0 0 0 0 0 0 0 500 Minor surgery 0 500 800 250 1,000 3,000 0 750 1,500 Antenatal care 300 500 500 0 1,000 1,000 500 500 1,000 Family planning 0 0 500 0 500 1,000 0 0 0 Medical care 300 500 500 1,500 3,000 5,000 250 1,250 3,500 Eye care 0 0 500 0 0 0 0 0 0 Mental health care 0 0 0 0 0 0 0 0 0 Dental health care 0 0 0 0 0 0 0 0 0 Delivery 1,000 2,000 3,000 0 5,000 8,000 0 3,500 6,000 Note: OPD, outpatient day. Low refers to the 25th percentile; high refers to the 75th percentile. Source: Health facility survey, 1999/2000 (Health Facility Questionnaire). facilities, facilities report keeping records of user Most patients (89 percent) interviewed in the fee revenues. As table 17 shows, there is some- exit polls report paying for the services received. times a great deal of variation between facilities, Interestingly, the proportion of clients paying is even within the same ownership category. higher in government than in private nonprofit Differences in user fees across ownership facilities (see table 19). Over 92 percent of categories are not very surprising, but there are patients report receiving some drugs following also considerable differences among govern- the consultation. In most cases--in particular, in ment facilities. This is partly reflected in government facilities--the patients did not pay regional differences as to whether facilities separately for the drugs received. charge for specific interventions (see figure 9). For example, most facilities in the eastern and Figure 9. Proportion of Government Facilities central regions report charging for reatten- Charging Fees, by Region dance of outpatients; in the western and north- ern regions the shares are considerably smaller. Most government facilities report not charging Delivery for drugs. For-profit and nonprofit facilities typically do charge. The pattern of exemptions differs across Family planning ownership categories. In general, exemptions are less common in nonprofit facilities than in Minor surgery government facilities and are even rarer in for- profit facilities. The exception is the exemptions in many facilities for facility staff and their rela- OPD (reattendance) tives (see table 18). On average, according to the in-charges at the sampled facilities, of every 100 0 10 20 30 40 50 60 70 80 90 100 Percent patients, approximately 16 in nonprofit and in Central Eastern Northern Western government facilities are exempted from pay- ment, compared with 7 in for-profit facilities. Note: OPD, outpatient day. Source: Health facility survey, 1999/2000. 26 Health Care on the Frontlines: Survey Evidence on Public and Private Providers in Uganda Table 18. Proportion of Facilities That Exempt Specific Patient Groups by Ownership Category Patient group Government Private for-profit Private nonprofit Patients with chronic diseases (e.g., tuberculosis) 71.8 0 23.8 Elderly 59.3 13.3 40.9 Very poor 75.3 33.3 59.1 Facility staff 75.3 73.3 81.8 Relatives of staff members 30.9 50.0 40.9 Local government officials 12.4 0 6.8 Relatives of local government officials 6.2 0 2.3 Local government politicians 12.4 0 4.6 Relatives of local government politicians 3.7 0 2.3 Members of the health unit management committee 26.3 0 27.9 Source: Health facility survey, 1999/2000 (Health Facility Questionnaire). The pattern of total payments, including that the information in table 20 is not con- consultation and drugs, is shown in table 20. trolled for the type of services provided or for Although a higher proportion of clients in subsidies received. The higher prices in private nonprofit facilities is treated free of charge, facilities may therefore partly reflect a differ- the mean and median payments are lower in ent service range than in government facilities government facilities, where almost 65 percent or a different financing structure (without sub- of clients pay USh 500 or less. By contrast, pri- sidies). The data on actual payments as report- vate for-profit and nonprofit providers are ed by clients appear broadly consistent with able to charge 40 to 50 percent of their clients the charges and exemption patterns reported more than USh 2,000. It is important to note by facility managers. Table 19. Payment for Services by Ownership Category and Region Paid for service Received drugs Paid separately for drugs Ownership category Government 90.2 94.1 3.1 Private for-profit 94.3 92.2 13.6 Private nonprofit 81.5 90.6 16.7 Total (all facilities) 88.5 92.7 8.7 Government facilities by region Central 93.6 95.0 3.6 Eastern 92.1 91.3 5.1 Northern 91.7 95.0 0.0 Western 83.1 96.0 0.8 Private (for-profit and nonprofit) facilities by region Central 78.8 90.4 7.9 Eastern 89.7 86.8 3.8 Northern 88.1 92.7 6.7 Western 96.0 97.3 46.3 Source: Health facility survey, 1999/2000 (Exit Poll). Public and Private Health Care Providers 27 Table 20. Amount Paid for Services, in Ugandan Shillings, by Ownership Category (percentage of clients) Government Private for-profit Private nonprofit Total (all facilities) No payment 9.1 5.0 17.5 10.7 1­500 55.7 19.7 17.8 38.4 501­1,000 20.6 11.1 10.7 16.1 1,001­2,000 10.3 13.3 15.6 12.3 2,001­5,000 4.2 34.1 23.0 15.0 5,000+ 0.1 16.9 15.4 7.5 Source: Health facility survey, 1999/2000 (Exit Poll). Utilization of User Fee Revenues support. Seven of the 10 sampled districts report receiving financial or other assistance Approximately half of the government and from donors for health provision at the facility nonprofit facilities report preparing a budget level. Financial support to the districts ranges for use of revenues; the remaining facilities from USh 175 million to USh 545 million. In report "spending the funds as they arrive." Pri- addition, many districts reportedly receive in- vate for-profit facilities have more independ- kind support, including drugs, supervision, ence in the management of revenues. User fee means of transport, rehabilitation, and equip- revenues are the main source of financing for ment (for example, laboratory equipment, gen- these facilities, and it is surprising that only 17 erators, or computers). percent of them report preparing a budget. At the facility level, the in-charge was asked Budgets for government facilities are approved whether the facility received any money apart by the HUMC, the district, or both. The from allowances and user fees for its opera- HUMC appears to have the primary responsi- tions in fiscal 1999/2000.24 Sixty-eight percent bility for approving the budgets of nonprofit of the government facilities and 84 percent of facilities, although donors or the village may the nonprofit facilities report receiving some occasionally play a role. money. Only one private for-profit facility As figure 10 shows, the way in which user fee revenues are utilized differs somewhat Figure 10. Spending of User Fee Revenues across ownership categories, but as a general by Ownership Category rule, allowances, staff wages, and drugs and medical expenditures account for about 60­70 100 percent of expenditures. It should be kept in 90 mind that the importance of user fee revenues 80 Other Bank savings as a source of financing varies considerably 70 Equipment across ownership categories, and this has 60 Transport Fuel and other implications for the composition of spending. 50 nonmedical expenditures Percent Drugs and other 40 medical expenditures Staff wages Other Financing Sources and Facility 30 Allowances Expenditures 20 10 0 General data on financing were collected at Government Private Private for-profit nonprofit both district and facility levels. At the district level, questions focused primarily on donor Source: Health facility survey, 1999/2000. 28 Health Care on the Frontlines: Survey Evidence on Public and Private Providers in Uganda receives funds from an outside source. For the (d) deliveries; and (e) inpatient days. Corre- facilities that do receive money, the district is sponding data were collected for the same the most important source for 89 percent of three months from the district level, making it the subsample, although some facilities report possible to verify the reliability of the facility- receiving money from the subcounty, the level data.25 Data on the number of vaccina- health subdistrict, a donor, or an NGO. tions carried out by the sampled facilities were The total amount of reported financial aid also collected, for BCG, polio, measles, tetanus varies considerably across facilities. In the case toxoid, and DPT. At the facility level, these of government facilities, the amounts are data covered the last six months of fiscal small, with a median of USh 245,000. By con- 1999/2000 and are calculated from the daily trast, the median for nonprofit facilities is USh records (tally sheets) for vaccinations. Corre- 3,150,000. In most cases facilities report that sponding data were collected at the district they receive all or a substantial proportion of level but only covered the last three months of their financing from the district. fiscal 1999/2000 (April­June 2000). The data on expenditures are less complete, particularly for nonprofit providers, and there OUTPATIENTS AND DELIVERIES. The number of may be differences across facilities in the outpatients seen per month varies considerably expenditures covered. Keeping these caveats in across facilities, from 15 to nearly 2,000.26 mind, we estimate that median spending is USh The average number of outpatients per month, 283,200 for government facilities and USh for all facilities, is 419; the median is 368. The 2,637,300 for nonprofit facilities. This is average monthly number of deliveries per facil- broadly consistent with the financing data. ity is 6.75 (median, 5). As can be seen from Spending