PROGRESS IN THE FACE OF INSECURITY IMPROVING HEALTH OUTCOMES IN AFGHANISTAN FULL REPORT © 2018 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this work is subject to copyright. Because The World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@ worldbank.org. Cover photo credit: © Rumi Consultancy/ World Bank. TABLE OF CONTENTS ACKNOWLEDGMENTS   v ACRONYMS   vii EXECUTIVE SUMMARY   ix INTRODUCTION AND STUDY OBJECTIVES    1 PART I STUDY METHODS   4 1 SCOPE AND METHODOLOGY   5 1.1 1.2 Study Design   5 Quantitative Analysis Methods   5 1.3 Structure and Outline   10 PART II KEY FINDINGS   12 2 HEALTH OUTCOMES, SERVICE COVERAGE AND HEALTH SYSTEMS PERFORMANCE    13 2.1 Trends in Health Outcomes in Afghanistan   13 2.2 Health Service Coverage   13 2.3 Health Systems Performance and Quality of Care   18 2.4 Is Health Facility Performance Related to Service Utilization?   21 i 3 IMPLICATIONS OF INSECURITY FOR HEALTH SERVICE COVERAGE AND HEALTH SYSTEMS PERFORMANCE   25 3.1 Insecurity Classification and Province Ranking   25 3.2 Differentials in Health Outcomes and Service Coverage by Insecurity Severity   25 3.3 Differentials in Health Facility Performance Indicators by Insecurity Severity   33 3.4 Key Section Conclusions and Considerations   40 4 IMPLICATIONS OF CONTRACTING TYPE FOR SERVICE COVERAGE AND HEALTH SYSTEMS PERFORMANCE    41 4.1 Differences in Service Coverage   41 4.2 Differences in Health Facility Performance Indicators   48 4.3 Key Section Conclusions and Considerations   49 PART III DISCUSSION AND CONCLUSIONS   52 REFERENCES   59 ii PROGRESS IN THE FACE OF INSECURITY LIST OF FIGURES FIGURE 1.  R ate of change in key health outcomes- Percentage point change per year    14 FIGURE 2.  R ates of change in service outcomes- Percentage points per year    15 FIGURE 3A.  National trends in reproductive and maternal interventions, 2003–2015    15 FIGURE 3B.  National trends in reproductive and maternal interventions, 2003–2015    16 FIGURE 3C.  National trends in childhood care-seeking interventions, 2003–2015    16 FIGURE 4A.  Composite coverage index levels and change by province, 2003 to 2010    17 FIGURE 4B.  Composite coverage index levels and change by province, 2010 to 2015    17 FIGURE 5A.  Overall province performance on CCI, 2003–2010    19 FIGURE 5B.  Overall province performance on CCI, 2010–2015    19 FIGURE 6.  National trends in health systems domains, 2004–2010    20 FIGURE 7.  National trends in health systems domains, 2011–2016    20 FIGURE 8A.  Health system overall mission scores and change by province, 2004 to 2010    22 FIGURE 8B.  Health system overall mission scores and change by province, 2011 to 2016    22 FIGURE 9A.  Provinces overall ranking in health systems performance, 2004–2010    23 FIGURE 9B.  Provinces overall ranking in health systems performance,2011–2016    23 FIGURE 10A.  Provinces by insecurity, 2003–2010    26 FIGURE 10B.  Provinces by insecurity, 2010–2015    26 FIGURE 11A.  Unadjusted change in service coverage by severity of insecurity, 2003–2010    27 FIGURE 11B.  Unadjusted change in service coverage by severity of insecurity, 2010–2015    27 FIGURE 12A.  Annual percentage point difference in service coverage by severity of insecurity, 2003–2010 (Reference: Minimal insecurity)*    32 FIGURE 12B.  Annual percentage point difference in service coverage by severity of insecurity, 2010–2015 (Reference: Minimal insecurity)*    32 FIGURE 13A.  Annual percentage point change in health systems performance by severity of insecurity, 2004–2010 (Reference: Minimal insecurity)*    39 FIGURE 13B.  Annual percentage point change in health systems performance indicators by severity of insecurity, 2011–2016 (Reference: Minimal insecurity)*    39 FIGURE 14A.  Annual percentage point difference in service coverageby type of contracting, 2003–2010 (Reference: Contracting-In)*    45 FIGURE 14B.  Annual percentage point difference in service coverage by type of contracting, 2010–2015 (Reference: Contracting-In)*    45 FIGURE 15.  Multivariable adjusted means of key RMNCH interventions by contracting type    46 FIGURE 16A.  Annual percentage point difference in health systems performance by type of contracting, 2004–2010 (Reference: Contracting-In)*    51 FIGURE 16B.  Average percentage point difference in health systems performance by type of contracting, 2011–2016 (Reference: Contracting-In)*    51 IMPROVING HEALTH OUTCOMES IN AFGHANISTAN iii LIST OF TABLES TABLE 1.  Afghanistan Health Systems Performance Domains and Indicators, 2004–2010 and 2011–2016    7 TABLE 2.  Annualized rates of change in coverage indicators-% point differences, 2003–2015    14 TABLE 3A.  RMNCH interventions coverage by insecurity status, 2003–2010    28 TABLE 3B.  RMNCH interventions coverage by insecurity status, 2010–2015    30 TABLE 4A.  Multivariable adjusted impact of insecurity on change in key RMNCH interventions    33 TABLE 4B.  Multivariable adjusted % point change in key RMNCH interventions by insecurity status, 2003–2010    34 TABLE 4C.  Multivariable adjusted % point change in key RMNCH interventions by insecurity status, 2010–2015    34 TABLE 5A.  Health systems composite indicators by insecurity status, 2004–2010    35 TABLE 5B.  Health systems composite indicators by insecurity status, 2011–2016    36 TABLE 6.  Multivariabe adjusted impact of confluct on change in key health systems indicators    38 TABLE 7A.  RMNCH interventions coverage by contracting mechanism, 2003–2010    42 TABLE 7B.  RMNCH interventions coverage by contracting mechanism, 2010–2015    43 TABLE 8A.  Multivariable adjusted impact of contracting mechanism on change in key RMNCH interventions    44 TABLE 8B.  Multivariable adjusted % point change in key RMNCH interventions by contracting group, 2003–2010    44 TABLE 8C.  Multivariable adjusted % point change in key RMNCH interventions by contracting group, 2010–2015    44 TABLE 9A.  Composite health systems indicators by contracting type, 2004–2010    49 TABLE 9B.  Composite health systems indicators by contracting type, 2011–2016    50 TABLE 10.  Multivariable adjusted impact of contracting mechanism on change in key health systems indicators    50 iv PROGRESS IN THE FACE OF INSECURITY ACKNOWLEDGMENTS T his report was prepared by the World Bank with support from a joint team from the Hospital for Sick Children Toronto, University of Toronto, the Aga Khan University and Ministry of Public Health, Afghanistan. Health development partners provided extensive comments at the protocol and report review stage. This task was led by Mickey Chopra (Lead Health Specialist) and Sayed Ghulam Dastagir Sayed (Senior Health Specialist) and included Aneesa Arur (Senior Health Specialist), Tawab Hashemi (Senior Health Specialist), Benjamin Loevinsohn (Lead Health Specialist) and Andre Medici (Senior Econo- mist). Martha P. Vargas provided valuable editorial assistance. We would like to thank Robert J. Saum (Afghanistan Country Director, 2012- 2016, World Bank), Shubham Chaudhuri (current Afghanistan Country Director, World Bank), Richard Spenser Hogg (Pro- gram Leader 2016, Afghanistan Country Management Unit), and Rekha Menon (Practice Manager for South Asia, Health, World Bank) for their support to this research. Sheryl Silverman, Raouf Zia, Aisha Faquir, Anugraha Palan, and Juhie Bhatia provided communications and outreach support. The joint team from the Hospital for Sick Children Toronto, University of Toronto and the Aga Khan University included (in alphabetical order): Nadia Akseer, Zulfiqar Bhutta, Jai Das, Zaid Bhatti, and Arjumand Rizvi. Zahra Feroz, Sehar Tejani, and Ahreen Allana assisted with transcribing interviews and group discussions. Special thanks to the Evaluation and Health Information System (EHIS) Department at the Minis- try of Public Health for organizing facilitating interviews and group discussions with key stakeholders (Dr. Sayed Ataullah Saeedzai, Husna Bakhshi, and Muzhgan Habibi); and to the French Medical Insti- tute for Mothers and Children (FMIC) Afghanistan for their assistance to University of Toronto with logistical arrangements. The team would like to acknowledge the financial support provided by the Afghanistan Reconstruc- tion Trust Fund to conduct the study. v ACRONYMS AHS Afghanistan Health Surveys ANC Antenatal Care ARI Acute Respiratory Infection ARTF Afghanistan Reconstruction Trust Fund BCG Bacillus Calmette-Guerin BHPS Basic Package of Health Services BRD Battle Related Death BSC Balance Scorecard CCI Composite Coverage Index CCT Conditional Cash Transfer CHS Community Health Supervisor CHW Community Health Worker CI Contracting-in CMW Community Midwives CO Contracting-out CPU Contraceptive Use (any method) CPUM Contraceptive Use (any modern method) DHS Demographic and Health Survey DPT Diphtheria, Pertussis, and Tetanus EHIS Evaluation and Health Information System EPHS Essential Package of Hospital Services FGD Focus Group Discussion FHA Family Health Action FHAG Family Health Action Group FHH Family Health House FMIC French Medical Institute for Mothers and Children FPS Family Planning Needs Satisfied HMIS Health Management Information Systems HP Health Practioners IDI In-depth Interview vii MCH Maternal Child Health MDG Millennium Development Goal MeSH Medical Subjects Headings MHT Mobile Health Team MICS Multiple Indicator Cluster Survey MoPH Ministry of Public Health NGO Non-Governmental Organization NMR Newborn Mortality Rates ORT Oral Rehydration Therapy PHD Provincial Health Department PNC Postnatal Case RBF Results-based financing RMNCH Reproductive, Maternal, Neonatal and Child Care Health SBA Skilled Birth Attendance TT Tetanus Toxoid UCDP Uppsala Conflict Data Programme UNDSS United Nations Department of Safety and Security UNFPA United Nations Population Fund USAID United States Agency for International Development viii PROGRESS IN THE FACE OF INSECURITY EXECUTIVE SUMMARY CONTEXT In recent years, armed insecurity in Afghanistan has intensified, putting health at greater risk. Since 2010, there has been a clear uptick in instability and insecurity in Afghanistan and an increasing share of the population lives in areas that are affected by high levels of insecurity. Maintaining service delivery and responding to intensified health needs under these circumstances is a key challenge facing Afghanistan’s health system. Scaling up service delivery strategies that are security-adaptive will be criti- cal in order to maintain and expand coverage with high quality services. Afghanistan has been at the forefront of establishing a large scale, public-private model for service delivery. All of the publicly funded health services in Afghanistan today involve some form of contracting with private entities or individuals. Two models predominate: ✦✦ Contracting-out (CO) or service delivery contracts with NGOs. Most services are deliv- ered by Non-Governmental Organizations (NGOs) under service delivery contracts with the Ministry of Public Health (MOPH). ✦✦ Contracting-in (CI) managers. In addition, in three provinces near Kabul, namely Parwan, Panjshir, and Kapisa, the MOPH contracts in managers to help strengthen services delivered using MOPH staff. After nearly a decade and a half of implementation, policy makers in Afghanistan are interested in understanding the extent to which these models are delivering good results and how current approaches can be further improved. These concerns set the stage for this study. OBJECTIVES AND METHODOLOGY This study is motivated by four main questions: ✦✦ Is the Afghan health system delivering good results to women and children in terms of service coverage and health systems performance? ix ✦✦ How has the escalating insecurity influenced these trends? ✦✦ Which model—CO or CI—delivered better results? ✦✦ How can service delivery in Afghanistan be improved and made resilient to insecurity and instability? To answer these questions, the study focuses mainly on quantitative analyses of household surveys (Multiple Indicator Cluster Surveys 2003 and 2010, as well as the Afghanistan Health Surveys 2012 and 2015) and health facility surveys (Afghanistan Balanced Scorecard Datasets, 2004–2016). In addition, sensitivity analyses were conducted with Demographic and Health Survey 2015 data. The study examines both service coverage and health systems performance results. Service cov- erage outcomes examined span the continuum of care: contraceptive coverage, antenatal care, skilled birth attendance, measles immunization, and use of oral rehydration salts for diarrhea. Health systems performance indicators examined correspond to performance domains used by the MOPH to moni- tor NGO contracts. These include client satisfaction and community involvement, human resources, physical capacity to deliver quality care, quality of service provision (i.e., process measures of quality), management systems at health facilities, and overall mission (i.e., equity of service use and in-patient satisfaction). Levels of insecurity over time were measured using battle-related deaths from the Uppsala Conflict Data Program, the international gold standard in analyses of conflict. The quantitative analyses focus on two main time windows: 2003–2010 and 2010/2011 to 2015/2016. The number of battle-related deaths in in each period was used to classify provinces into three categories: minimal, moderate, and high insecurity. Panel data linear and logistic regression methods were used to estimate the potential impact of con- tracting model; security and NGO types on changes in selected Reproductive Maternal Neonatal and Child Health (RMNCH) indicators for 2003–2010 and 2010–2015 periods; and health facility per- formance systems indicators for 2004–2010 and 2011–2016. The coverage models were adjusted for female illiteracy, percent rural population, and battle-related deaths. Health systems outcome models were adjusted for patient volumes, facility type, distance from provincial center, and region. KEY FINDINGS Afghanistan has made strong gains in health outcomes, coverage, and health systems performance All told, Afghanistan has made notable progress towards achieving the Millennium Develop- ment Goal (MDG) targets for improving maternal (MDG5) and child health (MDG4). According to United Nations (UN) estimates, maternal mortality rates (MMR) declined from 1,100 to 396 deaths per 100,000 live births from 2000 to 2015 (Alkema and others 2016). According to the latest Afghanistan Demographic Health Survey, the under 5 child mortality rate (U5MR) is now 55 per 1,000 live births, a significant reduction from the very high levels recorded in the early 2000s. Mirroring improvements in outcomes, there have been good gains in the coverage of mater- nal and child health services since 2003. Coverage of contraceptives increased until 2010 (+1.4 per- cent points per year), and declined between 2010 and 2015 (–0.5 percent points per year). All other x PROGRESS IN THE FACE OF INSECURITY maternal and child health service coverage indicators examined showed improvements across the 2000–2015 period (range: +0.3 percent points to +5.3 percent points per year), barring Tetanus Toxoid (TT) coverage, which declined over 2003–2010 and increased thereafter. Almost all provinces registered improvements in coverage between 2003 and 2015. The exceptions were in the Nimroz and Nuristan provinces during the years 2003–2010 and in Khost and Zabul over 2010–2015. Improvements achieved in health outcomes and service coverage in Afghanistan compare very favorably with improvements achieved by comparators. Afghanistan has achieved greater improvements in key maternal and child health outcomes and service coverage than regional com- parators. Improvements over time have also exceeded the global median for low and low-middle income countries over similar period of time (Akseer and others 2016; Arur and others 2011). Progress on Diphtheria Pertussis and Tetanus (DPT3) coverage in Afghanistan, however, lagged behind the global median, but not behind regional comparators. Nevertheless, in absolute terms there is considerable room for progress on both health outcomes and service coverage. Over 2004–2010, health systems performance improved considerably across the board. The rate at which individual health systems performance domains improved between 2004 and 2010 var- ies. The client and community performance domain shows the most remarkable improvements. Steep improvements were noted in physical capacity to deliver care, as well. Management systems and human resources for health domains also improved considerably. Overall mission and quality of service provi- sion remained relatively stable. Throughout 2011–2016, health systems performance continued to improve, although at a slower pace, with the exception of large improvements in physical capacity to deliver quality care. Service delivery has been resilient to insecurity Insecurity clearly presents a challenge to service delivery. In descriptive analyses, health facilities in low security provinces typically achieved greater increases in service coverage. However, improve- ments in coverage and health systems performance are apparent across the security spectrum with a few exceptions. Evidence of resilience to insecurity remains even after the analysis adjusts for confounders This holds true for service coverage as differences in improvement between higher (i.e., moder- ate and severe insecurity) and minimal insecurity provinces are small. Higher insecurity provinces show striking resilience to insecurity over 2003–2010 and made similar or statistically significantly greater gains relative to minimal insecurity provinces in improving coverage for all the measures examined, with the exception of relative progress on childhood vaccines coverage. Severe insecu- rity provinces made the largest gains in Oral Rehydration Therapy (ORT) use relative to minimal insecurity provinces (+8.2 percent points per year). During 2010–2015, higher insecurity areas achieved similar or greater relative improvements in Skilled Birth Attendance (SBA); DPT3 and measles coverage; and care seeking for ARI as compared to minimal insecurity provinces (range: no difference to +4.4 percent points per year). However, gains in contraceptive coverage, ANC, Bacillus Calmette-Guerin (BCG), and ORT coverage were statistically significantly smaller in more insecurity prone provinces as compared to minimal insecurity provinces after controlling for IMPROVING HEALTH OUTCOMES IN AFGHANISTAN xi confounders (range: -1.0 to -4.9 percent points per year), with ORT coverage most substantially affected by insecurity. Insecurity negatively impacted improvements in infrastructure, client assessment, and pro- vider knowledge in higher security facilities in 2011–2016 after adjusting for confounders (range: -1.8 to -2.8 percent points per year). There were no statistically significant differences in improvements between severe, moderate, and minimal security facilities on other health systems performance domains examined, with moderate and high security facilities achieving similar or greater improvements than minimal security facilities on functioning equipment, drugs and vaccine availability, patient counseling, and presence of a female health worker (range: no difference to +3.1 percent points per year). Contracting-in and contracting-out deliver comparable results Findings from the contracting model comparisons must take on board an important caveat. CI prov- inces are much closer to Kabul and smaller than are most CO provinces. This makes CI facilities easier to staff, supply, and manage on average than CO facilities and therefore easier to improve coverage in these provinces, as well. The analysis methods cannot control for these systematic advantages which are likely to bias findings in favor of CI provinces. CI provinces achieve greater improvements in maternal and child health coverage relative to CO provinces, but the absolute difference in improvements is small. Both CO and CI provinces achieved improvements in maternal and child health coverage over years 2003–2010, as well as over 2010–2015 with a few exceptions. Unadjusted comparisons of relative improvements in maternal and child health coverage find that CI provinces made greater gains in coverage over periods 2003–2010 and 2010–2015 as compared to CO provinces. After adjusting for confounders, CI provinces still achieved statistically significantly greater improvements on many service coverage indicators relative to CO prov- inces in over periods 2003–2010 and 2010–2015. However, the absolute difference in improvements achieved by the two approaches is relatively small, with the exception of ORT use where CI facilities achieved substantially greater improvements relative to CO facilities during 2010–2015 (range: no dif- ference to +6.9 percent points per year). The two contracting approaches also deliver similar results in terms of improvements in health systems performance, except in the case of drug availability. Adjusted comparisons of improvements over time show that CO facilities achieved similar or greater improvements in health systems performance over 2004–2010 relative to CI facilities (range: no difference to +3.1 percent points per year) with the exception of drug availability (-2.5 percent points per year). By contrast, over 2011–2016, CO facilities achieved greater improvements over time relative to CI comparators on several indicators, including functioning equipment, availability of drugs and vaccines, client physical assessment and client counseling (range: +1.6 percent points per year to +8.4 percent points per year). The availability of drugs in CO facilities increased to the greatest extent over this period. Contracting-out has performed well in high security settings and may present benefits over Contracting-in for such settings Insecurity resilience of service delivery seems to reflect Non-Governmental Organization (NGO) strategies, notably links with local communities and stakeholders. Links with local xii PROGRESS IN THE FACE OF INSECURITY communities and stakeholders were identified as a key potential drivers of insecurity resilience in ser- vice delivery in this study. NGOs recruit staff from local communities and build relationships with local powerbrokers, such strategies enabling NGOs to maintain service delivery in difficult contexts where there might be few alternative sources of medical services. This suggests that going forward there are likely benefits to embedding services closer to communities and strengthening ties with and accountability to local communities. The CO approach has clearly performed well in high and escalating security settings and NGO ability to respond quickly and with flexibility may explain good CO model performance. Since CI provinces are largely more secure than CO provinces, no evidence is currently available on the insecurity resilience of the CI model in high insecurity settings. The CO approach may have intrinsic benefits that explain these findings, notably nimble recruitment, timely salary payments, and flexibility with staff pay and flexible/ decentralized procurement. In addition, NGOs may be better able to access and deliver services in more insecure areas. International reviews of CO on the use of health services also find that the CO approach is effective in low- and middle-income countries, particularly in under-served areas and post- conflict settings. A recent systematic Cochrane (2009) review of the impact of CO in low- and middle-income countries (Lagarde and others 2009) finds that CO is an effective option particu- larly in settings where governments may have difficulties reaching populations. A literature review focused on contracting for primary care and nutrition services with broader inclusion criteria (Loevinsohn and others 2005) also concludes that successful approaches tend to maximize provider autonomy, and the review highlights that a focus on outputs and outcomes, rather than inputs, tends to lead to better results. Both reviews underscore the importance of robust evaluation and results monitoring. In light of this, it is important to shift back to true lump sum budgets for contracted NGOs. The current contracts given to NGOs are lump-sum, however, in actual fact, NGOs say they have to seek permission from the MOPH in order to transfer funds between line items, a cumbersome and time-con- suming process. This is troubling given the known benefits of provider autonomy to delivering good results, assuming that providers are held accountable for their performance, as is the case with current contracting models in Afghanistan. Substantially greater improvements in pharmaceutical and vaccine availability in CO facil- ities points to the importance of continuing decentralized procurement and supply chains. At the same time, there is a need to oversee drug quality through drug quality surveys and other approaches that independently assess whether drugs actually available at service delivery points meet quality standards. THE WAY FORWARD Effective purchasing of health services is key to delivering results: this involves a greater focus on outputs and outcomes. In general, effective purchasing of health services is more critical to delivering better health results and improving value from health spending than the question of public or private ownership of health service providers. The fundamental building block for this is the avail- ability of good performance data and purchaser capacity to use these data to better oversee provider performance. IMPROVING HEALTH OUTCOMES IN AFGHANISTAN xiii The Afghanistan health sector generates a wealth of data: these could be better used to drive performance improvements. The health sector in Afghanistan generates a wealth of data, including third party evaluation survey data and data generated by routine reporting systems. These could be used more extensively by the MOPH and Provincial Health Departments to actively drive improvements in performance in both CO and CI areas. There is also potential to expand the role of Provincial Health Departments to providing technical support to improving service delivery and decision making, rather than their more limited current focus on coordination and monitoring. In addition, the involvement of MOPH technical departments in monitoring service delivery could be strengthened. Strengthening citizen accountability and monitoring could improve both CO and CI per- formance. Findings on health systems performance improvement trends indicate that the Afghan health system has done very well on client and community engagement, and a key finding from this study is that links to communities may explain insecurity resilience. Increasing citizen involvement in monitoring service delivery may be a promising approach both to build insecurity resilience in service delivery and as a part of a broader democratic state building policy agenda. Rigorous research from other settings also points to the demonstrated value of community scorecard/citizen engagement approaches in improving service delivery (Nyqvist and others 2017). Going forward, it may be worthwhile testing innovative approaches that enable service beneficiaries to collect performance data as a complement to existing monitoring data sources, particularly in high security areas. xiv PROGRESS IN THE FACE OF INSECURITY INTRODUCTION AND STUDY OBJECTIVES A fghanistan has been at the forefront of establishing large scale public-private model for service delivery. The end of the Taliban regime in 2001 left Afghanistan with amongst the highest levels of maternal and child mortality in the world and an almost non-existent state provided health system. Much of the limited health care was being provided by NGOs, many of which were well established, but chronically underfunded. In this context, building upon this model made sense, especially with the political imperative to establish essential services across the country. Innovative approaches, such as the definition of a core basic package of health services (BPHS) at the primary level, and, later on, essential package of hospital services (EPHS) and the contracting out of the basic packages to local and international NGOs by the Ministry of Public Health (MOPH) through the use of donor resources, allowed Afghanistan to both re-establish services and to very rapidly increase access to women, children, and the poor. All of the publicly funded health services in Afghanistan today involve some form of contract- ing with private entities or individuals. Two models of contracting predominate in Afghanistan: ✦✦ Contracting-out (CO) or service delivery contracts with NGOs. Most services are delivered by Non-Governmental Organizations (NGOs) under service delivery contracts with the Ministry of Public Health (MOPH). All NGO contracts firstly focus on delivery of standard- ized packages of services defined by the MOPH; secondly, assign clear geographical respon- sibility to the NGOs (typically for whole provinces with populations ranging from about 150,000 to one million); and, thirdly, employ competitive selection of NGOs. As time has progressed, an increasing proportion of contracts have been awarded to Afghan NGOs: of 49 contracts awarded for service delivery, 72 percent are with local NGOs. ✦✦ Contracting-in (CI) managers. In addition, in three provinces near Kabul, namely Parwan, Panjshir and Kapisa, the MOPH contracts in managers to help strengthen services delivered using MOPH staff. This model involves the competitive recruitment of managers who are paid market-based salaries; a procedure for selectively increasing the salaries of MOPH health workers and field managers; provision of a level of funding similar to that provided to the NGOs; and the use of the same monitoring and evaluation mechanisms as in other provinces. Since the contracting models were introduced, Afghanistan has made significant gains in the health sector. Significant gains have been made in improving maternal, newborn, and child 1 survival; nutrition, health interventions coverage; and service availability to its populations. The large influx of financial assistance; strong local stewardship; development of sound and stable health policy frameworks; prioritization of investments in primary care and the introduction of a basic package of health services (BPHS); and essential package of hospital services (EPHS) delivered by non-governmental organizations (NGOs) have been among some of enablers of success (Akseer 2016 [1]). Afghanistan has achieved substantial improvements in health, but poor health status and services still threaten its economic development and ability to achieve the sustainable develop- ment goals (SDGs). Afghanistan has made significant progress on improving the health of its population over the last 15 years. These gains have been important of themselves, but have also contributed to the socio-economic development of the country. Healthier children will gain more from education and will turn into more productive workers. Healthier adults will suffer less absenteeism and earn higher incomes. While the improvements in health have been impressive given the context, there are some criti- cal remaining challenges. These include very high levels of malnutrition (stunting), a persistently high fertility rate, and getting basic health services to under-served rural communities. Reducing stunting, ensuring a fertility transition, and reaching the underserved remain major obstacles to achieving Afghanistan’s National Development Goals. In recent years, insecurity in Afghanistan has intensified and this poses clear risks for health. Since 2010, there has been a clear uptick in instability and insecurity in Afghanistan and an increasing share of the population lives in areas that are affected by high levels of insecurity. This trend puts health at greater risk. The direct and indirect impacts of insecurity and on health can be huge. Conflict and violence have an adverse effect on the health and well-being of populations, directly and indirectly. The collapse of health systems, deterioration of protective infrastructure, and displacement also lead to greater mor- tality and morbidity (Bhutta 2010 & 2016), in addition to threat of harm from weapons and brutal war practices resulting in death, injury, and disability. For every individual death caused by armed violence in a security zone, at least 3 to 15 other deaths occur due to malnutrition, infectious disease, or other side effect of insecurity (Save the Children 2014). The impacts of insecurity on health are also long-term. Insecurity also has long-term con- sequences with prolonged effects in the months and years following its conclusion (Bhutta 2010). The breakdown of infrastructure and institutions (including community networks, education, health services, and social welfare systems impairs coping and healing mechanisms, such as mental health wellness) further exposes individuals to long-term social, physical, and mental consequences. In fact, maternal depression, poor stimulation, unsafe learning environments, exposure to violence and trauma significantly contribute to behavioral and cognitive impairment in long-term development (Bhutta 2010 & 2016). Improving service delivery and responding to intensified health needs in a context is one of the biggest challenges facing Afghanistan’s health system. Scaling up service delivery strategies that are security adaptive will be critical to maintaining and expanding coverage with high quality services. The health services study focuses on four main questions to inform health policy in Afghanistan: a. Has the Afghan health system delivered good results in terms of Reproductive Maternal Neonatal and Child Health (RMNCH) service coverage and health systems performance? 2 PROGRESS IN THE FACE OF INSECURITY b. How has escalating insecurity influenced these trends? c. Has the CO or CI model delivered better results on service coverage and health systems performance? d. How can service delivery in Afghanistan be improved and made resilient to insecurity? The study also examines whether there is a relationship between type of NGO contracted, whether national Afghan, international, or consortium (a partnership between national and international NGO), and service coverage and health systems performance results achieved. Results on this are presented in the appendices. IMPROVING HEALTH OUTCOMES IN AFGHANISTAN 3 PART I STUDY METHODS CHAPTER SCOPE AND METHODOLOGY 1 1.1  STUDY DESIGN This study relies mainly on quantitative data analyses. These analyses focus on the 2003–2016 time period, focusing specifically on the critical transitionary windows of 2003/2004 to 2010/11 and 2010/11 to 2015/16. These time windows were selected for two main reasons: firstly, 2003/04 to 2010/11 marked the initial development phase of Afghanistan, including a strong focus from three donors (i.e., World Bank, European Union, USAID) in rapidly scaling up the BPHS (2003) and EPHS (2005) health services throughout the nation. This period also marked a comparatively “stable” security context than during the pre-2001 situation. In 2010, donor financing for health merged through the Afghanistan Reconstruction Trust Fund (ARTF) and the BPHS was revised to reflect gaps and emerg- ing health priorities. Additionally, the security context of the nation gradually deteriorated post-2010. Secondly, robust, comparable, and reliable datasets on health systems and household service coverage outcomes were available at these two-time points, as well. Interviews and focus group discussions were conducted with key stakeholders to supplement the quantitative data analyses and draw insights on how service delivery could be improved. Insights from this are integrated into this report. 