by government facilities is almost table 21, the number of outpatients is higher in exclusively on allowances. For nonprofit facil- government facilities, in part reflecting the ities, the most important categories of expendi- larger size of these units. Government facilities ture are wages, drugs, and fuel. in the central region see fewer outpatients, on average, than facilities in other regions but per- form more deliveries. Outputs and Efficiency Individual facilities sometimes exhibit notable variation from month to month. To a The survey collected detailed information on large extent, this variation appears to be idio- facility outputs at both facility and district lev- syncratic, although there is some evidence of els. A summary of this information is present- seasonal trends in the data, in particular in the ed here and is followed by a discussion of effi- number of outpatient visits. Specifically, uti- ciency in health care facilities. lization appears to increase in November­ January, possibly due to increased prevalence Facility Outputs of malaria.27 The number of deliveries appears to be lower for November­January; the num- Data on outpatients and deliveries were col- bers surge in May and June. lected for a 12-month period, July 1999­June Turning to the disaggregated data for the 2000. More disaggregated data on patient period April­June 2000, we note that approx- composition and inpatient numbers were col- imately 75 percent of all outpatients are new lected for the period April­June 2000. These patients. As figure 11 shows, the proportion data include (a) outpatient visits--children of reattenders is particularly low in govern- (new cases); (b) outpatient visits--adults (new ment facilities but is as high as 40 percent in cases); (c) outpatient visits--reattendance; for-profit facilities. The difference may reflect Public and Private Health Care Providers 29 Table 21. Number of Outpatients and Deliveries by Ownership Category Number of outpatients Number of deliveries Low Median High Low Median High All facilities Government 301.8 474.9 617.1 1.8 4.5 8.5 Private for-profit 127.0 204.6 351.4 3.3 7 11.3 Private nonprofit 152.1 251.7 510.9 1.5 4.3 9.8 Total (all facilities) 203.2 367.9 569.3 1.8 4.5 9.6 Government facilities Central 240.9 325.7 474.9 2.4 5.4 9.5 Eastern 347.3 608.5 836.0 1.8 4.8 9.9 Northern 421.3 519.9 782.6 1.8 2.6 4.1 Western 385.3 541.3 607.0 0.6 1.7 7.3 Total (all government facilities) 301.8 474.9 617.1 1.8 4.5 8.5 Note: Monthly figures are based on facilities for which data were available for at least 6 out of 12 months. Facilities that do not offer maternity services were not considered in compiling the number of deliveries. A small proportion (approximately 10 percent) of the sampled facilities was excluded because of lack of data. These were primarily private for-profit facilities. Low refers to the 25th percentile; high refers to the 75th percentile. Source: Health facility survey, 1999/2000 (Facility Data Sheet). differences in patient mix or the fact that for- approximately 40 tetanus, measles, and BCG profit and nonprofit facilities do not usually vaccinations and 115 DPT and polio vaccina- charge reattending clients, as government tions. (These averages exclude data on measles facilities do. vaccinations from three facilities that conduct- Approximately 35 percent of outpatients are ed vaccination campaigns during the period children. The proportions of new and reat- under consideration.)28 There is considerable tending patients and of patients under and over age five vary considerably among facilities. Figure 11. Composition of Outpatients Government facilities also show some regional by Ownership Category patterns in patient composition. For example, 100 in the eastern region more than 45 percent of outpatients are under age five, as against 35 percent nationally. 80 INPATIENT CARE. According to the information total 60 of gathered from the facility in-charges, over 60 percent of facilities provide inpatient care. This 40 percentage is higher for nonprofit and for- Percentage profit facilities and slightly lower for govern- 20 ment facilities. In the subsample of facilities for which data are available, the median number of inpatient days is 21. 0 Government Private Private Government Private Private for-profit nonprofit for-profit nonprofit VACCINATIONS. Data on vaccination numbers New Under age 5 Reattending Age 5 or older are based on aggregations of daily tally sheets. On average, facilities carry out per month Source: Health facility survey, 1999/2000. 30 Health Care on the Frontlines: Survey Evidence on Public and Private Providers in Uganda variation in vaccination activity for different minimization may not be an appropriate behav- ownership categories (see figure 12) and across ioral assumption; for example, staff allocation facilities of different types. For government may be driven by a policy of ensuring minimum facilities, the average number of vaccinations service standards. Consequently, we should not carried out is substantially higher in the eastern necessarily expect to observe allocative efficien- region, for all types of vaccines. cy in the public sector, and many studies of effi- ciency in the public sector therefore restrict Costs and Efficiency attention to technical efficiency.31 There are several techniques for analyzing Broadly speaking, the concept of efficiency has efficiency in service delivery (see, for example, to do with the relationship between inputs and Barnum and Kutzin 1993; Fried, Lovell, and outputs.29 In modern efficiency measurement it Schmidt 1993). Many of them, such as data is customary to distinguish between technical envelopment analysis and the econometric and allocative efficiency (Farrel 1957).30 Tech- analysis of cost functions, make intensive use nical efficiency refers to the maximization of of both data and analytical input. The scope of output using a given set of inputs; allocative this section is limited to an analysis of output efficiency reflects substitution between inputs per worker and labor cost per unit of output. with different prices to achieve minimum costs. Even in this limited analysis of efficiency, the These two measures can be combined to pro- methodological limitations have to be kept in vide a measure of total economic efficiency. mind. Most important, there is considerable Application to the public sector of the con- heterogeneity in service outputs: cept of allocative efficiency is often fraught 1. Quality may differ substantially across with methodological difficulties. First, the health care facilities and even between individ- choice of inputs is often beyond the control of ual cases at a given facility. For example, a thor- the individual facility, (at least in the public sec- ough outpatient consultation with a doctor is a tor), and where the facility can exercise discre- very different service from a rushed consulta- tion, price signals may be weak. Second, cost tion with a poorly trained nursing aide. The equipment and amenities of the facilities may not be of equal quality. In principle, it is possi- Figure 12. Number of Vaccinations per Month ble for the analysis to control for quality differ- by Ownership Category ences, but this has proved difficult in practice. 2. Within a particular service category, there DPT can be a noticeable variation in case mix and case complexity (severity) across facilities, and Tetanus this too causes problems for comparability. For Government example, "inpatient days" can range from Private nonprofit Measles cases involving simple interventions and limit- Private for-profit ed monitoring to highly complex cases that require a broad range of material and human Polio resources. Differences in case mix can arise from the socioeconomic characteristics of the BCG population in the provider catchment area, or 0 20 40 60 80 100 120 140 more complex cases may seek out providers with particular characteristics. Note: BCG, Bacillus Calmette-Guérin vaccine; DTP, diphtheria, pertussis, and tetanus vaccine. 3. Finally, in addition to problems relating to Source: Health facility survey, 1999/2000. the comparability of output measures in specif- Public and Private Health Care Providers 31 ic categories, most health care providers pro- OUTPUT PER HEALTH WORKER AND UNIT COSTS. vide a wide range of services. Even with a small Using this definition of output, table 22 pres- number of aggregated categories of interven- ents outputs per health worker for different cat- tions and services (for example, inpatient days egories of facility.33 Some patterns are evident: and outpatient visits), the issue arises of how to compare quantities of output across facilities · Across the total sample there are consid- with different service mixes. A standard tech- erable differences, ranging from 10.9 to nique for dealing with multioutput production 583.8 outpatient-equivalent service units is to construct an output index using market per worker. prices as weights. In the health sector, where · Output per worker varies to some extent output prices do not exist or are administra- across ownership categories, with low tively set, we must rely on ad hoc weights. numbers for nonprofit facilities.34 · There appear to be notable regional dif- CONSTRUCTING AN OUTPUT INDEX. The survey ferences, with low output per worker in was restricted to primary health care facilities, the central region for both government but even at this basic level many facilities deliv- and private facilities. er a wide range of services. We construct an output index based on the amount of staff time The reasons for these observed differences required to perform particular tasks. This out- are complex and cannot be explored fully patient-number-equivalent service unit (SU) here.35 It is, however, possible to conclude that index is defined as: workloads in facilities in the central region are lower than in other regions. Looking at output SU = 9 × inpatient days + per worker in a multivariate framework, it also 12 × deliveries + 0.5 × immunizations + appears that, overall, the presence of staff with 1 × outpatient numbers.32 higher qualifications does not have a positive Table 22. Health Worker Productivity by Ownership Category and Region (output per worker, including immunizations, expressed in service units) Minimum 25th percentile Median 75th percentile Maximum Ownership type Government 28.5 63.9 92.5 145.5 326.3 Private for-profit 17.8 58.7 79.6 124.1 583.8 Private nonprofit 10.9 52.2 69.0 106.8 210.3 Total (all facilities) 10.9 58.4 86.7 129.9 583.8 Government facilities by region Central 28.5 44.6 60.6 78.4 145.5 Eastern 50.8 100.0 147.2 169.5 236.9 Northern 67.9 88.6 92.7 127.6 195.0 Western 85.4 92.9 123.3 164.6 326.3 Private (nonprofit and for-profit) facilities by region Central 10.9 50.0 64.3 76.8 272.5 Eastern 21.3 52.2 66.0 84.0 583.8 Northern 17.8 44.1 89.4 160.6 206.2 Western 68.2 88.7 106.8 138.0 286.5 Source: Health facility survey, 1999/2000 (calculations based on Facility Data Sheet). 32 Health Care on the Frontlines: Survey Evidence on Public and Private Providers in Uganda impact. Clearly, no general conclusions can be neration and lower-than-average labor produc- drawn from this finding, and higher-qualified tivity in that region. staff may have a considerable effect on quality and on the range of services offered. Still, the issue of optimal staffing patterns, given the Client Perceptions costs of different types of worker, needs to be explored in greater detail. The great majority (88 percent) of the exit poll Labor cost per output (that is, unit cost) is respondents report that they normally seek care clearly related to productivity but also depends at the facility at which the interview took place. on the staff mix and the level of remuneration. Most of these individuals came to the facility to As noted above, average remuneration is high- receive treatment, although a considerable share er in government than in nonprofit and for- was seeking preventive care--that is, immuniza- profit facilities, and we would expect unit costs tions and antenatal care (see table 24). There are to reflect this difference. Indeed, as can be seen notable differences across ownership categories from table 23, the higher average remunera- and regions in the reasons given for attending a tion in government facilities appears to more specific facility. In particular, it is clear that non- than offset their higher productivity; unit costs profit facilities have an important role in immu- are higher in government than in for-profit and nizing children. The proportion of clients who nonprofit facilities. (The difference between report having come to the clinic to receive ante- government and nonprofit facilities is statisti- natal care is substantially higher for for-profit cally significant.) This difference is largely than nonprofit facilities. driven by the very high unit costs in govern- Clients were also asked about the care re- ment facilities in the central region, a finding ceived at the facility.36 Overall, patients report- consistent with the higher-than-average remu- ed that the service was friendly (see table 25). In Table 23. Unit Costs, Labor, by Ownership Category and Region (remuneration per output, including immunizations; Ugandan shillings) Minimum 25th percentile Median 75th percentile Maximum Ownership type Government 234.5 754.2 1,188.4 2,118.8 6,424.7 Private for-profit 147.5 505.0 740.2 1,686.3 4,009.5 Private nonprofit 140.6 427.2 670.8 1,365.6 4,036.7 Total (all facilities) 140.6 614.1 893.9 1,821.8 6,424.7 Government facilities by region Central 805.3 2,002.1 2,457.2 3,783.4 6,424.7 Eastern 410.9 532.1 736.6 1,142.2 2,266.7 Northern 673.9 820.6 888.9 1,242.7 1,807.0 Western 234.5 590.5 965.8 1,174.9 1,409.1 Private (nonprofit and for-profit) facilities by region Central 330.2 632.9 911.2 1,904.1 4,036.7 Eastern 147.5 547.6 692.5 1,365.6 1,801.5 Northern 140.6 225.0 423.4 1,201.0 4,009.5 Western 254.8 414.3 614.1 680.6 874.9 Source: Health facility survey, 1999/2000 (calculations based on Facility Data Sheet). Public and Private Health Care Providers 33 Table 24. Clients' Reasons for Coming to the Facility by Ownership Category and Region (percent) Receive Immunize Antenatal Family Minor Lab treatment child care planning Delivery surgery results Ownership type Government 89.4 6.1 9.6 0.8 0.4 0.4 1.3 Private for-profit 84.3 1.8 16.0 1.1 1.1 0.4 1.8 Private nonprofit 88.0 7.1 7.1 0.7 0.9 0 3.5 Total (all facilities) 88.1 5.6 10.1 0.8 0.7 0.3 2.0 Government facilities by region Central 92.5 4.4 15.8 1.4 0.7 0.7 3.0 Eastern 90.0 6.3 4.6 0 0.4 0 0.4 Northern 88.4 4.1 5.8 0.8 0 0.8 0 Western 83.2 10.7 8.1 0.7 0 0 0 Private (nonprofit and for-profit) facilities by region Central 88.8 7.2 7.6 0 0.8 0.4 7.6 Eastern 81.7 5.6 20.8 1.5 2.5 0 0 Northern 90.8 0 2.8 2.8 0 0 0.9 Western 86.0 4.0 8.0 0 0 0 0 Source: Health facility survey, 1999/2000 (Exit Poll). Table 25. Client Perceptions Regarding Services by Ownership Category and Region Friendly Information Advice on Prompt Information service about ailment medication attention about charges Ownership type Government 96.9 64.9 93.1 79.8 39.0 Private for-profit 99.3 76.3 95.6 94.6 49.6 Private nonprofit 99.1 77.2 93.3 84.8 51.0 Total (all facilities) 97.9 70.4 93.6 83.9 44.2 Government facilities by region Central 96.0 81.0 96.6 82.4 62.9 Eastern 99.2 73.3 96.6 92.9 19.5 Northern 93.4 43.8 77.4 50.4 57.0 Western 98.0 34.1 93.1 77.0 5.6 Private (nonprofit and for-profit) facilities by region Central 99.2 85.8 93.0 90.3 74.2 Eastern 99.0 89.3 95.3 94.9 34.1 Northern 98.2 58.7 91.1 79.6 78.5 Western 100.0 57.6 97.1 84.6 16.8 Source: Health facility survey, 1999/2000 (Exit Poll). 34 Health Care on the Frontlines: Survey Evidence on Public and Private Providers in Uganda some cases information about the clients' ail- to clients about charges, although in this case it ments appears to be limited, in particular in gov- is the eastern and western regions that had the ernment facilities in the northern and western lowest scores. regions. Most patients report having received Finally, clients were asked why they had directions for using the medicine prescribed, but chosen to visit the particular facility (see table enumerators claimed that in many cases the 26). Proximity is the most important factor clients did not sound sure. Attention was in gen- overall, particularly for government facilities. eral prompt, although the situation again Good treatment and staff characteristics are appears worse in the northern and western more important considerations for individuals regions. Similar regional disparities can be attending for-profit and nonprofit facilities. observed with respect to information provided Table 26. Clients' Main Reason for Choosing a Specific Facility by Ownership Category Government Private for-profit Nonprofit Total Proximity 48.6 30.2 39.2 42.8 Good treatment and service 32.1 48.5 38.9 36.9 Good health workers 5.3 13.4 11.7 8.5 Less expensive 12.7 5.5 7.3 10.0 Other 1.3 2.5 2.9 1.9 Note: Numbers may not sum to 100 percent because of rounding. Source: Health facility survey, 1999/2000 (Exit Poll). CHAPTER 5 Summary and Recommendations for Further Research T his paper has presented descriptive sta- Ownership and Health Facility tistics from a baseline survey of 155 pri- Performance mary health care facilities carried out in Uganda in 2000. It has given an There is no doubt that private for-profit and overview of the survey and the sample, dis- nonprofit health care providers are important cussed oversight, management, and competi- in Uganda. Indeed, Hutchinson (2001) finds tion in Ugandan frontline health care, and that the government, for-profit, and nonprofit explored inputs, costs, user fees, financing, sectors account for roughly equal shares of the outputs, and efficiency issues. It has also pro- country's health care. On the basis of data vided a client perspective, using data from the from household surveys, it appears that gov- exit poll of patients. Although the nature of the ernment facilities are the most important data makes exact measurement difficult, the providers of immunizations, modern deliveries, survey has demonstrated that it is possible to and reproductive health care, while the private collect data of sufficient quality to arrive at sector provides the bulk of curative care. The ballpark estimates and bounds for important importance of the private sector is in part a variables such as pricing, income, infrastruc- consequence of the almost total collapse of the ture, and human resources. At this stage, the government system during the 1970s. But pri- analysis of the survey data is largely diagnostic, vate sector provision has also been increasing- yielding sample averages for a number of vari- ly encouraged, and in a recent sector strategy ables. This section discusses issues that the government proposed that new forms of emerged from the first round of analysis, collaboration with the private sector be focusing on four that are considered of partic- explored (Ministry of Health 2000). The move ular importance: ownership and health facility toward increased reliance on the private sector performance, human resources, user fees and is consistent with policy trends in many other financing, and drug use. countries and is seen as a means of addressing what some consider to be endemic problems with public sector service delivery (see Birdsall and James 1993). 