1.2  QUANTITATIVE ANALYSIS METHODS Data Sources The report evaluates a range of national demographic, asset, health, and nutrition surveys conducted in Afghanistan post-2001. It focuses on the key datasets that were collected at critical time points to inform analyses and appraised to have adequate population coverage with comparable methods. This includes: Multiple Indicator Cluster Surveys (MICS) (2003, 2010/11) and the Afghan- istan Health Surveys (AHS) (2012, 2015). Sensitivity analyses were also conducted with the 2015 Demographic and Health Survey (DHS) (appendix), however, for comparability and consistency across survey estimates, inferences are derived largely from the AHS 2015. 5 Analyses also included the Afghanistan Balanced Scorecards (BSC) datasets which pro- vide comprehensive and rich information on health facilities assessment nationally and pro- vincially from 2004 to 2016. BSC indicators changed definitions frequently across sequential surveys. Generally, the 2004–2010 BSC survey definitions are comparable and 2011–2016 are alike (table 1). Analyses are presented separately for these periods. The original data was obtained and analyzed for all surveys. Indicators Reproductive, maternal, newborn and child health services coverage across the continuum of care was analyzed using the following key indicators from household surveys: Indicator Definition Current use of contraceptives Prevalence of current contraceptive use among married women 15–49 years (any method) (CPU) old, any method. Current use of contracep- Prevalence of current contraceptive use among married women 15–49 years tives (any modern method) old, any modern method. (CPUM) At least one antenatal care Percentage of women attended at least once during pregnancy by skilled health (ANC) checkup from a personnel. skilled provider 2 doses of tetanus toxoid in Percent of women who received at least two doses of tetanus-toxoid vaccine in pregnancy (TT) their last pregnancy. Skilled birth attendance (SBA) Percentage of live births attended by skilled health personnel. at last delivery Facility births Percentage of births delivered in a health facility. BCG vaccination (BCG) The percentage of children aged 12–23 months who have received at least one dose of Bacillus Calmette-Guérin vaccine. DPT3/Penta vaccination The percentage of one-year-olds who have received three doses of the com- (DPT3) bined diphtheria, tetanus toxoid, and pertussis (DTP3) vaccine in a given year. Measles vaccination (MSL) The percentage of children aged 12–23 months who have received at least one dose of measles-containing vaccine in a given year. Full immunization Percentage of children aged 12–23 months who have received at least 3 doses of DPT3, 3 doses of Polio, measles, and BCG. Care seeking for Acute Percentage of children under age five with symptoms of acute respiratory Respiratory Infection (ARI) illness (ARI) during the two-week period before the survey and, among children with symptoms of ARI, the percentage who were taken to an appropriate health provider. Oral Rehydration Therapy Percentage of children under age five ill with diarrhea during the two-week (ORT) use for diarrhea period before the survey and, among children ill with diarrhea, the percentage who received Oral Rehydration Solution (ORS) (ORS packet or pre-packaged ORS fluid). Composite Coverage Index The CCI is a weighted average of 8 core indicators: demand for family planning (CCI) satisfied, ANC 1+ visit, skilled birth attendance, BCG vaccine, 3 doses of DPT vaccine, measles vaccination, oral rehydration with continued feeding for diarrhea treatment, and care seeking for acute respiratory infection. 6 PROGRESS IN THE FACE OF INSECURITY TABLE 1.  Afghanistan Health Systems Performance Domains and Indicators, 2004–2010 and 2011–2016 AFGHANISTAN HEALTH SECTOR BPHS Balanced Scorecard 2004–2016 2004–2010 2011–2016 Domain A: Client and Community Overall Patient Satisfaction xx 1 Patient Perception of Quality Index xx Overall Client Satisfaction and Perceived Quality of Care Index xx Written Shura-e-sehie activities in community xx 2 Community Involvement and Decision-Making Index xx 3 Health Post Status Index (New) xx Domain B: Human Resources Health Worker Satisfaction Index xx 4 Revised Health Worker Satisfaction Index xx 5 Health Worker Motivation Index xx 6 Salary Payment Current xx xx Staffing Index: Meeting minimum staff guidelines xx 7 Revised Staffing Index: Meeting minimum staff guidelines xx Provider Knowledge Score xx Revised Provider Knowledge Score xx 8 Revised Provider Knowledge Score xx New Provider Knowledge Score xx Staff received training in last year xx 9 Revised Staff Received Training (in last year) xx Domain C: Physical Capacity Equipment Functionality Index xx 10 Revised Equipment Functionality Index xx Drug Availability Index xx 11 Pharmaceuticals and Vaccines Availability Index xx Laboratory Functionality Index (Hospitals & CHCs) xx 12 Laboratory Functionality Index (CHCs only) xx Clinical Guidelines Index xx 13 Revised Clinical Guidelines Index xx Infrastructure Index xx 14 Revised Infrastructure Index xx Domain D: Quality of Service Provision Patient History and Physical Exam Index xx 15 Client Background and Physical Assessment Index xx Patient Counseling Index xx 16 Client Counselling Index xx (continues on next page) IMPROVING HEALTH OUTCOMES IN AFGHANISTAN 7 TABLE 1.  Afghanistan Health Systems Performance Domains and Indicators, 2004–2010 and 2011–2016 (continued) Proper sharps disposal xx 17 Universal Precautions xx 18 Time Spent with Client xx xx Domain E: Management Systems HMIS Use Index xx 19 Revised HMIS Use Index xx 20 Financial Systems xx 21 Health Facility Management Functionality Index xx Domain F: Overall Mission Outpatient visit concentration index xx 22 New Outpatient visit concentration index xx xx Patient satisfaction concentration index xx 23 New Patient satisfaction concentration index xx xx Note: XX indicates availability of indicator in the respective time period; indicators highlighted in yellow are the same for both periods. The Composite Coverage Index (CCI) is a composite of overall health coverage and includes both curative and preventative child and maternal health interventions (Boerma and others 2008). This measure is calculated as a weighted coverage mean of eight essential interven- tions that represent broad categories of the continuum of care. The four categories are as follows: family planning, maternal and newborn care, immunization, and case management of sick children. Each continuum stage is given equal weight and the CCI is then calculated, as below. Note that FPS indicates family planning needs satisfied (related to contraceptive use) and CPNM refers to care seeking for ARI. CCI = 1 4 ( FPS + SBA + ANCS 2 DPT 3 + MSL + BCG ORT + CPNM 2 + 4 + 2 ) For health systems performance assessment, the report presents analyses of the standard com- posite domains as derived in the BSC methodology and detailed in table 1. Additionally, key com- ponent indicators were selected for detailed analyses. All domains and component indicators are scaled from 0 to 100, where higher values indicate better performance. ✦✦ Client and community ✦✦ Human resource ✦✦ Physical capacity ✦✦ Quality of service provision ✦✦ Management systems ✦✦ Ethics (after 2011) ✦✦ Overall/composite scores 8 PROGRESS IN THE FACE OF INSECURITY Descriptive Analyses and Stratifications National and provincial-level panel estimates were constructed for reproductive, maternal, neonatal, and child health (RMNCH) coverage and health facility assessment indicators to exam- ine changes over time. Performance ranking of provinces on RMNCH coverage was done based on composite coverage index improvement (percentage point increase) for two time periods between the years 2003–2010 and 2010–2015. Health systems performance ranking of provinces was based on the composite BSC score (percentage point increase) between 2004–2010 and 2011–2016. The best, mod- erate, and low-performing provinces were identified for each category based on the percentile distribu- tion and allocated in each category as follows: ✦✦ Low performing provinces: Provinces falling below 30th percentile. ✦✦ Moderate performing provinces: Provinces falling between 30th and 70th percentile ✦✦ High performing provinces: Provinces falling above 70th percentile Information was collected from MoPH on the contracts handed out to various NGOs (name of NGOs, contract dates, provinces). A number of NGOs have delivered health services in Afghanistan since 2003 and varying in size, operational capacity, and management. Analyses focused on two features of service delivery: contracting-in (CI) and contracting-out (CO), and among CO provinces on type of NGO/s contracted. The type of NGOs contracted were classified as national (Afghan NGO), international, or consortium (Afghan and international NGO jointly awarded a contract). A third stratification of interest was analysis by security/insecurity in the country. To do this, war-related casualty data (i.e., battle-related deaths) were analyzed with information from the Uppsala Conflict Data Programme (UCDP). The UCDP has recorded ongoing armed security data since 1970 and is the most commonly cited data source on global armed conflicts. UCDP defini- tions are commonly becoming a standard on how conflicts are studied and analyzed. The United Nations Department of Safety and Security (UNDSS) databases were additionally searched for province-level security ranking in Afghanistan; however, no robust and reliable time series security estimates were identified. We thus utilized available conflict data from Uppsala. Encounter-level count data on battle related deaths (BRDs) (defined as the use of armed force between warring parties in a conflict dyad, be it state-based or non-state, resulting in deaths) were obtained and totals generated for 34 provinces in Afghanistan from 2004 to 2015. Provinces were grouped into low-, moderate-, and high-intensity security zones by using the following classification for two time periods, 2003–2009 and 2010–2015: ✦✦ High-Intensity Insecurity Province (1,000 total BRD in any 3 consecutive years) ✦✦ Moderate-Intensity Insecurity Province (300–1,000 total BRD in any 3 consecutive years) ✦✦ Low-Intensity Insecurity Province (less than 300 total BRD in any 3 consecutive years) For sensitivity, the analyses also explored the BRD rates (per provincial population) as a measure of security/insecurity. War-related death rates were computed for 3- and 5-year periods and security groupings based on these were contrasted with the above based on BRD counts. Though IMPROVING HEALTH OUTCOMES IN AFGHANISTAN 9 we noted some variation in provinces’ security category, overall, these did not impact effect estimates and inferences. For stratified descriptive analyses by contracting mechanism, NGO type and insecurity, descriptive statistics were calculated including means/standard deviations, and frequencies/ proportions as appropriate. Means and mean differences were compared across subgroups using the student’s T-tests and one-way analysis of variance methods. Post-hoc comparisons were conducted with the Tukey’s multiple comparison methods, constraining Type 1 error rate at 0.05. Multivariable Analysis Methods The analyses examined the potential impact of contracting type, insecurity, and NGO types on changes in selected RMNCH indicators for 2003–2010 and 2010–2015 periods, and health facility performance systems indicators for 2004–2010 and 2011–2016 through panel data, linear and logistic regression methods. Outcomes for the coverage analysis were selected to represent various entry points on the continuum of care including contraceptive use, ante-natal care, skilled birth attendance, measles vaccination, BCG vaccination, DPT3/Penta vaccination, oral rehydration therapy, and care-seeking for ARI. For the health systems performance models, outcomes from multiple domains were selected as deemed important to the study context: these included equipment functionality, phar- maceutical and vaccine availability, functional infrastructure, client background and physical assessment, client counseling, availability of female health workers, and provider knowledge. The coverage models were adjusted for female illiteracy, % rural population and battle related deaths (BRD/10,000 popula- tion). Health systems outcome models were adjusted for patient volume, facility type, and distance from provincial center or region. It must be noted that though a range of potential confounders were identi- fied from the literature and expert opinion, the above factors were included for parsimony and to avoid over-fitting models. Additionally, model fit stats revealed that these covariates adequately represented the outcome and any important confounding by other variables of the main exposure-outcome effect. In addition, the main exposures, contract mechanism, and security level were also treated as covariates in the models where these factors were not included as the primary exposure. Generalized linear models and generalized estimating equations were used and population- averaged estimates were obtained through ‘xtreg’ and ‘xtgee’ routines in STATA for linear and logistic models, respectively. All provinces that were evaluated for at least two time points between 2003–2010 and 2010–2015 were included the analyses. Bivariate analysis was conducted to evaluate the independent effect of primary exposures on outcomes. The adjusted models were developed with time and primary exposure interaction with and without covariates to assess the impact of these exposures over time on the outcome. The results are reported as regression co-efficients with 95% confidence inter- vals. p-value less than 0.10 were considered statistically significant. 1.3  STRUCTURE AND OUTLINE This report is organized as follows: first, a brief overview is presented of the evolution of health major reforms and innovations post-2001 in Afghanistan. The next section describes trends in the coverage of key RMNCH interventions and health facility performance both at the national and 10 PROGRESS IN THE FACE OF INSECURITY the provincial level for the years 2003–2015, and an overall ranking of provinces according to perfor- mance is subsequently suggested. As a major contextual concern in Afghanistan, the effect-modifying role of security on service coverage and health systems performance is highlighted next. The report then proceeds to review the impact of the two contracting approaches (CO vs CI). Finally, we sum- marize inferences of key findings, lessons learned and implications for the future health system of Afghanistan. Analyses on the relationship between NGO type contracted (national, consortium, or international) and service coverage and health systems performance are presented in the appendices to this report. IMPROVING HEALTH OUTCOMES IN AFGHANISTAN 11 PART II KEY FINDINGS CHAPTER HEALTH OUTCOMES, SERVICE COVERAGE AND HEALTH SYSTEMS PERFORMANCE 2 2.1  TRENDS IN HEALTH OUTCOMES IN AFGHANISTAN Afghanistan has made notable progress towards achieving the Millennium Development Goal (MDG) targets for improving maternal (MDG5) and child health (MDG4). According to UN estimates, maternal mortality rates (MMR) declined from 1100 to 396 deaths per 100,000 live births from 2,000 to 2015 (Alkema and others 2016), and under 5 mortality rates (U5MR) reduced 34% (from 137 to 91 deaths per 1,000 live births), while newborn mortality rates (NMR) dropped 32% (from 53 to 36 deaths per 1,000 live births) (You and others 2016). The most recent Afghanistan Demographic and Health survey gives estimates of 55 and 22 per 1,000 live births for U5MR and NMR respectively (Afghanistan DHS 2017). Also notable is the lack of negative female gender bias in survival of newborns and children. Improvements achieved in health outcomes compare very favorably with improvements achieved in comparators. As figure 1 indicates, Afghanistan has achieved greater improvements in key maternal and child health outcomes than regional comparators. Improvements over time have also exceeded the global median for countries that started off at the same baseline levels as Afghanistan in 2003–04 (Akseer and others 2016; Arur and others 2011). Nevertheless, in absolute terms there is considerable room for progress. 2.2  HEALTH SERVICE COVERAGE National Trends in Key Maternal and Child Health Interventions Afghanistan has also made good progress on the coverage of all key maternal and child health interventions since 2003. Coverage of contraceptives increased until 2010 and declined between 2010 and 2015 (table 2, figures 3A-C). All other maternal and child health (MCH) service coverage indicators examined showed improvements across the 2010–2015 period, barring TT coverage, which declined in 2003–2010 and increased thereafter. Once again there is no significant gender differences in coverage of these critical life saving interventions. 