35 36 Health Care on the Frontlines: Survey Evidence on Public and Private Providers in Uganda Weaknesses in the delivery of health services between government, for-profit, and nonprofit in Ugandan government facilities are well doc- providers at the micro level. umented (and are similar to those in other As can be seen from table 27, it is difficult to developing countries). They include lack of draw firm conclusions concerning the relative drugs, equipment, and materials; lack of incen- merits of the different types of facility. A mul- tives for staff; staff shortages and absenteeism; tivariate analysis is required (and is in fact cur- and inappropriate staff attitudes (Ministry of rently under way). The survey captured some Health 1998). These problems lead to low lev- dimensions of quality that need to be analyzed els of utilization and effectiveness, as well as to further. Unobserved aspects of quality such as high costs. Indeed, according to a Ministry of staff attitudes may be important in explaining Health study (1997), the costs of basic services the preference for private providers that has are at least 50 percent higher in government been revealed in Ugandan household surveys than in nonprofit facilities. (Hutchinson 2001). Another issue that emerges It is often argued that these problems can be clearly from the survey is the complex interde- at least partly overcome in the private sector. pendencies between the government and the The for-profit sector is assumed to have the private sector. This is clear in the area of advantage of the profit incentive, while non- human resources, where salary levels in the profit providers are viewed as benefiting from government sector are creating what some greater organizational autonomy and flexibili- would call an uneven playing field. ty and from an organizational culture that pro- motes good performance. These are, of course, highly stylized arguments. The profit motive, Human Resources in combination with limited consumer infor- mation, may just as easily lead to overprovi- The survey finds that government facilities sion of health care or to provision of inappro- tend to have larger staffs, with higher qualifi- priate care. Similarly, organizational autonomy cations. Even so, over 60 percent of staff are and lack of accountability in the nonprofit sec- nursing aides or "other staff." This is consis- tor can result in poor coordination, duplica- tent with administrative data for 2000 show- tion, operational inefficiency, and the provi- ing that only 33 percent of established health sion of services that are not cost-effective. positions were filled by qualified staff, with the These arguments raise a number of impor- remainder either vacant or filled by unqualified tant empirical questions. What is the current nursing aides or other staff (Hutchinson 2001). policy and institutional framework for private There are some clear regional differences in sector operations? What services are currently this regard; government facilities in the central being provided by various types of provider? region are more likely to have qualified staff, What differences in quality and efficiency can and facilities in the eastern and western regions be observed across ownership types, and what are more likely to lack qualified staff. is driving these differences? What are the The evidence suggests a close link between implications of the existence of the private sec- the three types of provider through the labor tor for the overall planning and coordination market for health workers. Government dis- of health services and the efficiency of the pensaries pay higher salaries than private facil- health sector as a whole? ities, and for-profit facilities appear to pay This study can by no means answer all of more than nonprofits for qualified health staff. these questions. Nonetheless, the data present- The observed salary differences affect the ed in section 4 (and summarized in table 27) do movement of staff between provider organiza- yield a general picture of the differences tions and are highly relevant to Uganda's civil Summary and Recommendations for Further Research 37 Table 27. Differences in Health Care Facilities across Ownership Categories Issue Description Mix of services In general, government dispensaries offer a broader range of services than do private for-profit facilities. For example, for-profit facilities tend not to provide immunizations. Government facilities, however, are considerably less likely to offer laboratory services than are for-profit or nonprofit providers. Except for laboratory services, there appear to be no consistent differences in service range between government and nonprofit facilities. Staffing Government and nonprofit facilities are similar in size. For-profit facilities tend to have fewer staff. Salary level In general, government facilities pay staff more than do for-profit or nonprofit facilities. For- profit facilities appear to pay more than nonprofit facilities for qualified staff. User fees Exit polls clearly show that fees are higher in the private sector. Charges are, in general, higher in for-profit than in nonprofit facilities. Nonprofit facilities are at least as likely as government facilities to exempt clients from payment. Activity level The activity level is, in general, higher in government facilities, partly reflecting higher staffing levels. Specifically, the numbers of outpatients and inpatients are higher in government facilities than in for-profit or nonprofit facilities. Government facilities also perform more vaccinations, in particular in comparison with for-profit facilities, mainly because of the vertical programs of vaccine supply. Numbers of deliveries are similar for government and nonprofit facilities. The proportion of patients who are reattending for the same ailment is higher in private facilities. A considerable proportion of clients in for-profit facilities (16 percent) report coming to the facility for antenatal care. Output per Output per health worker is higher in government than in private facilities. This is particularly health worker true if government facilities in the central region, where staffing levels are comparatively high, are excluded. Due to higher average remuneration in government facilities, their unit labor cost per output is higher (if the central region is excluded). It should be noted that this information is difficult to interpret. For example, do high levels of output per worker represent efficiency, or do they indicate excessive workload and insufficient patient time and hence lower quality of service? Drug use There is considerable similarity across ownership types in the provision of drugs to clients. Apparent "overuse" of drugs does not appear to be restricted to government facilities, but there is some evidence of "overprescription" of antibiotics in government facilities run by staff with low or no qualifications. Client perceptions Clients are more likely to report "good treatment" or "good health workers" as a reason for attending private facilities, in particular for-profit facilities. In the case of government facilities, proximity is the most important reason for choosing a particular facility. service reform. They highlight the importance with for-profit facilities. Inequalities in remu- of looking at the entire health care labor mar- neration among and within government facili- ket and point to the need to compare civil serv- ties are also striking. There is a regional pat- ice pay not only within the public sector but tern, with noticeably higher compensation in also with its private counterparts (competi- the central region, for all categories of staff. tors). But differences are also driven by the source of Although staff in government dispensaries financing. For example, staff financed by the earn more than staff in the private sector, they subcounty or the facility receive considerably are much more likely to experience delays in less than staff in the same category who are salary payments, particularly in comparison financed by the district. The analysis shows 38 Health Care on the Frontlines: Survey Evidence on Public and Private Providers in Uganda that there are several different payrolls for for deliveries. They do not generally charge for health staff at the facility level, including staff drugs. originally employed by the central government; Charges are higher in for-profit facilities, staff recruited by district administrations prior and there is also more variation among facili- to decentralization; staff recruited under the ties in this category, perhaps reflecting quality conditional grant for primary health care from differences. Nonprofit facilities charge more the central government; and staff recruited by than government facilities but generally less subcounties and by facilities themselves (and than for-profit facilities. They exempt approx- paid for with user fees, until the fees were abol- imately the same proportion of clients from ished in 2001). This fragmentation in person- payment as do government facilities. Unlike nel management is bound to have an effect on government facilities, many private providers incentives and service quality at the facility charge per ailment, and reattendance is gener- level and hence should be brought into the ally free. Moreover, private providers tend to debate on civil service reform. charge for drugs, whereas government facilities The analysis shows a striking discrepancy do not. between facility and local government (district) In general, the information reported by the staff records. Only 56 percent of staff working facility in-charges appears to be consistent in dispensaries (with and without maternity with the data collected through the exit poll. units) were included in the district records. This is somewhat difficult to reconcile with This observation has major implications for earlier evidence that illicit charging is wide- personnel management in health care and for spread in government facilities in Uganda (Jitta the assessment of health management informa- 1996; McPake and others 1999; Mwesigye tion systems. 1996). As a rule, user fees can be expected to have an impact on utilization of health services. In a User Fees and Financing context such as Uganda, where "frivolous" use of health services is likely to be limited, the User fees have figured prominently in the poli- reduction in access may have an impact on cy agenda in Uganda (and other countries) for health outcomes. User fees do, however, have many years. The issue has gained importance potential for relaxing the financial constraint in Uganda following the abolition of user fees in the public sector, giving health workers and for primary health care in 2001. The available managers a financial stake in the facility and its data provide valuable information on charging activities and making households and commu- practices and the utilization of user fee rev- nities more demanding as to the services pro- enues. The survey of dispensaries also serves as vided. Through these channels, user fees may a baseline for analyzing the impact of the poli- result in expanded activity, improved quality, cy change in 2001 on both government and and greater efficiency (Griffin 1992; Shaw and private providers. Ainsworth 1994; Vogel 1991; World Bank On the basis of information provided by the 1987).37 facility in-charges, considerable differences in Currently, little evidence exists on the fees can be observed across and within owner- impact of user fees on utilization in Uganda. ship categories. Government facilities have Despite some methodological concerns, studies been charging fees since the late 1980s. Gov- on other countries do provide considerable evi- ernment facilities report charging USh 500 for dence that increases in user fees lead to a most services and approximately USh 2,000 reduction in utilization and, conversely, that a reduction in fees leads to an increase in utiliza- Summary and Recommendations for Further Research 39 tion. (For a review of the issues, see Reddy and ent from a comparison of drug use (measured Vandemoortele 1996.) It is to be hoped that by removal from stock) and patient numbers. ongoing analysis of household data will yield This observation can have many explanations, evidence on this issue that is specific to Ugan- including high need (patient or case mix), da. The effect of the policy change regarding overprescription, and leakage of drugs. Further user fees could also be assessed on the basis of analysis of the data may provide additional a second round of the facility survey. This information on what is driving this observa- would permit an analysis of how both user fee tion, but more detailed studies may also be and activity levels have changed in government required. and private providers in the two years follow- Second, evidence from the exit poll indicates ing the policy change. that drugs (in particular, antibiotics) are over- User fees are also expected to have an prescribed in both government and private impact on the supply side. Here, the survey facilities.38 The survey did not include consul- identifies some reasons for concern. The aboli- tation observations or "gold-standard" exami- tion of user fees can be expected to lead to an nations to assess "true" client need for drugs. increase in utilization, but how will it affect Still, the number and nature of the drugs actu- quality and resource availability at the facility ally received by patients suggest excessive and level? Many government facilities in the sam- inappropriate drug prescription. This problem ple report using revenues from user fees to pro- is by no means unique to Uganda. For exam- cure important supplies such as condoms, con- ple, Gilson and others (1993), reporting find- traceptives, detergents, and syringes. In a ings from Tanzania based on both retrospec- context where stock outages are common, a tive data (from patient registers) and reduction in user fee revenues could have a prospective data (from consultation observa- deleterious impact on facilities' capacity to tion), cite evidence that patients often receive a deliver services. In addition, a considerable large number of drugs, including one or more proportion of staff in government facilities was types of antibiotics.39 Moreover, 46 percent of financed by user fees in 2000. What happened all prescriptions issued for general consulta- to these staff following the abolition of fees? If tions were incorrect according to national the reduction in fees led to a reduction in the treatment guidelines. Problems included number of staff, what has been the impact on unnecessary or incorrect antibiotics, underpre- service delivery? These issues merit more study. scription, and incorrect dosage. Similar find- Further analysis of the data will seek to esti- ings are reported from other countries (see Fos- mate the level of user fee revenues in different ter 1993). categories of facility at the time of the survey, There are many possible reasons for inap- as well as the subsidies to government and propriate or excessive prescription of drugs.40 nonprofit facilities. Incorrect or inadequate diagnosis may be the consequence of poor skills of health workers, lack of effort, or lack of diagnostic equipment Drug Use and materials. Profit- or revenue-raising motives can lead to overprescription. Indeed, The survey highlights a number of important Gilson and others (1993) find that facilities run issues in relation to drug use and management. by religious organizations prescribed more Most important, it provides evidence, from drugs per visit across all conditions in Tanzania two sources, of excessive drug use in Uganda in than did other types of facility, and they sug- both government and private facilities. First, gest that this practice may stem from the need high and variable drug use per patient is appar- to raise revenues. Finally, prescription out- 40 Health Care on the Frontlines: Survey Evidence on Public and Private Providers in Uganda comes may be influenced by pressure from this is a case in which quality improvements clients with poor knowledge about health and can lead to cost reductions. Many studies have health care. A common example is the demand called for better training and supervision to for injectable drugs when oral preparations or improve the situation. The evidence on the no drugs at all would have been appropriate. effectiveness of this type of intervention is at Underprescription can have similar sources but best mixed (Loevinsohn, Guerrero, and Grego- may also result from a shortage of drugs. rio 1995; Ofori-Adeji and Arhinful 1996; Pare- These findings are important for two rea- des and others 1996; Rowe and others 2001), sons. First, they suggest that although drug but it does suggest the need for a mix of inter- stock-outs may be important in certain areas ventions, including measures to address the or at certain times, there are a lot of drugs in overall incentives of health workers. Although the Ugandan health system. Second, excessive this study does not provide firm answers and inappropriate drug use is not only ineffi- regarding these issues, it has demonstrated that cient but can also be harmful. In other words, the area merits further attention. APPENDIXES Appendix A. Methodology and Data Issues T his appendix describes the sample register. A reserve list of replacement facilities design, the implementation of the sur- was also drawn from the sample frame. vey, and the sources of the information Because of the unreliability of the register for collected, with attention to staff, drugs, private for-profit facilities, it was decided that vaccines, and outputs. for-profit facilities would be identified on the basis of information from the government facilities sampled.43 The administrative records Sample Design for facilities in the original sample were first reviewed at the district headquarters, where The starting point in designing the sample was some facilities that did not meet selection crite- the Ministry of Health health facility register ria and data collection requirements were for 1999. The register includes government, dropped from the sample. These were replaced private for-profit, and private nonprofit facili- by facilities from the reserve list. Overall, 30 ties but is known to be inaccurate with respect facilities were replaced.44 to the latter two. On the basis of existing infor- The sample was designed in such a way that mation, it was decided that the sample of 155 the proportion of facilities drawn from differ- facilities (dispensaries with and without mater- ent regions and ownership categories broadly nity units) would include 81 government, 30 mirrors that of the universe of facilities. private for-profit, and 44 private nonprofit Because no nationwide census of for-profit facilities. In the first stage in the sampling health facilities is available, it is difficult to process, 8 districts (out of 45) had to be assess the extent to which the sample is repre- dropped from the sample frame due to securi- sentative of this category. A census of health ty concerns.41 Ten