13 FIGURE 1. Rate of change in key health outcomes, percentage point change per year U5MR Stunting –3.50 –3.00 –2.50 –2.00 –1.50 –1.00 –0.50 0.00 KPK Global Median AFG Note: U5MR = Under 5 Mortality Rate; AFG = Afghanistan; KPK = Khyber Pakhtunkhwa Province. These comparators are presented due to geographic proximity and other contextual similarities, including security constraints. Interestingly, service coverage increased at a faster pace in the 2010–2015 period for most indicators. ANC and measles immunization coverage increased at a faster pace in 2003–2010 relative to 2010–2015. TT coverage, facility deliveries, SBA, BCG, and DPT3/Penta3 coverage, as well as the Composite Coverage Index increased at a faster pace during 2010–2015. These analyses examined annual percent point changes during each of the two time periods (table 2). The pace at which service coverage improved in Afghanistan compares very favorably to com- parators. As figure 3 illustrates, the annual percent point change in maternal and child health service coverage outcomes exceeds that of regional comparators. The only exception is DPT3 coverage wherein progress in Afghanistan lagged behind the global median for the subset of countries that started at the same baseline level as Afghanistan during 2003–04. TABLE 2. Annualized Rates of Change in Coverage Indicators, % Point Differences, 2003–2015 2003–2010 % Point 2010–2015 % point Annual Rate of Change Annual Rate of Change Contraceptives (Any method) 1.4 –0.6 Contraceptives (Any modern method) 1.4 –0.5 ANC by skilled provider 4.0 2.2 TT two or more shots –1.1 1.6 Facility deliveries 2.6 3.8 SBA 3.0 3.3 BCG 0.5 3.2 DPT3/Penta3 1.3 5.3 Measles 3.9 2.5 Fully immunized — 4.8 Care seeking for ARI 1.8 0.3 ORT 3.0 2.8 Composite coverage index 1.9 2.1 2003 values for full immunization were not available for calculation of annual rates of change. 14 PROGRESS IN THE FACE OF INSECURITY FIGURE 2. Rates of change in service outcomes, percentage points per year Diarrhea (ORS) DPT3 CPR Modern ANC SBA –3.00 –2.00 –1.00 0 1.00 2.00 3.00 4.00 5.00 Baloch. KPK Global Median AFG FIGURE 3A. National trends in reproductive and maternal interventions, 2003–2015 MICS MICS AHS AHS 100 90 80 70 Coverage (%) 60 50 40 30 20 10 0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Year Contraceptives (Any method) ANC by skilled provider TT two or more shots Contraceptives (Any modern method) SBA Facility deliveries IMPROVING HEALTH OUTCOMES IN AFGHANISTAN 15 FIGURE 3B. National trends in reproductive and maternal interventions, 2003–2015 MICS MICS AHS AHS 100 90 80 70 Coverage (%) 60 50 40 30 20 10 0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Year BCG DPT3/Penta3 Measles Fully immunized FIGURE 3C. National trends in childhood care-seeking interventions, 2003–2015 MICS MICS AHS AHS 100 90 80 70 Coverage (%) 60 50 40 30 20 10 0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Year Care seeking for ARI ORT Provincial Trends in Key Maternal and Child Health Intervention There are differences in improvements in coverage over time among provinces, but almost all provinces made improvements in the composite coverage index in both time periods. We evalu- ated changes in coverage at the provincial level using the composite coverage index (figures 4A–4B, Appendix A). Overall, almost all provinces made improvements during 2003–2010 and 2010–2015. The picture differs slightly when individual coverage indicators are examined separately, although it remains a largely positive one. More results are presented in appendix A. 16 PROGRESS IN THE FACE OF INSECURITY FIGURE 4A. Composite coverage index levels and change by province, 2003–2010 Composite Coverage Index 2003–2010 *estimates coloured as follows: red = MICS 2003, blue = MICS 2010; vertical line represents percentage point change between years 100 80 60 % 40 20 0 Nimroz Nuristan Badghis Farah Jawzjan Baghlan Herat Kunduz Kabul Laghman Balkh Badakhshan Zabul Helmand Samangan Bamyan Ghazni Faryab Khost Parwan Uruzgan Daykundi Saripul Kapisa Paktiya Paktika Ghor Kandahar Nangarhar Takhar Kunar Logar Panjsher Wardak Note: A single point is presented for provinces with an estimate for only that year. FIGURE 4B. Composite coverage index levels and change by province, 2010–2015 Composite Coverage Index 2010–2015 *estimates coloured as follows: blue = MICS 2010, black = AHS 2015; vertical line represents percentage point change between years 100 90 80 70 60 % 50 40 30 20 10 0 Khost Zabul Uruzgan Kandahar Nangarhar Kabul Faryab Logar Helmand Panjsher Ghor Saripul Herat Laghman Kunar Ghazni Kunduz Paktika Badakhshan Nuristan Baghlan Samangan Farah Paktiya Balkh Bamyan Wardak Badghis Takhar Daykundi Parwan Kapisa Jawzjan Nimroz Note: A single point is presented for provinces with an estimate for only that year. IMPROVING HEALTH OUTCOMES IN AFGHANISTAN 17 Ranking of Provinces Despite improvements, the share of the Afghan population living in provinces with slow- est pace of improvements has increased over time from 23% to 39%. The CCI trends for the provinces show varying patterns for provinces with an improvement in performance for almost all provinces from 2003–2015 (figures 5A–5B). The exceptions are Nimroz and Nuristan during the years 2003–2010 and Khost and Zabul in 2010–2015. Figures 5A–5B show the provinces which were identified as high, moderate, or minimal performing based on the relative ranking. For the period 2003–2010, the high-performing provinces were Faryab, Ghazni, Khost, Kunar, Logar, Paktika, Panjsher, Parwan, Saripul, and Uruzgan; while for 2010–2015, the high-performing provinces based on CCI were Badghis, Bamyan, Daykundi, Jawzjan, Kapisa, Nimroz, Paktiya, Parwan, Takhar, and Wardak. Parwan was the only high-performing province during the two time periods. Population distribution in the three health service coverage performance categories during 2003–2010 were 23% (minimal), 53% (moderate), and 23% (high), and over 2010–2015 were 39% (minimal), 42% (moderate), and 20% (high). HEALTH SYSTEMS PERFORMANCE 2.3  AND QUALITY OF CARE National Trends Figures 6 and 7 present health facilities performance domains in Afghanistan from 2004 to 2016. Domain definitions and component variables were modified post-2010 to make the balanced score card tougher in view of improvements in facility performance. Therefore, trends are contrasted separately for the 2004–2010 and 2011–2016 time periods in these analyses. Over 2004–2010, most health systems performance domains showed considerable improve- ments. Overall, health systems broad domains have changed variably in Afghanistan from 2004 to 2010 (figure 6). The nation seems to be consistently performing best in the client and community area with scores as high as 80/100. Steep improvements were noted for physical capacity, which moved from a low score of 44 to 74 from 2004 to 2010. Scores for management systems and human resources were improving and well above 70/100 before 2010, while overall mission and quality of service provision were relatively unchanged.1 2011–2016 onwards, health systems performance continued to improve, although at a slower pace, with the exception of large improvements in physical capacity to deliver high quality care. As figure 7 shows, while health systems performance continued to improve over 2011–2016, the pace of improvements is slower. The main exception is physical capacity to deliver quality care which continued to improve at a fast pace. Another important development throughout this period has been the steady increase in female health workers, increasing from fewer than 600 in 2002 to around 2,500 in 2014 and almost 90% of health facilities with at least one female health worker. Please refer to table 1 for more details on each health systems domain. 1 18 PROGRESS IN THE FACE OF INSECURITY FIGURE 5A. Overall province performance on CCI, 2003–2010 Jawzjan Balkh Kunduz NORTH EAST Badakhshan Takhar NORTH Samangan Faryab Baghlan Sari Pul Panjsher Nuristan Badghis Bamyan Parwan Kapisa EAST Kunar CENTRAL Laghman Kubal CENTRAL Wardak Nangarhar Herat Ghor HIGHLAND Logar WEST Daykundi Ghazni Paktiya Khost SOUTH EAST Uruzgan Paktika Farah Zobel LEGEND SOUTH Region Kandahar Hilmand Ranking Nimroz High Moderate Minimal FIGURE 5B. Overall province performance on CCI, 2010–2015 Jawzjan Balkh Kunduz NORTH EAST Takhar Badakhshan NORTH Samangan Faryab Baghlan Sari Pul Panjsher Nuristan Badghis Bamyan Parwan Kapisa EAST Kunar CENTRAL Laghman Kubal CENTRAL Wardak Nangarhar Herat Ghor HIGHLAND Logar WEST Daykundi Ghazni Paktiya Khost SOUTH EAST Uruzgan Paktika Farah Zobel LEGEND SOUTH Region Kandahar Nimroz Hilmand Ranking High Moderate Minimal IMPROVING HEALTH OUTCOMES IN AFGHANISTAN 19 FIGURE 6. National trends in health systems domains from 2004–2010 100 90 80 70 60 Score/100 50 40 30 20 10 0 2004 2005 2006 2007 2008 2009/10 Year Domain A: Client and Community Domain C: Physical Capacity Domain E: Management Systems Domain B: Human Resources Domain D: Quality of Service Provision Domain F: Overall Mission Average composite score (Median) Note: Trends are interpreted separately for 2004–2010 and 2011–2016 time periods due to a change in domain definitions after 2010. FIGURE 7. National trends in health systems domains from 2011–2016 100 90 80 70 60 Score/100 50 40 30 20 10 0 2011/12 2012/13 2015 2016 Year Domain A: Client and Community Domain C: Physical Capacity Domain E: Management Systems Domain B: Human Resources Domain D: Quality of Service Provision Domain F: Overall Mission Average composite score (Median) Note: Trends are interpreted separately for 2004–2010 and 2011–2016 time periods due to a change in domain definitions after 2010. 20 PROGRESS IN THE FACE OF INSECURITY Provincial Trends All provinces, except Zabul, made improvements in health systems domains during 2004– 2010, while all provinces, except Kapisa, Kunar and Badghis, made health systems improve- ments during 2011–2016. During 2004–10, most provinces showed improvement in all the six domains evaluated (see supplementary figures in appendix). Client and community showed improve- ment in the range of 3.8% to 44.2% across provinces, physical capacity from 3.8% to 44.2%, human resource from –12.5% to 40.9%, quality of service provision from –24.6% to 25.1%, management sys- tems from –18.9% to 54%. During 2011–2016, client and community showed improvement in the range of –10.3% to 17.9% between provinces, physical capacity from 5.4% to 52.2%, human resource from –12.7% to 19.1%, quality of service provision from –17.9% to 29.2%, management systems from –21.1% to 27.5%. Figures 8A and 8B show the overall improvement during the two time periods. With the exception of Zabul, all provinces showed improvements ranging from 1.3% to 25.5% in 2004–2010. In 2011–2016, all except three (Kapisa, Kunar, Badghis) showed improvement which ranged from 0.1% to 48.3% across provinces. Overall, the improvements were greater over the 2004–2010 period than during the 2011–2016 period, as indicated by steeper slopes and greater absolute score gains within the earlier period. It should be noted that there are some provinces that saw worsening performance in some specific areas/components, but these disappear in the aggregate. Ranking of Provinces The best performing provinces in terms of health systems performance varied slightly from the CCI service utilization ranking: the share of the Afghan population living in mini- mal improvement provinces increased from 37% to 45% between 2004–2010 and 2011–2016. Composite health systems scores were used to further rank provinces based on overall performance (figures 9A–9B). Health facilities are functioning best in Baghlan, Faryab, Herat, Jawzjan, Khost, Kunar, Laghman, Logar, Nuristan, and Paktika in 2003–2010; while for the period 2011–2016, the high performing provinces are Badakhshan, Balkh, Daykundi, Farah, Faryab, Helmand, Nangarhar, Paktiya, Saripul, and Zabul. Faryab was the only high-performing province during the two time peri- ods. Population distribution in the three health systems performance categories over 2004–2010 were 37% (minimal), 38% (moderate) and 25% (high); and over 2011–2016 were 45% (minimal), 26% (moderate) and 30% (high). IS HEALTH FACILITY PERFORMANCE 2.4  RELATED TO SERVICE UTILIZATION? Service utilization is not consistently correlated with health facility performance with the exception of care-seeking for ARI. In evaluating the performance of the provinces for both the service coverage and facility performance, during the 2003–2010 period, five provinces (Faryab, Khost, Kunar, Logar) were rated as best performing in both areas, while two provinces (Daykundi and Paktiya) were rated as best performing in both areas during 2010–2015. We examined correlations between coverage IMPROVING HEALTH OUTCOMES IN AFGHANISTAN 21 FIGURE 8A. Health system overall mission scores and change by province, 2004 to 2010 Overall Performance (2004–2010) *estimates coloured as follows: red = 2004 blue = 2010; vertical line represents percentage point change between years 100 90 80 70 60 % 50 40 30 20 10 0 Zabul Badakhshan Bamyan Helmand Urozgan Kabul Ghazni Sar-E-Pul Kunduz Balkh Samangan Badghis Nimroz Farah Parwan Herat Nooristan Jawzjan Faryab Khost Laghman Baghlan Daykundi Paktya Kapisa Kunarha Paktika Panjsher Kandahar Takhar Ghor Nangarhar Logar Wardak Note: A single point is presented for provinces with an estimate for only that year. FIGURE 8B. Health system overall mission scores and change by province, 2011 to 2016 Overall Performance (2011–2016) *estimates coloured as follows: blue = 2011 black = 2016; vertical line represents percentage point change between years 100 90 80 70 60 % 50 40 30 20 10 0 Badghis Uruzgan Parwan Laghman Kundiz Jawzjan Balkh Helmand Khost Bamyan Ghazni Nimroz Farah Samangan Daykundi Kabul Badakhshan Faryab Zabul Sar-E-Pul Baghlan Herat Nuristan Kapisa Paktya Paktika Kunar Kandahar Ghor Panjsher Nangarhar Logar Takhar Wardak Note: A single point is presented for provinces with an estimate for only that year. 22 PROGRESS IN THE FACE OF INSECURITY FIGURE 9A. Provinces overall ranking in health systems performance, 2004–2010 Jawzjan Balkh Kunduz NORTH EAST Badakhshan Takhar NORTH Samangan Faryab Baghlan Sari Pul Panjsher Nuristan Badghis Bamyan Parwan Kapisa EAST Kunar CENTRAL Laghman Kubal CENTRAL Wardak Nangarhar Herat Ghor HIGHLAND Logar WEST Daykundi Ghazni Paktiya Khost SOUTH EAST Uruzgan Paktika Farah Zobel LEGEND SOUTH Region Kandahar Hilmand Ranking Nimroz High Moderate Minimal FIGURE 9B. Provinces overall ranking in health systems performance, 2011–2016 Jawzjan Balkh Kunduz NORTH EAST Badakhshan Takhar NORTH Samangan Faryab Baghlan Sari Pul Panjsher Nuristan Badghis Bamyan Parwan Kapisa EAST Kunar CENTRAL Laghman Kubal CENTRAL Wardak Nangarhar Herat Ghor HIGHLAND Logar WEST Daykundi Ghazni Paktiya Khost SOUTH EAST Uruzgan Paktika Farah Zobel LEGEND SOUTH Region Kandahar Hilmand Ranking Nimroz High Moderate Minimal IMPROVING HEALTH OUTCOMES IN AFGHANISTAN 23 of SBA, measles, care seeking for ARI and CCI and health facility indicators in 2004, 2011, and 2016 (appendix B). In 2004, there was no significant correlation between coverage and health facility perfor- mance indicators. In 2011, increased care seeking for ARI among children was moderately correlated to the female health worker index (Pearson r = 0.387, p = 0.031). In 2015/16, increased ARI care seeking was positively correlated with female health worker index (r = 0.38, p = 0.031), provider knowledge score (r = 0.457, p = 0.01), and borderline significantly with equipment functionality (r = 0.332, p = 0.068) and functional infrastructure (r = 0.333, p = 0.067). The CCI was also positively associated with pharma- ceuticals and vaccines availability (r=0.372, p = 0.043), while higher SBA was linked to higher functional infrastructure (r = 0.387, p = 0.034). 24 PROGRESS IN THE FACE OF INSECURITY CHAPTER 3 IMPLICATIONS OF INSECURITY FOR HEALTH SERVICE COVERAGE AND HEALTH SYSTEMS PERFORMANCE I n this section, we explore the extent to which escalating insecurity explains observed trends. Data sources, security indicators/classifications, and methods are detailed in the methodology section, above. 3.1 SECURITY CLASSIFICATION AND PROVINCE RANKING To assess the relationship of security with key indicators, we relied on battle-related death data from Uppsala databases, as described in the methodology. Annual battle-related deaths for each province are shown in appendix C. We divided the provinces as low-, moderate-, and high-intensity secu- rity zones based on the criteria previously specified (figures 10A-10B). There has been an escalation in insecurity in Afghanistan and the share of the population living in severely insecure areas has increased substantially over time. Population distribution across three insecure zones during 2003–2010 were as follows: 50% (minimal), 33% (moderate) and 17% (severe); and during 2010–2015, 16% (minimal), 44% (moderate) and 40% (severe). Many provinces moved from low to moderate and from moderate to severe insecurity during the time periods 2003–2010 to 2010–2015. According to our classification, Zabul was the only province to show improvement, as it moved from the high- to moderate-intensity group. No province under the CI approach was classified as severely insecure for the two time periods (identified with asterisks on the maps), given their proximity to Kabul. Parwan and Panjsher were classified as regions with minimal insecurity during the two time periods, while Kapisa moved from minimal intensity in 2003–2010 to moderate intensity over 2010–2015. DIFFERENTIALS IN HEALTH OUTCOMES 3.2  AND SERVICE COVERAGE BY INSECURITY An analysis of the RMNCH indicators to evaluate whether there is a relationship between insecurity with coverage of essential interventions is presented in tables 3A and 3B, respec- tively, for the 2003–2010 and 2010–2015 time periods. Results suggest that service delivery has been resilient to insecurity. 25 FIGURE 10A. Provinces by insecurity, 2003–2010 Jawzjan Balkh Kunduz Badakhshan Takhar Samangan Faryab Baghlan Sari Pul Panjsher Nuristan Badghis Bamyan Parwan Kapisa Kunar Laghman Kubal Wardak Herat Nangarhar Ghor Logar Daykundi Ghazni Paktiya Khost Uruzgan Paktika Farah LEGEND Zobel Insecurity (Year 2003–08) Kandahar Minimal Hilmand Nimroz Moderate Severe FIGURE 10B. Provinces by insecurity, 2010–2015 Jawzjan Balkh Kunduz Badakhshan Takhar Samangan Faryab Baghlan Sari Pul Panjsher Nuristan Badghis Bamyan Parwan Kapisa Kunar Laghman Kabul Wardak Hirat Nangarhar Ghor Logar Daykundi Ghazni Paktya Khost Uruzgan Paktika Farah Zabul LEGEND Insecurity (Year 2009–15) Kandahar Low intensity Hilmand Nimroz Moderate Severe 26 IMPROVING ACCESS TO AND QUALITY OF HEALTH SERVICES FIGURE 11A. Unadjusted change in service coverage by severity of insecurity, 2003–2010 70 60 56.8 52.5 50 40 37 30.5 30 26.3 26.1 25.7 24.5 20.3 19.3 20.4 20 18.3 17 15 12.2 12 11.3 7.4 8.4 10 3.8 0 –10 – 4.7 –7.7 –8.2 –20 –14.6 Contraceptive ANC SBA BCG DPT3/Penta Measles ORT Care seeking Use (Modern) for ARI Minimal insecurity Moderate insecurity High insecurity FIGURE 11B. Unadjusted change in service coverage by severity of insecurity, 2010–2015 80 68.7 60 53.1 50.5 48.7 42.3 36.2 40 29.2 31.4 26.6 20.2 20 14.5 12.6 11.4 10.6 7.5 10.8 3.7 6.5 1.3 0.1 1 0 –2.2 –20 –12.3 –27.3 –40 Contraceptive ANC SBA BCG DPT3/Penta Measles ORT Care seeking Use (Modern) for ARI Minimal insecurity Moderate insecurity High insecurity THE AFGHANISTAN HEALTH SERVICES STUDY 27 Higher insecurity provinces achieved improvements in service coverage in during the time periods 2003–2010 and 2010–2015. Severe and moderate insecurity provinces achieved increases in service coverage on most indicators, with the exception of immunization coverage which declined over 2003–2010. During 2010–2015, ORT coverage declined substantially in severe and moder- ate insecurity provinces as did contraceptive use, with the latter declining only in severe insecurity provinces. The pace of improvements varied over time and by indicator. During 2003–2010, severe inse- curity provinces achieved statistically significantly greater improvements in contraceptive coverage (p < 0.01) and ORT use (p < 0.01) than did minimal insecurity provinces. Moderate insecurity provinces achieved statistically significantly greater improvements in BCG (p < 0.01) and measles coverage (p < 0.01) relative to minimal insecurity provinces. On the remaining indicators, provinces with minimal insecurity achieved similar (p > 0.10) or greater increases than did provinces with moderate or severe insecurity. Over 2010–2015, severe insecurity provinces achieved statistically significantly greater improvements in DPT3 coverage (p < 0.01) compared to minimal insecurity provinces. On all other coverage indicators presented here, provinces with severe and moderate insecurity achieved fewer or statistically similar improvements (p > 0.10) in coverage compared to provinces with minimal inse- curity (see tables 3A and 3B). TABLE 3A. RMNCH Interventions Coverage by Insecurity Status, 2003–2010 Minimal vs. Minimal Minimal Moderate Severe Moderate vs. Severe Indicator Survey N Insecurity N Insecurity N Insecurity p-value p-value Mean% 12,048 5.2 5,654 20.3 3,260 6.6 0.000 0.013 2003 Contraceptives Mean% 10,042 12.7 5,801 22.8 2,231 25.1 0.000 0.000 (Any method) 2010 Mean 7.5 2.5 18.5 0.000 0.000 Difference Mean% 12,033 4.4 5,647 16.2 3,255 6.3 0.000 0.001 2003 Contraceptives Mean% 10,042 11.8 5,801 20.1 2,231 24.6 0.000 0.000 (Any modern method) 2010 Mean 7.4 3.8 18.3 0.000 0.000 Difference Mean% 6,295 8.7 2,890 31.8 1,702 10.5 0.000 0.330 2003 ANC by skilled Mean% 2,714 45.7 1,808 56.3 352 36.8 0.007 0.068 provider 2010 Mean 37.0 24.5 26.3 0.000 0.031 Difference 28 IMPROVING ACCESS TO AND QUALITY OF HEALTH SERVICES TABLE 3A. RMNCH Interventions Coverage by Insecurity Status, 2003–2010 (continued) Minimal vs. Minimal Minimal Moderate Severe Moderate vs. Severe Indicator Survey N Insecurity N Insecurity N Insecurity p-value p-value Mean% 6,370 5.4 2,919 28.3 1,719 7.7 0.000 0.011 2003 Mean% Facility deliveries 2,786 27.3 1,820 45.3 356 15.3 0.000 0.013 2010 Mean 21.9 17.0 7.6 0.000 0.001 Difference Mean% 6,380 7.4 2,915 29.5 1,720 8.3 0.000 0.384 2003 Mean% SBA 2,786 33.4 1,820 49.9 356 23.3 0.000 0.037 2010 Mean 26.1 20.3 15.0 0.000 0.021 Difference Mean% 2,505 41.6 1,283 19.2 857 49.3 0.000 0.010 2003 Mean% BCG 1,429 60.9 816 71.8 247 34.8 0.026 0.000 2010 Mean 19.3 52.5 –14.6 0.000 0.000 Difference Mean% 2,504 23.3 1,283 45.5 857 11.7 0.000 0.000 2003 Mean% DPT3/Penta3 1,400 31.8 793 37.8 240 7.0 0.168 0.000 2010 Mean 8.4 –7.7 –4.7 0.001 0.007 Difference Mean% 2,516 27.7 1,285 17.2 851 28.6 0.001 0.733 2003 Mean% Measles 1,413 39.9 817 47.7 244 20.4 0.076 0.003 2010 Mean 12.2 30.5 -8.2 0.000 0.006 Difference Mean% 3,110 45.0 1,121 54.4 487 36.2 0.004 0.025 2003 Mean% Care seeking for ARI 1,696 57.0 1,041 65.7 212 61.8 0.020 0.546 2010 Mean 12.0 11.3 25.7 0.967 0.126 Difference Mean% 4,800 25.9 2,071 41.6 865 17.2 0.000 0.007 2003 Mean% ORT 2,059 46.3 1,044 58.6 337 74.0 0.034 0.000 2010 Mean 20.4 17.0 56.8 0.400 0.000 Difference THE AFGHANISTAN HEALTH SERVICES STUDY 29 TABLE 3B. RMNCH Interventions Coverage by Insecurity Status, 2010–2015 Minimal vs. Minimal Minimal Moderate Severe Moderate vs. Severe Indicator Survey N Insecurity N Insecurity N Insecurity p-value p-value Mean% 4,426 14.1 7,329 16.4 6,319 20.7 0.099 0.000 2010 Contraceptives Mean% 4,880 15.2 12,013 16.9 7,828 20.0 0.213 0.002 (Any method) 2015 Mean 1.1 0.5 –0.7 0.737 0.349 Difference Mean% 4,426 12.7 7,329 14.4 6,319 19.8 0.212 0.000 2010 Contraceptives Mean% 4,880 14.0 12,013 14.5 7,828 17.6 0.685 0.008 (Any modern method) 2015 Mean 1.3 0.1 –2.2 0.520 0.082 Difference Mean% 1,223 50.2 1,954 49.9 1,697 46.7 0.933 0.350 2010 ANC by skilled Mean% 1,636 79.5 3,915 60.5 2,400 50.4 0.000 0.000 provider 2015 Mean 29.2 10.6 3.7 0.000 0.000 Difference Mean% 1,228 32.3 1,974 37.6 1,760 27.3 0.195 0.195 2010 Mean% Facility deliveries 1,740 59.6 4,185 55.4 2,495 50.2 0.990 0.025 2015 Mean 27.3 17.8 22.9 0.081 0.536 Differences Mean% 1,228 36.1 1,974 44.7 1,760 32.6 0.042 0.395 2010 Mean% SBA 1,770 56.3 4,305 52.2 2,606 47.1 0.957 0.017 2015 Mean 20.2 7.5 14.5 0.018 0.344 Differences Mean% 1,429 60.9 816 71.8 247 34.8 0.026 0.000 2010 Mean% BCG 818 92.3 2,002 84.3 1,166 83.5 0.001 0.000 2015 Mean 31.4 12.6 48.7 0.000 0.566 Differences Mean% 1,400 31.8 793 37.8 240 7.0 0.168 0.000 2010 Mean% DPT3/Penta3 818 84.9 2,002 74.0 1,166 75.7 0.000 0.002 2015 Mean 53.1 36.2 68.7 0.000 0.002 Differences 30 IMPROVING ACCESS TO AND QUALITY OF HEALTH SERVICES TABLE 3B. RMNCH Interventions Coverage by Insecurity Status, 2010–2015 (continued) Minimal vs. Minimal Minimal Moderate Severe Moderate vs. Severe Indicator Survey N Insecurity N Insecurity N Insecurity p-value p-value Mean% 1,413 39.9 817 47.7 244 20.4 0.076 0.003 2010 Mean% Measles 818 82.2 2,002 74.2 1,166 70.9 0.008 0.000 2015 Mean 42.3 26.6 50.5 0.002 0.388 Difference Mean% 1,696 57.0 1,041 65.7 212 61.8 0.020 0.546 2010 Mean% Care seeking for ARI 520 68.4 1,388 72.2 542 72.6 0.318 0.378 2015 Mean 11.4 6.5 10.8 0.449 0.992 Difference Mean% 2,059 46.3 1,044 58.6 337 74.0 0.034 0.000 2010 Mean% ORT 891 47.4 1,813 46.2 863 46.7 0.766 0.875 2015 Mean 1.0 –12.3 -27.3 0.052 0.000 Difference After controlling for confounders (maternal literacy, contracting type and rural resi- dence), this finding of insecurity resilience persists and differences in improvement between higher and minimal insecurity provinces are small. Higher insecurity provinces show strik- ing resilience to insecurity in 2003–2010 and make statistically significantly (p < 0.10) greater gains relative to low insecurity provinces in improving coverage of contraceptive use, Ante Natal Care (ANC), Skilled Birth Attendance (SBA), care seeking for ARI (Acute Respiratory Infec- tions) and ORT (Oral Rehydration Therapy) use during this period after adjusting for maternal literacy, contracting type and rural residence. However, relative progress in scaling up childhood vaccines coverage was lacking in the high insecurity provinces (figure 12A; table 4A, 4B). During 2010–2015, gains in contraceptive coverage, ANC, Bacillus Calmette-Guerin (BCG), and ORT were statistically significantly smaller in more insecurity prone provinces compared to min- imal insecurity provinces after controlling for confounders (figure 12B; tables 4A, 4C). However, higher insecurity areas achieved similar relative improvements in SBA, DPT3, and measles coverage and care seeking for ARI as compared to minimal insecurity provinces. ORT coverage was most affected by insecurity as improvements in coverage for this indicator were 4.9 percent points lower in severe insecurity provinces compared to minimal insecurity provinces over a period of 5 years or about 1 percent point per year (see figures 12 A-B, tables 4A-C). Crude box-plots, complete regres- sion models, and covariate-adjusted prevalence plots for each indicator over time are included in appendix D. THE AFGHANISTAN HEALTH SERVICES STUDY 31 FIGURE 12A. Annual percentage point difference in service coverage by severity of insecurity, 2003–2010 (Reference: Minimal insecurity)* 10 8.2 8 6 4.1 4 3.5 3.0 3.1 2.9 2.8 2.1 2.1 2 1.5 1.5 0.5 0.2 0 –0.7 –1.2 –2 –2.2 –4 Contraceptive ANC SBA BCG DPT3/Penta Measles ORT Care seeking Use for ARI Moderate versus Minimal insecurity Severe versus Minimal insecurity Note: *Red values indicate statistical significance at p = 0.10 or less. Model adjusted for the main effects of insecurity, time, and covariates including maternal illiteracy, contracting type, and rural residence. Positive numbers mean that provinces with severe or moderate insecurity achieved greater improvements than did provinces with minimal insecurity. FIGURE 12B. Annual percentage point difference in service coverage between by severity of insecurity, 2010–2015 (Reference: Minimal insecurity)* 6 4.4 4 2.4 2 1.8 0.2 0.0 0 –0.9 –1.0 – 0.4 –1 –2 –1.1 –1.8 –1.6 –1.7 –4 –3.1 –3.1 –6 – 4.9 Contraceptive ANC SBA BCG DPT3/Penta Measles ORT Care seeking Use for ARI Moderate versus Minimal insecurity Severe versus Minimal insecurity Note: Red values indicate statistical significance at p = 0.10 or less. Model adjusted for the main effects of insecurity, time, and covariates, including maternal illiteracy, contracting type, and rural residence. Positive numbers mean that provinces with severe or moderate insecurity achieved greater improvements than did provinces with minimal insecurity. 32 IMPROVING ACCESS TO AND QUALITY OF HEALTH SERVICES TABLE 4A. Multivariable Adjusted Impact of Insecurity on Change in Key RMNCH Interventions Insecurity* time Interaction MICS 2003–MICS 2010 MICS 2010–AHS 2015 [Reference = Minimal] Coef. 95% CI p-value Coef. 95% CI p-value Contraceptive Use: Moderate insecurity* time 1.03 (0.73, 1.32) 0.000 –0.48 (–0.73, –0.22) 0.000 Contraceptive Use: Severe insecurity* time 1.22 (0.74, 1.69) 0.000 –0.67 (–0.99, –0.34) 0.000 ANC: Moderate insecurity* time 0.43 (0.06, 0.8) 0.024 –0.87 (–1.22, –0.51) 0.000 ANC: Severe insecurity* time 0.66 (–0.05, 1.37) 0.070 –0.86 (–1.34, –0.37) 0.001 SBA: Moderate insecurity* time 0.53 (0.2, 0.86) 0.002 –0.25 (–0.59, 0.09) 0.149 SBA: Severe insecurity* time 0.43 (–0.38, 1.24) 0.294 0.08 (–0.52, 0.68) 0.796 BCG: Moderate insecurity* time 0.05 (–0.45, 0.55) 0.851 –0.46 (–1, 0.08) 0.096 BCG: Severe insecurity* time –1.33 (–2.17, -0.49) 0.002 0.46 (–0.35, 1.28) 0.261 DPT3/Penta: Moderate insecurity* time –0.07 (–0.51, 0.38) 0.769 –0.22 (–0.74, 0.31) 0.415 DPT3/Penta: Severe insecurity* time –2.11 (–3.05, –1.17) 0.000 1.79 (0.94, 2.64) 0.000 Measles: Moderate insecurity* time 0.47 (0.01, 0.94) 0.044 –0.07 (–0.58, 0.43) 0.779 Measles: Severe insecurity* time –1.15 (–2.03, –0.27) 0.010 0.41 (–0.43, 1.24) 0.342 ORT: Moderate insecurity* time –0.43 (–0.97, 0.1) 0.111 –0.39 (–0.86, 0.08) 0.101 ORT: Severe insecurity* time 2.78 (2.06, 3.5) 0.000 –1.27 (–1.9, –0.64) 0.000 Care seeking ARI: Moderate insecurity* time 0.59 (0.16, 1.02) 0.008 –0.36 (–0.8 0.09) 0.118 Care seeking ARI: Severe insecurity* time –0.30 (–1.29, 0.69) 0.555 –0.23 (–1.22, 0.76) 0.651 *Models adjusted for the main effects of insecurity and time, and covariates including maternal illiteracy, contracting type and rural residence; complete results in the appendix. DIFFERENTIALS IN HEALTH FACILITY PERFORMANCE 3.3  INDICATORS BY INSECURITY SEVERITY Health systems performance improvements have been achieved across the insecurity spec- trum; while facilities in higher insecurity provinces achieve improvements over time in health systems performance in unadjusted comparisons of changes over time. During 2003–2010, with the exception of quality of service provision (process quality) in severe insecurity provinces, severe, and moderate insecurity provinces achieved improvements on all health systems performance domains examined (table 5A). During 2011–2016, with the exception of the overall mission domain (equity) in moderate insecurity provinces, severe and moderate insecurity provinces achieved improvements in all other health systems performance domains (table 5B). Crude box-plots, com- plete regression models, and covariate-adjusted mean plots for each indicator over time are included in appendix E. Although facilities in severe insecurity provinces achieved improvements in 2003–2010, the rate of (unadjusted) improvement is smaller than in minimal insecurity provinces on some health systems performance indicators. Severe insecurity provinces made smaller improvements than min- imal insecurity provinces on client and community (p < 0.10) and physical capacity (p < 0.01). The average facility in a severe insecurity area has a statistically significantly lower quality score than in a min- imal insecurity facility (9 percentage points lower over 2003–2010, p = 0.012). Facilities in moderate insecurity provinces achieved similar levels of improvement (p > 0.10) over time as did minimal insecurity facilities in the same time period. THE AFGHANISTAN HEALTH SERVICES STUDY 33 34 TABLE 4B. Multivariable Adjusted % Point Change in Key RMNCH Interventions by Insecurity Status, 2003–2010 Minimal Moderate Severe MICS MICS 2003– MICS MICS 2003– MICS MICS 2003– 2003 2010 2010 2003 2010 2010 2003 2010 2010 Average Average Average Minimal vs. Minimal IMPROVING ACCESS TO AND QUALITY OF HEALTH SERVICES Adjusted Adjusted % Point Adjusted Adjusted % Point Adjusted Adjusted % Point Moderate vs. Severe % % Change % % Change % % Change p-value p-value Contraceptive Use 11.02 12.8 0.1  7.12 19.77 1.6  8.41 26.07 2.2 0.000 0.000 ANC 17.96 46.84 0.3 14.46 49.32 4.4  8.47 38.94 3.8 0.024 0.070 SBA 16.12 35.91 0.3 12.54 39.29 3.3  8.25 27.36 2.4 0.002 0.294 BCG 33.93 59.97 0.2 42.73 69.17 3.3 38.16 34.12 –0.5 0.851 0.002 DPT3/Penta 29.41 37.47 0.2 23.87 29.81 0.7 21.75  5.44 –2.0 0.769 0.000 Measles 25.49 40.32 0.2 21.05 45.48 3.1 28.85 20.7 –1.0 0.044 0.010 ORT 27.24 49.29 0.2 40.25 53.91 1.7  7.52 74.87 8.4 0.111 0.000 Care seeking for ARI 47.49 56.84 0.1 40.45 63.98 2.9 54.1 56.1 0.3 0.008 0.555 TABLE 4C. Multivariable Adjusted % Point Change in Key RMNCH Interventions By Insecurity Status, 2010–2015 Minimal Moderate Severe MICS AHS 2010– MICS AHS 2010– MICS AHS 2010– 2010 2015 2015 2010 2015 2015 2010 2015 2015 Average Average Average Minimal vs. Minimal Adjusted Adjusted % Point Adjusted Adjusted % Point Adjusted Adjusted % Point Moderate vs. Severe % % Change % % Change % % Change p-value p-value Contraceptive Use 12.46 14.75 0.4 20.73 16.55 –0.7 27.69 19.42 –1.4 0.000 0.000 ANC 50.47 72.61 3.7 53.55 57.19 0.6 45.74 49.59 0.6 0.000 0.001 SBA 41.33 50.17 1.5 43.84 47.26 0.6 35.76 45.91 1.7 0.149 0.796 BCG 68.28 89.02 3.5 70.75 85.46 2.5 48.26 83.77 5.9 0.096 0.261 DPT3/Penta 39.55 81.11 6.9 35.55 74.4 6.5  9.09 76.94 11.3 0.415 0.000 Measles 52.35 79.91 4.6 46.4 74.27 4.6 33.37 71.63 6.4 0.779 0.342 ORT 40.15 44.41 0.7 53.17 47.94 –0.9 71.86 46.5 –4.2 0.101 0.000 Care seeking for ARI 40.46 64.31 4.0  62.3 75.94 2.3 51.19 69.22 3.0 0.118 0.651 TABLE 5A. Health Systems Composite Indicators by Insecurity Status, 2004–2010 Minimal Moderate Severe Minimal vs. Minimal Insecurity Insecurity Insecurity Moderate vs. Severe Indicator Survey N Group N Group N Group p-value p-value Mean% 21 60.71(18.01) 8 65.98(6.86) 5 64.68(11.07) 0.690 0.862 2004 Mean% 21 79.37(7.22) 8 79.48(6.59) 5 67.54(14.38) 0.999 0.021 Domain A Client and 2010 Community Mean 18.66(15.36) 13.5(10.66) 2.86(8.34) 0.693 0.067 Difference (SD) Mean% 20 51.17(7.35) 8 52.86(8.3) 5 56.12(6.81) 0.853 0.397 2004 Mean% 20 68.77(8.57) 8 73.73(7.08) 5 63.64(16.54) 0.354 0.698 Domain B Human 2010 Resources Mean 17.6(9.99) 20.86(11.46) 7.52(15.29) 0.766 0.186 Difference (SD) Mean% 20 46.99(7.87) 8 48.1(6.28) 5 54.82(11.94) 0.944 0.154 2004 Mean% 20 75.571(8.46) 8 74.18(4.98) 5 67.76(7.98) 0.902 0.123 Domain C Physical Capacity 2010 Mean 29.33(9.89) 26.08(8.67) 12.94(6.18) 0.678 0.003 Difference (SD) Mean% 20 48.6(8.24) 8 38.76(8.4) 5 43.56(15.4) 0.050 0.549 2004 Mean% 20 48.92(10.90) 8 49.69(8.84) 5 38(12.9) 0.966 0.136 Domain D Quality of 2010 Service Provision Mean 0.32(13.41) 10.93(7.17) -5.56(12.15) 0.107 0.598 Difference (SD) Mean% 20 64.97(13.1) 8 70.56(13.47) 5 49.6(29.21) 0.692 0.160 2004 Mean% 20 78.98(12.97) 8 79.1(12.93) 5 65.16(22.71) 0.999 0.154 Domain E Management 2010 Systems Mean 14.86(20.11) 8.54(24.19) 15.56(17.19) 0.749 0.998 Difference (SD) Mean% 20 50.48(1.81) 8 49.63(2.15) 5 49.98(0.39) 0.494 0.842 2004 Mean% 20 50.31(2.68) 8 50.69(3.52) 5 50.58(2.49) 0.948 0.981 Domain F Overall Mission 2010 Mean –0.18(3.78) 1.06(3.69) 0.6(2.75) 0.698 0.905 Difference (SD) (continues on next page) THE AFGHANISTAN HEALTH SERVICES STUDY 35 TABLE 5A. Health Systems Composite Indicators by Insecurity Status, 2004–2010 (continued) Minimal Moderate Severe Minimal vs. Minimal Insecurity Insecurity Insecurity Moderate vs. Severe Indicator Survey N Group N Group N Group p-value p-value Mean% 20 52.15(4.19) 8 51.38(3.42) 5 53.62(8.65) 0.925 0.819 2004 Mean% 20 66.41(6.50) 8 67.89(4.15) 5 58.88(10.05) 0.856 0.074 Overall Means (Provincial) 2010 Mean 14.9(6.54) 16.51(6.29) 5.26(4.57) 0.812 0.012 Difference (SD) TABLE 5B. Health Systems Composite Indicators by Insecurity Status, 2011–2016 Minimal Moderate Severe Minimal vs. Minimal Insecurity Insecurity Insecurity Moderate