districts were randomly care facilities in selected districts, carried out in selected from the remaining districts, implying the context of the Delivery of Improved Ser- that roughly one-quarter of the eligible dis- vices for Health (DISH) project supported by tricts were sampled.42 the U.S. Agency for International Development From the selected districts, a sample of gov- (USAID), suggests that about 63 percent of all ernment and private nonprofit facilities was facilities operate on a for-profit basis, while drawn randomly from the Ministry of Health government and nonprofit providers run 26 41 42 Health Care on the Frontlines: Survey Evidence on Public and Private Providers in Uganda and 11 percent of facilities, respectively. This the instruments were pretested in Mukono and would suggest an undersampling of private Mpigi Districts. The purposes of the exercise providers in the survey. It is not clear, however, were to assess the feasibility of the survey tools whether the DISH districts are representative in data collection and to provide a basis for of other districts in Uganda in terms of the review and finalization of the instruments. market for health care. Also, any characteriza- Nowithstanding the training and considerable tion of the health care market is dependent on field testing of the survey instruments, enumer- the definition of a "facility" or "provider." ators sometimes encountered problems in the field; these typically stemmed from poor record-keeping or the unreliability of existing Survey Implementation records, although minor problems in instru- ment design also occurred. The survey was designed and implemented by the World Bank in collaboration with the Mak- erere Institute for Social Research and the Ugan- Specific Data Issues dan Ministry of Health. The survey team con- sisted of a team leader, five supervisors, and five STAFFING AND STAFF REMUNERATION. The most research assistants. Together they formed five complete source of data on facility staffing is separate teams for the fieldwork. One team was the Facility Data Sheet. Through this instru- assigned to each region; the fifth team acted as a ment, information about all the staff members support group to the central region team, which working at the facility was collected, including had the largest number of health facilities to names, positions, remuneration, and source of cover. The team leader supervised the teams dur- finance. Because of occasional nonresponse or ing the survey period. Each team spent at least data recording errors, data for some staff are two days in its district interviewing the district incomplete. (For example, data on salaries are health official and extracting data from the dis- not available for 10 percent of the facility- trict records. On average, each team also spent level staff.) another one and a half days interviewing the in- The analysis of staff remuneration focuses charge at each facility and reviewing facility on staff for whom data were provided in the records. The total number of days spent by each facility-level questionnaire. In order to calcu- team in the field depended on the number of late the facility-level wage bill and staff costs, it facilities in the region. The fieldwork was car- is necessary to impute salary payments and ried out during October­December 2000. A lunch allowances for those observations where total of 155 health facilities (dispensaries with- these data are missing. Salaries are predicted out and with maternity units) was surveyed. For from a simple regression of salary on dummy the exit poll, exactly 10 interviews per facility variables for position, source of financing, and were carried out in approximately 85 percent of the ownership and regional location of the the facilities. In the remaining facilities the tar- facility in question. In cases where missing val- get of 10 interviews was not met, as a result of ues remain because of missing values in the low activity levels. regressors, predicted values are derived from a Before taking the survey to the field, the simpler model. To allow for the heterogeneous entire research team was trained for over three nature of "other" staff, salaries for this catego- weeks by the Ugandan team leader and World ry are estimated separately for government Bank staff. The training acquainted enumera- facilities on the basis of data on the source of tors with the instruments and techniques to be finance. In nongovernment facilities the medi- used in data collection. Following the training, an salary for "other" staff is used. For lunch Appendixes 43 allowances, values can be imputed from the poor record-keeping. For outpatient data, implicit rules in the data: the allowance is complete facility-level data (that is, data for all assumed to be (a) zero for nongovernment months) are available for approximately 75 facilities; (b) USh 66,000 for more qualified percent of government and nonprofit facilities. staff in government facilities; (c) USh 44,000 For the remaining facilities, data are missing for staff with lower qualifications; and (d) zero for some or all months. For facilities that offer for "other" staff. maternity services, data on number of deliver- ies are missing for approximately 10 percent of DRUGS. Drug supply data are based on facility government and nonprofit facilities. Some of stock cards. In the case of private for-profit the surveyed facilities began offering maternity facilities, stock cards were typically missing, services only recently, which partly explains and no reliable information could be collected. the lack of historical data. In the remaining In private nonprofit facilities, records pertained facilities, data are available for some or all only to drugs received from the district and did months. Again, data are less complete for pri- not reflect sources of drugs, dates, new stocks, vate for-profit providers. The analysis of the and stock balances.45 In government facilities, number of outpatients and deliveries is restrict- records were more consistent, although some- ed to the subsample of facilities for which data times incomplete. In particular, very few facili- are available for at least 6 out of 12 months. ties kept records on ergometrine and oral rehy- Data on vaccinations are based on aggrega- dration salts. Because of these data limitations, tions of daily tally sheets. For approximately the analysis in the report includes only a small 10 percent of relevant government and private proportion (about 30 percent) of nonprofit nonprofit facilities, no data on vaccination facilities and no private for-profit facilities. numbers are available for any month. For for- Data are available for approximately 75 per- profit facilities, data are missing for three out cent of government facilities. of eight facilities. In addition, data are missing for most months for some facilities. The analy- VACCINE SUPPLY. Data were collected on vaccine sis is restricted to facilities for which data are supply for six months. This information is far available for at least four out of six months; it from complete. Despite reports from some pri- excludes facilities for which data are not avail- vate facilities that they receive vaccination sup- able or are available for only one month. The plies from the district, no corresponding data resulting subsample comprises about 80 per- could be collected at the district level. The cent of the government and private nonprofit analysis in this report is therefore restricted to facilities that perform vaccinations and 60 per- the subsample of government and private non- cent of the for-profit facilities. profit facilities that actually perform immu- Patient composition is analyzed using nizations and for which data were available. monthly average number of patients in various Observations were excluded if there were not categories. The averages are based on disaggre- data for at least two out of six months for the gated data for three months. These data are respective vaccines. In practice, this meant reasonably complete, although less so for pri- excluding approximately 20 percent of the vate for-profit facilities. For the disaggregation facilities in the subsample. by age category, data in some facilities refer only to new patients. Assuming that the age OUTPUTS. Output data were collected for 12 composition of reattending patients is broadly months; more detailed data were collected for similar to that of new patients, this fact should 3 months. For some facilities, data were partly not bias the findings. or completely lacking, typically as a result of Appendix B. Consistency between Facility and District Records D ata on the number of outpatients, First, health facility staff may believe that deliveries, and vaccinations for the reported patient numbers will not have any three months April­June 2000 were real effect on the activities and resources of the collected at both facility and district facility and that the numbers will not be levels, permitting analysis of the consistency checked. In that case, they may decide that it is between the two sources. not worth expending effort on accurately reporting patient numbers. Second, if resources are allocated on the basis of reported patient Outpatient Numbers and Deliveries numbers, or if these records are used to define user fee revenue targets for which the facility is Here, attention is restricted to government held accountable, there may be incentives to facilities for which comparable data are avail- over- or underreport patient numbers. The able for all three months. Figure B.1 shows the seemingly poor reliability of patient data at the average number of outpatients and deliveries district level also casts some doubt on official per month for April­June 2000, as recorded at statistics on utilization. From the perspective the facility level, against the corresponding of information management, data quality is data as recorded at the district level. As is clear not the only concern; it is also worrying that from the figure, consistency between the two complete information is available at the district sources is fairly poor. Although there is no level for only 60 percent of the government clear pattern, at least for outpatient numbers facilities in the sample. there appears to be a tendency for facilities to overreport output statistics to districts, in rela- tion to the data recorded in patient registers. In Vaccinations some cases this overreporting is considerable. Patient registers appear to be the most reli- Like outpatient and delivery data, vaccination able source, as they are for the facility's use records exist at both facility and district levels. only. Taking these numbers as given, there are Facility staff record vaccinations on a tally several possible explanations for the discrepan- sheet, and monthly totals are communicated to cy with the data recorded at the district level. the district. The general impression from the 44 Appendixes 45 Figure B.1. Comparing District and Facility Output Data Outpatients Deliveries 4,000 120 3,500 100 3,000 80 2,500 data data 2,000 60 District 1,500 District 40 1,000 20 500 0 0 0 1,000 2,000 3,000 4,000 0 20 4 0 6 0 8 0 100 120 Facility Data Facility Data Source: Health facility survey, 1999/2000. data is that there is a considerable discrepancy for the three months considered is approxi- between facility and district data for many mately 33 percent, and the mean is over 100 facilities. Indeed, consistent data are available percent. Conversely, for underreporting facili- for only a minority of government and non- ties, the median discrepancy is approximately profit facilities. For most vaccinations, there is 15 percent, with a similar mean. On average, it a tendency toward overreporting. In particular, appears that for both government and non- there is a long tail of a small number of facili- profit facilities, district-level data overstate ties that overreport by several hundred per- vaccination numbers compared with what is cent. For example, for BCG vaccinations, the recorded at the facility level. median discrepancy for overreporting facilities Notes 1 Bidani and Ravallion (1997) do find that hold-level behavior and outcomes). Such a public spending has a large effect on the linkage is, however, possible and could be health status of the poor, but only a small considered in future facility surveys. effect on the aggregate health status of the poor and the nonpoor taken together. 7 The sample design and the survey imple- mentation are discussed in greater detail in 2 The public expenditure tracking survey appendix A. See also Asiimwe (2001). (PETS) is another variant of this tool (Reinikka and Svensson 2002). 8 Given that for-profit facilities were selected on the basis of proximity to the sample of 3 Provider or facility surveys are not entirely government facilities, as described in new. The Living Standards Measurement appendix A, it is not possible to draw any Study (LSMS) surveys have included health general conclusions about their location. facility modules on an ad hoc basis (see, for example, Alderman and Lavy 1996). A 9 In many districts, donors also perform sim- number of the Demographic and Health ilar supervisory functions. Surveys (DHSs) carried out in over 50 developing countries have included service 10 This is particularly true for private facili- provider components. The Family Life Sur- ties, for which only limited information is veys implemented by the RAND Corpora- available at the district level. tion have combined health provider surveys with surveys of households. These surveys, 11 This count is based on staff member names, however, rarely collect information on pub- which is the most complete variable. Issues lic and other expenditures. For a review of relating to staffing data are discussed in health facility surveys, see Lindelöw and detail in appendix A. Wagstaff (2003). 12 The other category includes porters, watch- 4 The survey instruments can be found under men, records assistants, and sweepers. In "Tools" at . most cases no information is available about the exact positions of these staff, but 5 For a discussion of the evidence on the approximately 21 percent receive a month- divergence between local and national pri- ly wage of more than 70,000 Ugandan orities in the allocation of resources in the shillings (USh), suggesting that a propor- health sector, see Akin, Hutchinson, and tion of them may be considered "quali- Strumpf (2001). fied." 6 The facility survey was not linked to a house- 13 In cases where staff in private facilities are hold survey. Hence, the data do not permit financed by the district, records may exist. an analysis of interactions between the No records were available for staff in the demand and supply sides (for example, the sampled for-profit facilities. For private impact of facility characteristics on house- nonprofit facilities, district-level records 46 Notes 47 existed for approximately 20 percent of 22 Nineteen of the 91 facilities in the subsam- the staff. ple report receiving supplies for measles immunization days. These supplies primari- 14 The matching of facility and district records ly benefited the central region. is based on the names of health workers. 23 The figure presents data for polio and BCG 15 Exchange rate as of November 1, 2000: 1 vaccines. Similar patterns can be observed U.S. dollar (US$) = 1,770.00 Ugandan for other vaccines. shillings (USh). 24 For government facilities, most expendi- 16 Schedule B for medical personnel applies. tures on facility inputs--including the most Within each of the 10 categories of staff, important inputs, staff and drugs--are there is a gradation of salaries correspon- made at administrative levels higher than ding to the experience of the staff member. the actual facility. In general, therefore, we would not expect substantial financial 17 In the salary structure for fiscal 2000/2001 transfers to government facilities. the lunch allowance was integrated into the salary payment. This consolidation does 25 Appendix B presents a discussion of consis- not apply to nonmedical staff deployed in tency between data from the facility and the health service. district levels. 18 Private nonprofit staff financed by the dis- 26 Data for some facilities were incomplete or trict receive lunch allowances in accordance missing, typically because of poor record- with the rules that apply in government keeping. The discussion here of the num- facilities. bers of outpatients and deliveries focuses primarily on monthly averages for the 12 19 Stock cards were not generally available in months for which data are available. The private for-profit facilities. Due to incom- analysis is restricted to the subsample of pleteness or lack of records in government facilities for which data are available for at and nonprofit facilities, the averages are least 6 out of 12 months. Data issues are calculated on the basis of a subsample of discussed in further detail in appendix A. facilities--approximately 90 percent of government facilities and 45 percent of pri- 27 For the sample as a whole, the average vate nonprofit facilities. monthly number of outpatients ranges from 354 (April 2000) to 488 (November 1999). 20 There is considerable variation in reported dosage across facilities, in particular for 28 The analysis here excludes facilities for children. which no data are available or where data are available for only one month. The 21 Apparent "overuse" of drugs does not resulting subsample comprises about 80 appear to be restricted to government facil- percent of the government and private non- ities. Although some variation is to be profit facilities that perform vaccinations expected due to differences in need and to and nearly 60 percent of the for-profit facil- measurement error, the observed differences ities that do so. See appendix A and B for are greater than what can be explained by further details. these factors alone. 48 Health Care on the Frontlines: Survey Evidence on Public and Private Providers in Uganda 29 In the broadest sense, efficiency can be like. Conversely, high output per worker viewed as concerning the relationship may reflect excessive workloads rather than between inputs and outcomes. Using this "productivity." broad concept of efficiency would, however, require data that are not typically available. 36 These perception-based variables must be interpreted with caution because the crite- 30 The efficiency concepts were originally ria used in assessing care may differ in non- developed in relation to firm performance. random ways across individuals and See Fried, Lovell, and Schmidt (1993) for a because clients may fear that a negative comprehensive treatment. response may create problems for health facility staff or the respondent. 