vs. Severe Indicator Survey N Group N Group N Group p-value p-value Mean% 7 78.96(4.55) 15 75.68(8.48) 11 76.29(13.71) 0.757 0.847 2011 Mean% 7 79.16(6.35) 15 81.25(5.3) 11 79.58(9.45) 0.792 0.992 Domain A Client and 2016 Community Mean 0.2(8.05) 5.23(7.83) 3.29(6.98) 0.331 0.681 Difference (SD) Mean% 7 50.47(6.8) 15 46.51(5.38) 11 52.54(6.28) 0.332 0.758 2011 Mean% 7 57.97(4.64) 15 54.15(8.85) 11 56.85(6.22) 0.498 0.948 Domain B Human 2016 Resources Mean 7.5(5.23) 7.06(9.41) 4.32(6.81) 0.992 0.685 Difference (SD) Mean% 7 33.59(2.75) 15 36.2(8.23) 11 41.35(9.12) 0.747 0.116 2011 Mean% 7 76.56(4.93) 15 77.95(9.15) 11 77.61(8.58) 0.927 0.963 Domain C Physical 2016 Capacity Mean 42.97(3.75) 41.37(11.83) 36.26(14.42) 0.952 0.472 Difference (SD) Mean% 7 46.33(7.69) 15 50.68(8.55) 11 51.2(8.12) 0.489 0.449 2011 Mean% 7 58.99(9.21) 15 60.54(12.65) 11 60.17(9.09) 0.947 0.973 Domain D Quality of 2016 Service Provision Mean 12.66(11.89) 9.46(14.17) 8.97(8) 0.830 0.801 Difference (SD) 36 IMPROVING ACCESS TO AND QUALITY OF HEALTH SERVICES TABLE 5B. Health Systems Composite Indicators by Insecurity Status, 2011–2016 (continued) Minimal Moderate Severe Minimal vs. Minimal vs. Insecurity Insecurity Insecurity Moderate Severe Indicator Survey N Group N Group N Group p-value p-value Mean% 7 42.33(3.9) 15 43.23(8.24) 11 53.95(18.23) 0.985 0.132 2011 Mean% 7 51.13(14.53) 15 50.35(14.7) 11 54.68(13.35) 0.992 0.865 Domain E Management 2016 Systems Mean 8.8(12.95) 5.53(8.12) 0.74(8.74) 0.733 0.200 Difference (SD) Mean% 7 46.41(2.47) 15 48.15(3.53) 11 47.48(4.28) 0.553 0.816 2011 Mean% 7 41.47(7.41) 15 42.6(11.3) 11 50.39(11.06) 0.969 0.200 Domain F Overall Mission 2016 Mean –4.94(6.23) –5.95(11.15) 2.91(11.27) 0.975 0.277 Difference (SD) Mean% 7 54.6(4.29) 15 54.55(4.9) 11 56.53(10.41) 0.999 0.843 2011 Mean% 7 62.21(4.6) 15 62.38(7.19) 11 65.36(7.65) 0.998 0.620 Overall Means 2016 (Provincial) Mean 7.61(4.71) 7.24(6.43) 8.84(13.7) 0.996 0.960 Difference (SD) During 2011–2016, facilities in severe and moderate insecurity provinces achieved similar (unadjusted) improvements in health systems performance as did facilities in minimal insecurity provinces. Remarkably, improvements in health systems performance at severe and moderate facil- ities are statistically non-distinguishable (p > 0.10) from those at minimal insecurity facilities in the 2011–2016 period. During 2004–2010, differences in health systems performance between higher and low insecurity provinces are minimal after adjusting for confounders. During 2004–2010, moderate and severe insecurity facilities made statistically significantly smaller improvements in functioning equip- ment (p < 0.01) after adjusting for key confounders (table 6; figure 13A). Moderate insecurity facilities also achieved smaller improvements (p < 0.10) in infrastructure relative to minimal insecurity facilities. However, the differences in improvement as compared to minimal insecurity facilities are very small (less than 3.3 percent points over 2004–2010). The main exception is pace of improvements in functioning equipment at severe insecurity facilities, which was 12.3 percentage points lower in severe insecurity facilities relative to minimal insecurity facilities (p = 0.01). In 2011–2016 as well, differences in health systems performance between higher and lower insecurity provinces remain minimal after adjusting for confounders. Adjustments for confound- ers found that insecurity negatively impacted improvements in infrastructure, client assessment, and provider knowledge in severe and/or moderate insecurity facilities as compared to minimal insecurity THE AFGHANISTAN HEALTH SERVICES STUDY 37 TABLE 6. Multivariable Adjusted Impact of Conflict on Change in Key Health Systems Indicators 2004–2010 2011–2016 Insecurity* time Interaction [Reference = Minimal] Coef. 95% CI p-value Coef. 95% CI p-value Equipment functionality index: Moderate insecurity* time –3.26 –5.07–1.45 <0.0001 1.7 0.61–2.7 0.002 Equipment functionality index: Severe insecurity* time –12.37 –21.38–3.36 0.01 0.24 –1.22–1.6 0.750 Drug availability index: Moderate insecurity* time 0.18 –2.77–3.14 0.903 — — — Drug availability index: Severe insecurity* time 13.81 –0.95–28.57 0.067 — — — Pharmaceuticals and vaccines availability index: — — — 3.12 1.96–4.2 <0.0001 Moderate insecurity* time Pharmaceuticals and vaccines availability index: — — — 0.6 -0.95–2.1 0.448 Severe insecurity* time Infrastructure index: Moderate insecurity* time –2.26 –4.75–0.24 0.077 — — — Infrastructure index: Severe insecurity* time 6.99 –5.38–19.36 0.268 — — — Functional infrastructure index: Moderate insecurity* time — — — –1.85 –3.64–0.07 0.042 Functional infrastructure index: Severe insecurity* time — — — –2.84 –5.21–0.46 0.019 Patient history and physical examination index: 0.98 –0.77–2.74 0.271 — — — Moderate insecurity* time Patient history and physical examination index: –4.00 –12.79–4.79 0.372 — — — Severe insecurity* time Client background and physical assessment index: — — — –0.16 –1.27–0.95 0.779 Moderate insecurity* time Client background and physical assessment index: — — — –2.75 –4.24–1.2 <0.0001 Severe insecurity* time Patient counseling index: Moderate insecurity* time –0.92 –3.44–1.60 0.474 — — — Patient counseling index: Severe insecurity* time 6.85 –5.77–19.46 0.288 — — — Client counseling index: Moderate insecurity* time — — — 0.73 –1.04–2.4 0.420 Client counseling index: Severe insecurity* time — — — –1.28 –3.66–1.09 0.290 Female health worker index: Moderate insecurity* time — — — -0.19 –1.73–1.3 0.813 Female health worker index: Severe insecurity* time — — — 1.6 –0.46–3.6 0.128 Provider knowledge score: Moderate insecurity* time — — — 0.80 –0.23–1.81 0.127 Provider knowledge score: Severe insecurity* time — — — –1.84 –3.21–0.46 0.009 Note: *Models adjusted for the main effects of insecurity and time, and covariates including patient volume, facility type, geographic region, and contracting type; complete results in the appendix; (–) indicates unavailable indicators. facilities during 2011–2016 (table 6; figure 13B). However, the pace of improvements in health sys- tems performance was only slightly smaller in severe insecurity areas compared to minimal insecurity areas (between 1.8 and 2.8 percent points lower over 5 years). Severe insecurity facilities achieved statistically significantly greater improvements in functioning equipment (p < 0.01) and availability of drugs (p < 0.01) relative to minimal insecurity facilities, although the difference in improvements is relatively small. There were no statistically significant differences in improvements between severe, moderate, and minimal insecurity facilities on other health systems performance domains examined over 2011–2016. Crude boxplots, adjusted means, and complete regression models are included in appendix E. 38 IMPROVING ACCESS TO AND QUALITY OF HEALTH SERVICES FIGURE 13A. Annual percentage point change in health systems performance by severity of insecurity, 2004–2010 (Reference: Minimal insecurity)* 20 15 13.8 10 7.0 6.9 5 0.2 1.0 0 –2.3 –0.9 –5 –3.3 – 4.0 –10 –15 –12.4 Functioning equipment Drug availability Infrastructure Patient history and Patient counseling examination Moderate versus Minimal insecurity Severe versus Minimal insecurity Note: *Red values indicate statistical significance at p = 0.10 or less. Model adjusted for the main effects of insecurity and time, and covariates including patient volume, facility type, geographic region, and contracting type. Positive numbers mean that provinces with severe or moderate insecurity achieved greater improvements than did provinces with minimal insecurity. FIGURE 13B. Annual percentage point change in health systems performance indicators by severity of insecurity, 2011–2016 (Reference: Minimal insecurity)* 4 3.1 3 2 1.7 1.6 0.7 0.8 1 0.6 0.2 0 –0.2 –1 –0.2 –1.3 –2 –1.9 –1.8 –3 –2.8 –2.8 –4 Functioning Drugs & Infrastructure Patient physical Patient Female health Provider equipment vaccines assessment counseling worker knowledge availability Moderate versus Minimal insecurity Severe versus Minimal insecurity Note: *Red values indicate statistical significance at p = 0.10 or less. Model adjusted for the main effects of insecurity and time, and covariates including patient volume, facility type, geographic region, and contracting type. Positive numbers mean that provinces with severe or moderate insecurity achieved greater improvements than did provinces with minimal insecurity. THE AFGHANISTAN HEALTH SERVICES STUDY 39 KEY SECTION CONCLUSIONS 3.4  AND CONSIDERATIONS Overall, service delivery has been resilient to insecurity. Improvements in service coverage, as well as health systems performance, are evident across the insecurity spectrum in unadjusted compar- isons. After adjusting for key factors that might influence this finding, it is apparent that even in those instances where improvements over time have been smaller in higher insecurity facilities, the difference in pace of improvements is relatively small. Indicative findings from interviews and focus group discus- sions with key stakeholders indicate that insecurity resilience may be driven by specific NGO strategies that result in stronger links with local communities. Implementers and policymakers in Afghanistan highlight the challenges posed by insecurity to delivering services effectively. Interviews and discussions with stakeholders conducted as a part of this study suggest that security has been a challenge, resulting in increased difficulty with identifying and retaining suitable human resource, supply management, coordination, and supervising and monitoring of health facilities. Evidence from other settings also underscores the devastating impact that insecurity has on service delivery. In this context, relatively small differences in performance between higher and minimal insecurity facilities are an important achievement. This finding of insecurity resilience merits further investigation with health management information systems data. These conclusions are based on survey data. Survey teams may have been unable to access facilities in the most insecure areas. This could mean that the facilities eventually sur- veyed were better-performing relative to others in high insecurity areas, and thus could result in an under- estimation of the impact of insecurity on health facility performance. Future survey strategies should clearly describe the sampling frames and geospatial distribution within provinces. Exploring variations in performance between facilities using health management information system data could yield more granular insights into good practices with insecurity adaptation. 40 IMPROVING ACCESS TO AND QUALITY OF HEALTH SERVICES CHAPTER 4 IMPLICATIONS OF CONTRACTING TYPE FOR SERVICE COVERAGE AND HEALTH SYSTEMS PERFORMANCE 4.1  DIFFERENCES IN SERVICE COVERAGE Differences in RMNCH interventions coverage by contracting mechanism are presented in tables 7A-7B, appendix F). At baseline (2003) before these contracting models were implemented, CO provinces had significantly higher coverage of ANC (p = 0.002), facility births (p = 0.000), SBA (p = 0.001) and DPT3 coverage (p = 0.012). In 2010, CI provinces had higher levels of ANC (p = 0.002) and lower coverage of ORT (p = 0.029). In 2015, CI provinces had significantly higher coverage in all interventions (p < 0.05) relative to CO provinces. Unadjusted comparisons of changes in service coverage between 2003–2010 and 2010–2015 clearly indicate that CI provinces achieved greater absolute gains in coverage (i.e., mean differ- ences between time points) for some RMNCH indicators (p < 0.10) than did CO provinces. During 2003–2010, CI provinces achieved greater increases in coverage than CO provinces on facility delivery and SBA (p < 0.01), and care seeking for ARI (p < 0.10). CO provinces accomplished greater increases than CI provinces on ANC coverage (p < 0.01). However, in 2010–2015, CI provinces achieved statis- tically significantly greater increases (p < 0.10) on most indicators than CO provinces (see table 7A–B) This report explored the impact of contracting approach through multivariable analy- sis on change in contraceptive use, ANC, SBA, BCG, DPT3/Penta, measles vaccinations, care-seeking for ARI and ORT use for the 2003–2010 and 2010–2015 time periods. Model results, covariate-adjusted prevalence, and slopes over time are displayed in tables 8A–C and figures 14A–B and 15. During 2003–2010, CO provinces made smaller improvement in ANC, SBA, and care seeking for ARI, but greater improvement in ORT coverage as compared to CI provinces after adjusting for confounders. However, the difference in improvements is small. Over 2003– 2010, covariate-adjusted improvements in ANC, SBA, and care seeking for ARI were slower among CO provinces relative to CI (tables 8A and 8B): on average, coverage of these interventions in CO provinces increased only about 3.8%, 2.8%, and 1.6% points per year, while average gains in CI prov- inces were 5.9%, 4.4%, and 3.5% points per year, p = 0.003, 0.007, 0.052, respectively. ORT coverage, however, improved more rapidly in CO provinces during the same time period (3.0% vs 1.3% points, p = 0.064) (see tables 8A and 8B). However, the difference in the pace of improvements is small, with a maximum difference of 2.1 percentage points over 2003–2010. 41 TABLE 7A. RMNCH Interventions Coverage by Contracting Mechanism, 2003–2010 Contracting In Contracting vs. Contracting Indicator Survey N Contracting IN N Out Out p-value Mean% MICS 2003 1,064 7.7 19,898 10.4 0.158 Contraceptive any method Mean% MICS 2010 1,175 18.2 16,899 17.5 0.752 Mean Difference 10.5 7.1 0.159 Mean% MICS 2003 1,061 6.0 19,874 8.6 0.111 Contraceptive any modern method Mean% MICS 2010 1,175 16.5 16,899 16.1 0.840 Mean Difference 10.5 7.4 0.130 Mean% MICS 2003 577 7.9 10,310 16.6 0.002 ANC by skilled provider Mean% MICS 2010 228 60.2 4,646 48.2 0.005 Mean Difference 52.3 31.6 0.000 Mean% MICS 2003 590 3.4 10,418 13.4 0.000 Facility delivery Mean% MICS 2010 230 38.8 4,732 32.4 0.213 Mean Difference 35.4 19.0 0.000 Mean% MICS 2003 587 5.4 10,428 14.9 0.001 Skilled birth attendance Mean% MICS 2010 230 44.3 4,732 38.2 0.197 Mean Difference 38.8 23.3 0.000 Mean% MICS 2003 242 46.6 4,403 35.6 0.060 BCG Mean% MICS 2010 137 60.9 2,355 61.6 0.920 Mean Difference 14.3 26.0 0.183 Mean% MICS 2003 241 15.2 4,403 28.5 0.012 DPT3/PENTA3 Mean% MICS 2010 137 23.8 2,296 31.4 0.111 Mean Difference 8.6 2.9 0.276 Mean% MICS 2003 243 27.7 4,409 24.3 0.571 Measles Mean% MICS 2010 136 42.2 2,338 40.2 0.786 Mean Difference 14.5 15.9 0.832 Mean% MICS 2003 317 36.1 4,401 47.0 0.089 Care seeking for ARI Mean% MICS 2010 175 63.2 2,774 60.3 0.470 Mean Difference 27.1 13.3 0.061 Mean% MICS 2003 509 27.3 7,227 29.8 0.567 ORT Mean% MICS 2010 217 41.0 3,223 54.1 0.029 Mean Difference 13.7 24.3 0.202 42 PROGRESS IN THE FACE OF INSECURITY TABLE 7B. RMNCH Interventions Coverage by Contracting Mechanism, 2010–2015 Contracting IN Contracting vs. Contracting Indicator Survey N Contracting IN N Out Out p-value Mean% MICS 2010 1,175 18.2 16,899 17.5 0.752 Contraceptive any method Mean% AHS 2015 1,939 26.5 22,782 17.6 0.000 Mean Difference 8.3 0.1 0.012 Mean% MICS 2010 1,175 16.5 16,899 16.1 0.840 Contraceptive any modern method Mean% AHS 2015 1,939 23.6 22,782 15.4 0.000 Mean Difference 7.0 -0.7 0.011 Mean% MICS 2010 228 60.2 4,646 48.2 0.005 ANC by skilled provider Mean% AHS 2015 588 80.4 7,363 59.3 0.000 Mean Difference 20.2 11.1 0.013 Mean% MICS 2010 230 38.8 4,732 32.4 0.213 Facility delivery Mean% AHS 2015 623 71.6 7,797 53.6 0.000 Mean Difference 32.8 21.2 0.078 Mean% MICS 2010 230 44.3 4,732 38.2 0.197 Skilled birth attendance Mean% AHS 2015 640 69.4 8,041 50.3 0.000 Mean Difference 25.2 12.2 0.025 Mean% MICS 2010 137 60.9 2,355 61.6 0.920 BCG Mean% AHS 2015 308 98.3 3,678 84.9 0.000 Mean Difference 37.4 23.3 0.000 Mean% MICS 2010 137 23.8 2,296 31.4 0.111 DPT3/PENTA3 Mean% AHS 2015 308 94.3 3,678 75.9 0.000 Mean Difference 70.5 44.5 0.000 Mean% MICS 2010 138 8.9 2,361 16.1 0.065 Fully immunized Mean% AHS 2015 308 76.9 3,678 67.4 0.019 Mean Difference 67.9 51.4 0.006 Mean% MICS 2010 136 42.2 2,338 40.2 0.786 Measles Mean% AHS 2015 308 82.1 3,678 74.1 0.028 Mean Difference 39.9 33.9 0.301 Mean% MICS 2010 175 63.2 2,774 60.3 0.470 Care seeking for ARI Mean% AHS 2015 149 95.3 2,301 70.8 0.000 Mean Difference 32.1 10.5 0.000 Mean% MICS 2010 217 41.0 3,223 54.1 0.029 ORT Mean% AHS 2015 217 74.8 3,350 45.9 0.000 Mean Difference 33.8 –8.3 0.000 IMPROVING HEALTH OUTCOMES IN AFGHANISTAN 43 TABLE 8A. Multivariable Adjusted Impact of Contracting Mechanism on Change in Key RMNCH Interventions Contracting type* time Interaction [Reference = MICS 2003- MICS 2010 MICS 2010- AHS 2015 contracting in] Coef. 95% CI p-value Coef. 95% CI p-value Contraceptive Use –0.32 (–0.87 0.23) 0.253 –0.52 (–0.9 –0.14) 0.008 ANC –0.94 (–1.56 –0.33) 0.003 –0.43 (–0.85 –0.01) 0.045 SBA –0.96 (–1.66 –0.27) 0.007 –0.46 (–0.95 0.02) 0.058 BCG 0.39 (–0.4 1.18) 0.334 –2.18 (–3.21 –1.14) 0.000 DPT3/Penta –0.17 (–0.95 0.61) 0.664 –1.86 (–2.66 –1.05) 0.000 Measles 0.13 (–0.76 1.02) 0.777 –0.11 (–0.86 0.64) 0.776 ORT 0.61 (–0.04 1.26) 0.064 –1.74 (–2.37 –1.12) 0.000 Care seeking for –0.61 (–1.22 0.01) 0.052 –1.93 (–2.85 –1.01) 0.000 ARI Note: *Models adjusted for the main effects of contracting type and time, and covariates including maternal illiteracy, insecurity, and rurality of residence; complete results in the appendix. TABLE 8B. Multivariable Adjusted % Point Change in Key RMNCH Interventions by Contracting Group, 2003–2010 Contracting IN Contracting OUT MICS 2003 MICS 2010 2003–2010 MICS 2003 MICS 2010 2003–2010 p-value for Average % Average % interaction Adjusted % Adjusted % Point Change Adjusted % Adjusted % Point Change term Contraceptive Use 10.81 22.41 1.5 10.47 17.04 0.8 0.253 ANC 13.56 60.95 5.9 16.93 47.28 3.8 0.003 SBA 9.89 44.94 4.4 15.14 37.35 2.8 0.007 BCG 45.09 60.79 2.0 35.64 60.11 3.1 0.334 DPT3/Penta 19.88 25.99 0.8 28.28 32.32 0.5 0.664 Measles 24.6 40.33 2.0 24.49 43.18 2.3 0.777 ORT 32.63 43.23 1.3 28.87 53.19 3.0 0.064 Care seeking for ARI 36.75 64.37 3.5 47.16 60.19 1.6 0.052 TABLE 8C. Multivariable Adjusted % Point Change in Key RMNCH Interventions by Contracting Group, 2010–2015 Contracting IN Contracting OUT MICS 2010 AHS 2015 2010–2015 MICS 2010 AHS 2015 2010–2015 P-value for Average % Average % interaction Adjusted % Adjusted % Point Change Adjusted % Adjusted % Point Change term Contraceptive Use 24.33 30.89 1.1 18.58 16.03 –0.4 0.008 ANC 64.85 79.14 2.4 49.93 57.5 1.3 0.045 SBA 55.38 70.26 2.5 41.4 46.61 0.9 0.058 BCG 67.66 98.09 5.1 66.15 84.96 3.1 0.000 DPT3/Penta 31.18 94.17 10.5 34.93 75.88 6.8 0.000 Measles 52.06 81.21 4.9 44.74 74.12 4.9 0.776 ORT 35.46 71.67 6.0 51.07 45.95 –0.9 0.000 Care seeking for ARI 53.23 93.67 6.7 55.61 70.99 2.6 0.000 FIGURE 14A. Annual percentage point difference in service coverage by type of contracting, 2003–2010 (Reference: Contracting-In)* 4 3 2 1.7 1.1 1 0.3 0 –1 –0.3 –0.7 –2 –1.6 –2.1 –1.9 –3 –4 Contraceptive ANC SBA BCG DPT3/Penta Measles ORT Care seeking Use for ARI Note: *Red values indicate statistical significance at p = 0.10 or less. Model adjusted for the main effects of contracting type and time, and covariates including maternal illiteracy, insecurity, and rurality of residence. Positive numbers mean the CO provinces achieved greater improvements than did CI provinces. FIGURE 14B. Annual percentage point difference in service coverage by type of contracting, 2010–2015 (Reference: Contracting-In)* 0 0 –2 –1.5 –1.1 –1.6 –2 –4 –3.7 – 4.1 –6 –8 – 6.9 –10 –12 Contraceptive ANC SBA BCG DPT3/Penta Measles ORT Care seeking Use for ARI Note: *Red values indicate statistical significance at p = 0.10 or less. Model adjusted for the main effects of contracting type and time, and covariates including maternal illiteracy, insecurity, and rurality of residence. Positive numbers mean the CO provinces achieved greater improvements than did CI provinces. IMPROVING HEALTH OUTCOMES IN AFGHANISTAN 45 FIGURE 15. Multivariable adjusted means of key RMNCH interventions by contracting type MICS 2003–MICS 2010 MICS 2010–AHS 2015 Contraceptive any method Contraceptive any method .3 .35 .25 .3 Proportion Proportion .2 .25 .15 .1 .2 .05 .15 2003 2010 2010 2015 Year Year Contracting IN Contracting OUT Contracting IN Contracting OUT ANC by skilled provider ANC by skilled provider .8 .9 .6 .8 Proportion Proportion .4 .7 .2 .6 0 .5 2003 2010 2010 2015 Year Year Contracting IN Contracting OUT Contracting IN Contracting OUT Skilled birth attendance Skilled birth attendance .5 .8 .4 Proportion .7 Proportion .3 .6 .2 .1 .5 0 .4 2003 2010 2010 2015 Year Year Contracting IN Contracting OUT Contracting IN Contracting OUT BCG BCG .7 1 .6 .9 Proportion Proportion .8 .5 .7 .4 .6 .3 .5 2003 2010 2010 2015 Year Year Contracting IN Contracting OUT Contracting IN Contracting OUT 46 PROGRESS IN THE FACE OF INSECURITY FIGURE 15. Multivariable adjusted means of key RMNCH interventions by contracting type (continued) MICS 2003-MICS 2010 MICS 2010-AHS 2015 DPT3/PENTA3 DPT3/PENTA3 .35 1 .3 .8 Proportion Proportion .25 .6 .2 .15 .4 .1 .2 2003 2010 2010 2015 Year Year Contracting IN Contracting OUT Contracting IN Contracting OUT Measles Measles .6 .9 .5 .8 Proportion Proportion .4 .7 .3 .6 .2 .5 .1 .4 2003 2010 2010 2015 Year Year Contracting IN Contracting OUT Contracting IN Contracting OUT ORS use ORS use .6 .8 .5 Proportion Proportion .6 .4 .4 .3 .2 .2 2003 2010 2010 2015 Year Year Contracting IN Contracting OUT Contracting IN Contracting OUT Care seeking for ARI Care seeking for ARI .7 1 .6 Proportion Proportion .8 .5 .4 .6 .3 .4 2003 2010 2010 2015 Year Year Contracting IN Contracting OUT Contracting IN Contracting OUT IMPROVING HEALTH OUTCOMES IN AFGHANISTAN 47 CO provinces made smaller improvements than did CI provinces on many service cov- erage indicators between 2010 and 2015 after adjusting for confounders; however, the dif- ference in improvements is small with the exception of ORT. From 2010 to 2015, coverage gains were significantly higher in CI provinces for almost all indicators assessed (p < 0.05) (see figure 14B, tables 8A and 8C). Respectively for CI versus CO provinces, greatest improvements were observed for BCG (average increase 5.1% vs 3.1% points, p = 0.000), DPT3/Penta (10.5% vs 6.8% points, p = 0.000), ORT (6.0% vs -0.9% points, p = 0.000) and care seeking for ARI (6.7% vs 2.6% points, p = 0.000) (see table 9C). However, the difference in improvements is small. There is one exception: annual increases in ORT use were 6.9 percent points lower in CO provinces than in CI provinces over 2010–2015. Crude boxplots and complete regression models are included in appendix F. DIFFERENCES IN HEALTH FACILITY 4.2  PERFORMANCE INDICATORS Unadjusted comparisons indicate that CI and CO facilities find no significant differences in health facility performance over the 2004–2010 period. Health systems performance broad domains were comparable between CO and CI provinces, 2004–2010 (table 9A, p > 0.10). By contrast, unadjusted comparisons between 2011–2016 find that CO facilities achieved greater improvements relative to CI facilities on the client and community domain. In fact, CI regressed in this period, while CO made improvements in this domain (table 9B). The pace of improve- ments in client and community performance was about 9.2 percent points higher in CO facilities than in CI facilities over 2011–2016 (p < 0.05). Adjusted comparisons of improvements over time in health systems domains in the CO and CI models find greater improvements in CI facilities relative to CO facilities in 2004–2010. Table 10 and figure 16A present the impact of contracting type on health systems performance indica- tors for 2004–2010. After adjusting for covariates, CO facilities achieved greater improvements from 2004–2010 in the patient counseling index (β = 3.1, p = 0.001) relative to CI facilities, but smaller improvements in drug availability (β = –2.47, p = 0.024). However, the difference in pace of improve- ments between CO and CI facilities is relatively small, with a maximum difference of 3 percentage points over 2004–2010. Over 2011–2016, however, CO facilities achieved greater improvements over time relative to CI comparators in adjusted comparisons. From 2011 to 2016, however, CO facilities achieved greater improvements than CI facilities on several indicators, including equipment functionality (p = 0.009), the availability of pharmaceuticals and vaccines (p < 0.0001), appropriate client back- ground and physical assessment (p = 0.07), and client counseling (p = 0.03). Table 10 and figure 16B have more details. Covariate-adjusted mean plots for each indicator over time are included in appendix G. However, the difference in rate of improvement over time is relatively small with the exception of improvements in drug availability, which were 8.4 percentage points higher per year in CO facilities than in CI facilities. Crude boxplots and complete regression models are included in appendix G. 48 PROGRESS IN THE FACE OF INSECURITY TABLE 9A. Composite Health Systems Indicators by Contracting Type, 2004–2010 Contracting In Contracting In Contracting Out vs. Contracting Indicator Survey N Group N Group Out p-value Mean% 2004 3 66.92(21.19) 30 62.11 (14.76) 0.605 Domain A Client and Mean% 2010 3 81.99(9.52) 31 77.23 (9.19) 0.399 Community Mean Difference (SD) 15.07(16.33) 15.12 (14.49) 0.995 Mean% 2004 3 50.14(7.12) 30 52.55(7.6) 0.602 Domain B Human Mean% 2010 3 69.59(8.04) 31 68.47(10.75) 0.863 Resources Mean Difference (SD) 19.45(1.04) 16.6(12.17) 0.691 Mean% 2004 3 43.57(8.34) 30 48.94(8.42) 0.300 Domain C Physical Mean% 2010 3 73.05(4.91) 31 74.2(8.27) 0.816 Capacity Mean Difference (SD) 29.48(13.21) 25.71(10.55) 0.568 Mean% 2004 3 55.29(12.81) 30 44.45(9.57) 0.078 Domain D Quality of Mean% 2010 3 53.43(8.35) 31 46.67(11.32) 0.323 Service Provision Mean Difference (SD) –1.86(19.77) 2.39(12.48) 0.568 Mean% 2004 3 70.87(21.96) 30 63.31(16.84) 0.474 Domain E Management Mean% 2010 3 82.73(16.77) 31 76.42(15) 0.495 Systems Mean Difference (SD) 11.87(30.32) 13.59(19.8) 0.891 Mean% 2004 3 51.1(1.01) 30 50.08(1.81) 0.346 Domain F Overall Mean% 2010 3 48.17(1.96) 31 50.63(2.78) 0.145 Mission Mean Difference (SD) –2.93(1.26) 0.56(3.56) 0.103 Mean% 2004 3 53.3(1.85) 30 52.07(4.98) 0.678 Overall Means Mean% 2010 3 67.6(6.36) 31 65.47(7.21) 0.626 (Provincial) Mean Difference (SD) 14.3(6.35) 13.78(7.29) 0.907 KEY SECTION CONCLUSIONS 4.3  AND CONSIDERATIONS Policy conclusions on the relative performance of CO and CI must take on board the fact that CI provinces are much closer to the capital city and smaller than are most CO provinces. This makes CI facilities easier to staff, supply, and manage on average than CO facilities, and therefore easier to improve coverage in these provinces, as well. The analysis methods cannot control for these sys- tematic advantages which are likely to bias findings in favor of CI provinces. Despite this, it is clear that CO and CI approach have delivered similar results with the excep- tion of the availability of drugs and vaccines. Although there are some differences in which contracting type delivers greater improvements in service coverage and health systems performance, these differ- ences are relatively small for almost all health systems performance domains. Substantially greater improvements in pharmaceutical and vaccine availability in CO facili- ties points to the importance of continuing decentralized procurement and supply chains. This IMPROVING HEALTH OUTCOMES IN AFGHANISTAN 49 TABLE 9B: Composite Health Systems Indicators by Contracting Type, 2011–2016 Contracting In Contracting In Contracting Out vs. Contracting Indicator Survey N Group N Group Out p-value Mean% 2011 3 79.67(2.09) 30 76.25(10.21) 0.572 Domain A Client and Mean% 2016 3 74.83(3.68) 31 80.81(6.99) 0.158 Community Mean Difference (SD) –4.83(2.08) 4.37(7.47) 0.044 Mean% 2011 3 53.69(1.94) 30 48.91(6.57) 0.225 Domain B Human Mean% 2016 3 53.33(6.27) 31 56.04(7.49) 0.550 Resources Mean Difference (SD) –0.36(4.33) 6.91(7.77) 0.124 Mean% 2011 3 32.83(2.42) 30 37.79(8.4) 0.322 Domain C Physical Mean% 2016 3 71.18(2.41) 31 78.18(8.17) 0.154 Capacity Mean Difference (SD) 38.35(2.92) 40.2(12.24) 0.799 Mean% 2011 3 50.55(11.03) 30 49.85(8.11) 0.890 Domain D Quality of Mean% 2016 3 53.1(13.73) 31 60.78(10.34) 0.239 Service Provision Mean Difference (SD) 2.55(24.16) 10.73(10.19) 0.253 Mean% 2011 3 42.24(3.55) 30 47.05(13.35) 0.544 Domain E Management Mean% 2016 3 40.08(7.46) 31 53.06(13.98) 0.126 Systems Mean Difference (SD) –2.17(5.77) 5.3(9.79) 0.207 Mean% 2011 3 47.65(0.05) 30 47.52(3.76) 0.954 Domain F Overall Mean% 2016 3 36.1(8.23) 31 45.72(10.83) 0.146 Mission Mean Difference (SD) –11.55(8.28) –1.9(10.81) 0.145 Mean% 2011 3 56.1(1.49) 30 55.13(7.31) 0.822 Overall Means Mean% 2016 3 56.73(3.57) 31 63.95(6.81) 0.082 (Provincial) Mean Difference (SD) 0.63(5.06) 8.58(9.08) 0.149 TABLE 10: Multivariable Adjusted Impact of Contracting Mechanism on Change in Key Health Systems Indicators Contracting Type* Time Interaction 2004–2010 2011–2016 [Reference = Contracting In] Coef. 95% CI p-value Coef. 95% CI p-value Equipment functionality index –1.05 –2.36–0.25 0.114 2.15 0.53–3.77 0.009 Drug availability index –2.47 –4.62–0.32 0.024 — — — Pharmaceuticals and vaccines availability — — — 8.39 6.68–10.10 <0.0001 index Infrastructure index 0.57 –1.20–2.35 0.526 — — — Functional infrastructure index — — — 0.82 –1.79–3.42 0.538 Patient history and physical examination 0.33 –0.98–1.65 0.622 — — — index Client background and physical — — — 1.55 –0.13–3.24 0.071 assessment index Patient counseling index 3.12  1.25–4.99 0.001 — — — Client counseling index — — — 2.97 0.27–5.66 0.031 Female health worker index — — — –0.16 –2.43–2.12 0.894 Provider knowledge score — — — –0.53 –2.09–1.03 0.504 Note: *Models adjusted for the main effects of contracting type and time, and covariates including patient volume, facility type, distance of facility from province centre, geographic region, and insecurity; complete results in the appendix. (–) indicates unavailable indicators. FIGURE 16A: Annual percentage point difference in health systems performance by type of contracting, 2004–2010 (Reference: Contracting-In) * 4 3.12 3 2 1 0.57 0.33 0 –1 –1.05 –2 –3 –2.47 Functioning Drug availability Infrastructure Patient history and Patient counseling equipment examination Note: *Red values indicate statistical significance at p = 0.10 or less. Model adjusted for the main effects of contracting type and time, and covariates including patient volume, facility type, distance of facility from province center, geographic region, and insecurity. Positive numbers mean the CO facilities achieved greater improvements than did CI facilities. FIGURE 16B. Average percentage point difference in health systems performance by type of contracting, 2011–2016 (Reference: Contracting-In) * 9 8.39 8 7 6 5 4 2.97 3 2.15 2 1.55 0.82 1 0 –1 –0.16 –0.53 Functioning Drugs & Infrastructure Patient Patient Female health Provider equipment vaccines physical counseling worker knowledge availability assessment Note: *Red values indicate statistical significance at p = 0.10 or less. Model adjusted for the main effects of contracting type and time, and covariates including patient volume, facility type, distance of facility from province center, geographic region, and insecurity. Positive numbers mean the CO facilities achieved greater improvements than CI facilities. is particularly striking since CO provinces are more remote and greater than the CI provinces, implying that it may be more difficult to keep them adequately supplied. The current system of decentralized procurement and supply chains appears to be functioning better, and these findings therefore suggest that it should be continued. Future supply chain and commodities assessments should also include information on batch numbers and expiration dates to assess utilization. At the same time, there is a need to oversee drug quality through drug quality surveys and other approaches that independently assess whether drugs actually available at service delivery points meet quality standards. IMPROVING HEALTH OUTCOMES IN AFGHANISTAN 51 PART III DISCUSSION AND CONCLUSIONS A fghanistan has made noteworthy gains towards improving maternal and child health and the country has made considerable progress towards achieving the Millen- nium Development Goals for maternal and child health (MDGs 4 and 5). Service coverage of key reproductive maternal neonatal and child health interventions increased between 2003 and 2015 across the country, with no indications of a slowdown in the 2010–2015 period, despite the increase in insecurity. Health systems performance continued to improve considerably at the national level during 2004–2016. From 2011–2016 improvements largely continued, but at a slower pace and with the exception of physical capacity to deliver high quality care, which has continued to register substantial improvements. These trends are mirrored at the provincial level. While some differences are evident by individual indicators, there were overall increases in service coverage as assessed by the composite coverage index2 in nearly all provinces over 2003–2015, except for Nimroz and Nuristan during the years 2003–2010 and Khost and Zabul over 2010–2015. Similarly, all provinces (except Zabul) made progress on health systems domains over 2004–2010, while all provinces (except Kapisa, Kunar, and Badghis) achieved health systems improvements over 2011–2016. Improvements achieved in health outcomes and service coverage compare very favorably with improvements achieved by comparators. Afghanistan has achieved greater improvements in key maternal and child health outcomes and service coverage than regional comparators. Improve- ments over time have also exceeded the global median for countries that started off at the same baseline levels as Afghanistan during 2003–04 (Akseer and others 2016; Arur and others 2011). Nevertheless, in absolute terms, there is considerable room for progress which will be needed to achieve the SDG health targets. Escalating insecurity is one of the main challenges to improving health outcomes in Afghan- istan. Insecurity has large direct and indirect adverse impacts on health and well-being. Escalations 2 This overall assessment of coverage was based on improvements in the composite coverage index which is a widely used measure calculated as the weighted mean of eight essential interventions spanning the continuum of care, including both preventive and curative care. 53 in armed conflict increase health needs, while also weakening health systems and service delivery. Analyses of battle-related deaths using data from the Uppsala Conflict Data Program, a gold standard in international analyses of insecurity, finds a substantial increase in insecurity after 2010, particularly in the South, East, and North Eastern regions. Service delivery has been generally resilient to insecurity. The report finds that health service delivery has been resilient to insecurity with a few exceptions. Service coverage increases over time have been smaller in severe or moderate insecurity provinces than in minimal insecurity provinces. However, improvements in coverage are apparent across the insecurity spectrum. Furthermore, after adjusting for confounders that could influence the relationship between insecurity and RMNCH service coverage, we find that the difference in the pace of improvements between higher and minimal insecurity provinces is relatively small. Similarly, health systems