31 Where allocative efficiency in the health sector has been addressed, this has typical- 37 It should be noted that there are few cases ly been done by comparing the ratio of in which the expected positive impact of marginal products to the ratio of remuner- user fees has materialized. ation for the respective staff categories. Marginal products are calculated on the 38 Our initial analysis shows that overpre- basis of an estimated production function scription is associated with the presence of with different categories of staff among its unqualified staff and with government arguments. facilities. More detailed analysis is required to explore this linkage further (for example, 32 The weights in the output index are based to examine its association with availability on the index used in the health sector in of labs and frequency of testing). Mozambique. The general findings of this section are robust to changes in the weights 39 Forty-two percent of prescriptions included applied. chloroquine and 36 percent an antibiotic drug; 29 percent were per injection. 33 Needless to say, these findings are sensitive to the definition of output, as well as to the 40 Separate from the issue of prescription is category of workers used as the denomina- the question of patient compliance. As a tor. The findings, however, do not change consequence of low levels of education, substantively if only output per medical rushed consultations, and inappropriate worker is considered. staff attitudes, compliance can be expected to be low in developing countries, including 34 The difference in mean output per worker Uganda. between private nonprofit and government facilities is statistically significant; other dif- 41 The eight districts were Bundibugyo, Gulu, ferences in average output per worker for Kabarole, Kasese, Kibaale, Kitgum, Kotido, different ownership categories are not. and Moroto. 35 Low output per worker may be attributable 42 The study districts were Mpigi, Mukono, to shirking (low effort) by staff but also to and Masaka in the central region; Mbale, demand factors related to charging prac- Iganga, and Soroti in the east; Arua and tices, quality, the demographic and socio- Apac in the north; and Mbarara and economic characteristics of the catchment Bushenyi in the west. area, the availability of substitutes, and the Notes 49 43 Specifically, the x private facilities in region 44 The specific reasons for dropping facilities y were identified by the in-charges in the are discussed in Asiimwe (2001). first x randomly drawn government facili- ties in region y; each in-charge was asked to 45 These facilities also had other drug use identify the closest private dispensary or records that listed the quantity of drugs private dispensary with a maternity unit. bought and removed from stock but not stock balance and drug use per day. References The word "processed" describes informally nomic Research Consortium. May. reproduced works that may not be commonly Processed. available through library systems. Farrel, M. J. 1957. "The Measurement of Pro- ductive Efficiency." Journal of the Royal Akin, John, Paul Hutchinson, and Koleman Statistical Society, Series A (general), 120 Strumpf. 2001. "Decentralization and (3): 253­90. Government Provision of Public Goods: Filmer, Deon, Jeffrey S. Hammer, and Lant H. The Public Health Sector in Uganda." Pritchett. 2000. "Weak Links in the MEASURE Evaluation Working Paper Chain: A Diagnosis of Health Policy in WP-01-35. MEASURE Evaluation, Car- Poor Countries." World Bank Research olina Population Center, University of Observer 15 (2): 199­224. North Carolina, Chapel Hill. ------. 2002. "Weak Links in the Chain II: A Alderman, Harold, and Victor Lavy. 1996. Prescription for Health Policy in Poor "Household Responses to Public Health Countries." World Bank Research Services: Cost and Quality Tradeoffs." Observer 17 (1): 47­66. World Bank Research Observer 11 (1): Foster, S. 1993. "Economic Aspects of the Pro- 3­22. duction and Use of Pharmaceuticals: Evi- Asiimwe, Delius. 2001. "Cost Efficiency and dence and Gaps in Research." In Anne Incentives in Health Care Delivery in Mills and Kenneth Lee, eds., Health Eco- Uganda: Study Report." Makerere Insti- nomics Research in Developing Countries. tute of Social Research, Kampala. Oxford, U.K.: Oxford University Press. Barnum, Howard N., and Joseph Kutzin. Fried, Harold O., C. A. Knox Lovell, and Shel- 1993. Public Hospitals in Developing ton S. Schmidt. 1993. The Measurement Countries: Resource Use, Cost, Financ- of Productive Efficiency. New York: ing. Baltimore, Md.: Johns Hopkins Uni- Oxford University Press. versity Press. Gilson, L., S. Jaffar, S. Mwankusye, and T. Bidani, Benu, and Martin Ravallion. 1997. Teuscher. 1993. "Assessing Prescribing "Decomposing Social Indicators Using Practice: A Tanzanian Example." Interna- Distributional Data." Journal of Econo- tional Journal of Health Planning and metrics 77 (1): 125­39. Management 8: 37­58. Birdsall, Nancy, and Estelle James. 1993. Griffin, C. 1992. "Welfare Gains from User "Health, Government, and the Poor: The Charges for Government Health Ser- Case for the Private Sector." In James N. vices." Health Policy and Planning 7: Gribble and Samuel H. Preston, eds., The 177­80. Epidemiological Transition: Policy and Hay, Roger. 1998. "Health Services in Ugan- Planning Implications for Developing da." Report to the Africa Region, Macro- Countries. Washington, D.C.: National economics 2, World Bank, Washington, Academy of Sciences. D.C. Processed. Devarajan, Shantayanan, and Ritva Reinikka. Hutchinson, Paul. 2001. "Combating Illness." 2002. "Making Services Work for Poor In Ritva Reinikka and Paul Collier, eds., People." Presented to the African Eco- Uganda's Recovery: The Role of Farms, 50 References 51 Firms, and Government. Regional and Ofori-Adeji, David, and Daniel K. Arhinful. Sectoral Studies. Washington, D.C.: 1996. "Effect of Training on the Clinical World Bank. Management of Malaria by Medical Jitta, J. 1996. "Evaluation of User Charges in Assistants in Ghana." Social Science and Uganda." Child Health and Development Medicine 42 (8): 1169­76. Centre, Makerere University, Kampala. Okello, D. O., R. Lubanga, D. Guwatudde, Processed. and A. Sebina-Zziwa. 1998. "The Chal- Lindelöw, Magnus, and Adam Wagstaff. 2003. lenge to Restoring Basic Health Care in "Health Facility Surveys: An Introduc- Uganda." Social Science and Medicine 46 tion." Policy Research Working Paper (1): 13­21. 2953. Development Research Group, Paredes, Patricia, Manuela de la Peña, Enrique World Bank, Washington, D.C. Flores-Guerra, Judith Diaz, and James Loevinsohn, B. P., E. T. Guerrero, and S. P. Trostle. 1996. "Factors Influencing Physi- Gregorio. 1995. "Improving Primary cian's Prescribing Behaviour in the Treat- Health Care through Systematic Supervi- ment of Childhood Diarrhoea: Knowl- sion: A Controlled Field Trial." Health edge May Not Be the Clue." Social Policy and Planning 10 (2): 144­53. Science and Medicine 42: 1141­53. McPake, Barbara, Delius Asiimwe, Francis Pritchett, Lant. 1996. "Mind Your P's and Q's: Mwesigye, Mathias Ofumbi, Lisbeth The Cost of Public Investment Is Not the Orthenblad, Pieter Streefland, and Asaph Value of Public Capital." Policy Research Turinde. 1999. "Informal Economic Working Paper 1660. Policy Research Activities of Public Health Workers in Department, World Bank, Washington, Uganda: Implications for Quality and D.C. Accessibility of Care." Social Science and Reddy, Sanjay, and Jan Vandemoortele. 1996. Medicine 49 (7): 849­65. "User Financing of Basic Social Services: Ministry of Health. 1997. "Health Expendi- A Review of Theoretical Arguments and ture in Uganda." Government of Uganda, Empirical Evidence." Office of Evalua- Entebbe. tion, Policy and Planning, United Nations ------. 1998. "Health Policy Paper, Version Children's Fund (UNICEF), New York. XVII." Government of Uganda, Entebbe. Reinikka, Ritva. 2001. "Recovery in Service ------. 2000. "Health Sector Strategic Plan." Delivery: Evidence from Schools and Government of Uganda, Kampala. Health Centers." In Ritva Reinikka and Möller, L. C. 2002. "Uganda and the Millenni- Paul Collier, eds., Uganda's Recovery: um Development Goals." Human Devel- The Role of Farms, Firms, and Govern- opment Network, World Bank, Washing- ment. Regional and Sectoral Studies. ton, D.C. Processed. Washington, D.C.: World Bank. Musgrove, Philip. 1996. Public and Private Reinikka, Ritva, and Jakob Svensson. 2001. Roles in Health: Theory and Financing "Explaining Leakage of Public Funds." Patterns. World Bank Discussion Paper Policy Research Working Paper 2709. 339. Washington, D.C. Development Research Group, World Mwesigye, Francis. 1996. "Effects of User Bank, Washington, D.C. Charges on Quality of Curative Services ------. 2002. "Assessing Frontline Service in Rural Health Units in Uganda." Health Delivery." Development Research Group, Planning Department, Ministry of Health, World Bank, Washington, D.C. Entebbe. Processed. 52 Health Care on the Frontlines: Survey Evidence on Public and Private Providers in Uganda Republic of Uganda. 2000. "Tracking the Flow Fees and Insurance: Lessons from Sub- of and Accountability of UPE Funds." Saharan Africa." Africa Technical International Development Consultants, Department, World Bank, Washington, Ltd. Ministry of Education and Sports, D.C. Kampala. Vogel, Ronald J. 1991. "Cost Recovery in the ------. 2002. "Uganda Poverty Reduction Health-Care Sector in Sub-Saharan Strategy Paper Progress Report 2002." Africa." International Journal of Health Kampala. Planning and Management 6: 167­91. Rowe, Alexander K., Faustin Onikpo, Marcel World Bank. 1987. Financing Health Services Lama, François Cokou, and Michael S. in Developing Countries: An Agenda for Deming. 2001. "Management of Child- Reform. A World Bank Policy Study. hood Illness at Health Facilities in Benin: Washington, D.C. Problems and Their Causes." American ------. 1999. "Rapid Assessment of Data Journal of Public Health 91 (10): Availability in Health Care Units." World 1625­35. Bank with Makerere Institute of Social Shaw, R. Paul, and Martha Ainsworth. 1994. Research. Processed. "Financing Health Services through User