performance improvements are also evident across the inse- curity spectrum, and differences in the pace of health systems performance improvement between high and low security provinces were minimal, as well. Stronger links between service providers and NGOs with local communities may have been a key contributor to insecurity resilience. CO provinces experienced higher levels of insecurity than did CI provinces with two of three CI provinces falling in the minimal intensity insecurity category during both time periods; and the third province moving from minimal to moderate intensity insecurity during the 2010–2015. Interviews and discussions with stakeholders as a part of this study point to spe- cific NGO strategies, notably links with local communities, as a potential drivers of insecurity resilience, as links with local communities and stakeholders were identified as key potential drivers of insecurity resilience in service delivery in the health services study, where NGOs recruit staff from local commu- nities and build relationships with local powerbrokers. Such strategies enable NGOs to maintain service delivery in difficult contexts where there might be few alternative sources of medical services. Going forward, there may be benefits to embedding services closer to communities by strengthening ties with and accountability to local communities. Recent studies have also highlighted other potential explanatory factors for the resilience of service delivery in insecure settings. In particular two recent World Bank studies are relevant in this regard. The study, “Critical Administrative Constraints to Service Delivery: Improving Public Services in Afghanistan’s Transformational Decade” (2014), suggests that the contracting approach has enabled the health sector to bypass some of the administrative and political economy bottlenecks that have stymied other sectors (though the study also points out that this may be at the cost of political buy-in). The more recent study, “Social Service Delivery in Violent Contexts: Achieving Results against the Odds” (2017), argues that it is crucially the nature of the elite bargaining at sub-national level that facilitates sustained service delivery. This finding of insecurity resilience merits further exploration with health management infor- mation systems data. Health service coverage and health systems performance data in this report are sourced from household and health facility surveys. Every type of data source has its own drawbacks and, in this case, a key concern could be that survey teams may not have been able to access the worst affected facilities. Health management information system data in Afghanistan are considered to be relatively complete and of good quality. It may therefore be worth exploiting these data further to explore these relationships prospectively, not only to confirm findings of resilience, but also to gain more granular insights into good practices to adapt to levels of insecurity. Yet, absolute levels of coverage are low for 54 PROGRESS IN THE FACE OF INSECURITY many health services, and there is a need to continue to build on the improvements that have so far been achieved. The location and size of CI provinces gives them systematic advantages over CO provinces for which the analysis cannot control. Findings from the contracting model comparisons must take on board an important caveat. CI provinces are much closer to Kabul and smaller than most CO provinces. This makes CI facilities easier to staff, supply and manage on average than CO facilities and therefore to improve coverage in these provinces as well. The analysis methods cannot control for these systematic advantages which are likely to bias findings in favor of CI provinces. CO and CI provinces achieve similar improvements in service coverage, with some excep- tions. Both CO and CI provinces achieved improvement in maternal and child health coverage in 2003–2010, as well as over 2010–2015 and with a few exceptions. Unadjusted comparisons of relative improvements in maternal and child health coverage over time find that CI provinces made greater gains in coverage in both 2003–2010 and 2010–2015 compared to CO provinces. After adjusting for con- founders, CI provinces still achieved statistically significantly greater improvements on most service coverage indicators relative to CO provinces in both 2003–2010 and 2010–2015. However, the absolute difference in the pace of improvements achieved by the two approaches is not very large. Improvements in ORT use is a noteworthy exception to this: the rate of increase in ORT use were 6.9 percent points lower in CO provinces over 2010–2015 than in CI provinces. The two contracting approaches also deliver similar results in terms of improvements in health systems performance, except in the case of drug availability where CO facilities showed much greater improvements in 2011–2016. Adjusted comparisons of improvements over time show that CO facilities achieved similar or greater improvements in health systems performance in 2004–2010 with the exception of drug availability. By contrast, in 2011–2016, CO facilities achieved greater improvements over time relative to CI comparators on several indicators including functioning equipment, availability of drugs and vaccines, client physical assessment and client counseling. The availability of drugs increased to a far greater extent in CO facilities than in CI facilities over this period. Even after controlling for confounders, the pace of improvements in drug availability were 8.4 percent points higher in CO facilities than in CI facilities over 2010–2015. The CO approach has clearly performed well in high and escalating conflict settings, and NGOs’ ability to respond quickly and with flexibility may explain good CO model performance. Since CI provinces are largely more secure than CO provinces no evidence is currently available on the insecurity resilience of the CI model in high insecurity settings. The CO approach may have intrinsic ben- efits that explain these findings, notably via nimble recruitment, greater mobility, timely salary payments and flexibility with staff pay and flexible/ decentralized procurement. In addition, NGOs may be better able to access and deliver services in more insecure areas. International reviews of CO on the use of health services also find that the CO approach is effective in low- and middle-income countries, particularly in under-served areas and post- insecurity settings. A recent systematic Cochrane (2009) review of the impact of CO in low- and middle-income countries (Lagarde and others 2009) finds that CO is an effective option particularly in settings where governments may have difficulties reaching populations. A literature review focused on contracting for primary care and nutrition services with broader inclusion criteria (Loevinsohn and others 2005) also concludes that contracting approaches whether contracting of service delivery IMPROVING HEALTH OUTCOMES IN AFGHANISTAN 55 (or CO) or management contracting (or CI) have achieved impressive and rapid results in terms of scaling up service delivery. Other studies in Afghanistan also point to the contribution of contracting approaches in improving coverage and equity. Arur and others (2010) found that both contracting-in and con- tracting-out approaches were associated with substantial increases in service use from 2004 to 2005 compared with non-contracted facilities, which was about 29% for outpatient visits (p < 0.01), and up to 41% (p < 0.01) for female patients, 68% for the poorest quintile (p < 0.01) and 27% for children aged under 5 years (p < 0.05). Contracting may also have the potential to reduce inequalities. Alonge and others (2015) studied levels of inequity in access between the poor and non-poor in Afghanistan. The adjusted odds of a poor client attending a health facility over time increased significantly for facilities that were contracted out, with odds ratio of 2.00 and 2.82 respectively (p-value 0.001). Evidence points to the importance of a focus on provider autonomy and results, not inputs, in delivering better results, including improvements in equity. Similar to other research find- ings from Afghanistan (Arur 2008), the Loevinsohn and others (2005) review finds that success- ful approaches tend to maximize provider autonomy, and highlights that a focus on outputs and outcomes, rather than inputs tends to lead to better results. Both the Cochrane (2009) review and Loevinsohn review, discussed here, underscore the importance of robust evaluation and results monitoring. In light of these findings, it is important to shift back to true lump sum budgets for contracted NGOs. The current contracts given to NGOs are lump-sum, but, in actual fact, NGOs say they have to seek permission from the MOPH to transfer funds between line items, which is a cumbersome and time-consuming process. This is troubling given the known benefits of provider autonomy in delivering good results, assuming that providers are held accountable for their performance, as is the case with cur- rent contracting models in Afghanistan. Substantially greater improvements in pharmaceutical and vaccine availability in CO facilities point to the importance of continuing decentralized procurement and supply chains. At the same time, there is a need to oversee drug quality through drug quality surveys and other approaches that independently assess whether drugs actually available at service delivery points meet quality standards. The Way Forward The health sector has shown itself to be innovative and dynamic with the ability to make a major contribution to overall development objectives. The ministry has achieved success on a num- ber of fronts: i. A regulatory framework has been established to engage the non-state sector in service delivery; ii. Regular data collection is implemented in over 2,000 facilities. The use of third party monitors for data collection can give life to the accountability function of the Citizens Charter; iii. Health campaigns, such as polio eradication and ending preventable maternal and newborn deaths, have been promoted and Community Development Centers (CDCs) that are empow- ered to play a leading role in improving development of their communities; and 56 PROGRESS IN THE FACE OF INSECURITY Donor funding has been aligned to employ, train, and supervise more than 2,000 nutrition iv.  female counsellors predominantly in rural areas: this is also an example of the sector’s poten- tial to contribute to creating quality jobs for women in rural areas, whilst strengthening the health system. While the track record of improvements in the health sector is heartening, there is a need to build on these improvements. Despite the improvements observed over time, absolute levels of coverage for maternal and child health services are low and health systems performance is no longer improving at the same pace. Given the pace of acceleration needed to reach the SDG targets for health, greater focus is needed on quality and scale of care to reach marginalized and remote populations. Opportunities exist to improve the performance of both contracting-out and contracting-in models Previous research in Afghanistan points to the importance of three key factors in deliver- ing results: robust and independent results monitoring; provider autonomy; and performance incentives. Previous differences among contracting-out models implemented with World Bank, Euro- pean Union, and United States Agency for International Development (USAID) funding in Afghanistan offer critical insights into the elements within contracting-out models that have driven results. Prior research into the impact of contracting-out on service utilization and quality using health facility survey data indicates that robust and independent monitoring, coupled with high degree of provider autonomy and credible links between payments to NGOs and performance, delivered the best results (Arur 2008). By contrast, additional resources without these elements failed to deliver greater improvements than in areas without any additional interventions. Effective purchasing of health services is key: this involves a greater focus on outputs and outcomes. In general, effective purchasing of health services is more critical to delivering better health results and improving value from health spending than are concerns of public versus private ownership of health service providers. In fact, many high performing health systems have largely private service provision or a mix of public and private provision with public financing (see, for example, The Commonwealth Fund’s 2015 report on international profiles of health care systems). The fundamental building block for this is the availability of good performance data and purchaser capacity to use these data to better oversee provider performance. The Afghanistan health sector generates a wealth of data: these could be better used to drive performance improvements. The health sector in Afghanistan generates a wealth of data, including third party evaluation survey data and data generated by routine reporting systems. These could be used more extensively by the MOPH and Provincial Health Departments to actively drive improvements in performance in both CO and CI areas. There is also potential to expand the role of Provincial Health Departments to provide technical support for improved service delivery and decision-making, rather than their current more limited focus on coordination and monitoring. In addition, the involvement of MOPH technical departments in monitoring service delivery could be strengthened. Strengthening citizen accountability and monitoring could improve both CO and CI perfor- mance. Findings on health systems performance improvement trends indicate that the Afghan health system has done very well on client and community engagement, and a key finding from this study is IMPROVING HEALTH OUTCOMES IN AFGHANISTAN 57 that links to communities may explain insecurity resilience. Increasing citizen involvement in mon- itoring service delivery may be a promising approach both in building insecurity resilience in service delivery and as a part of a broader democratic state-building policy agenda. Rigorous research from other settings also points to the demonstrated value of community scorecard/citizen engagement approaches in improving service delivery (Nyqvist and others 2017). Escalating insecurity makes robust and independent monitoring of contracted NGOs more difficult, and survey-based monitoring needs to be complemented with other monitoring strat- egies. Survey-based contract monitoring strategies are becoming increasingly difficult to implement in high insecurity areas, as survey teams may not be able to access some areas. This underscores the importance of complementing survey-based monitoring methods with other strategies, such as commu- nity-based monitoring strategies, as well as variants of survey methods, such as phone survey. Exploiting routinely generated data would also be valuable. 58 PROGRESS IN THE FACE OF INSECURITY REFERENCES Advancing Partners & Communities. 2014. Country Profile: Afghanistan Community Health Programs. Arlington, VA: Advancing Partners and Communities. Alonge O, Gupta S, Engineer C, Salehi AS, and Peters D. 2015. “Assessing the pro-poor effect of differ- ent contracting schemes for health services on health facilities in rural.” Afghanistan Health Policy and Planning 30:1229–1242 Afghanistan 2015 Demographic and Health Survey Available at: http://dhsprogram.com/pubs/pdf/ SR236/SR236.pdf. Agrasada, GV, J Gustafsson, E Kylberg, and U Ewald. 2005. “Postnatal peer counseling on exclu- sive breastfeeding of low-birthweight infants: a randomized, controlled trial.” Acta Paediatr. 94: 1109–1115. doi: 10.1080/08035250510025752. [1] Akseer, N, AS Salehi, et al. 2016. “Achieving maternal and child health gains in Afghanistan: a count- down to 2015 country case study.” Lancet Global Health 4 (6): e393–413. [2] Akseer, N, Z Bhatti, A Rizvi, AS Salehi, T Mashal, and ZA Bhutta. 2016. “Coverage and inequalities in maternal and child health interventions in Afghanistan.” BMC Public Health 16 (2): 797. Alkema, L., et al. 2016. “Global, regional, and national levels and trends in maternal mortality between 1990 and 2015, with scenario-based projections to 2030: a systematic analysis by the UN Maternal Mortality Estimation Inter-Agency Group.” Lancet 387 (10017): 462–474. Amudhan, S, K Mani, SK Rai, CS Pandav, and A Krishnan. 2013. “Effectiveness of demand and supply side interventions in promoting institutional deliveries—a quasi-experimental trial from rural north.” India Int J Epidemiol 42: 769–780. Arur, A. 2008. “Contracting for health services in Afghanistan-an analysis of changes in service use and quality of care.” Ph.D. Dissertation, Johns Hopkins University. Arur, A, D Peters, P Hansen, MA Mashkoor, LC Steinhardt, and G Burnham. 2010. “Contracting for Health Services and Curative Care Use in Afghanistan between 2004 and 2005.” Health Policy and Planning 25. Bailey, S, K Savage, and S O’Callaghan. 2008. Cash transfers in emergencies: A synthesis of World Vision’s experience and learning. A report commissioned by World Vision International. Monrovia, CA: World Vision International. Bailey, S, and K Hedlund. 2012. The impact of cash transfers on nutrition in emergency and transitional contexts: a review of the evidence. HPG Commissioned Reports. London, UK: Overseas Development Institute. 59 Bari, S, I Mannan, MA Rahman, GL Darmstadt, MHR Seraji, AH Baqui, S El Arifeen, SM Rahman, SK Saha, NU Ahmen, S Ahmed, M Santosham, RE Black, and PJ Winch. 2006. “Bangladesh Projahnmo-II Study Group. Trends in Use of Referral Hospital Services for Care of Sick Newborns in a Community- based Intervention in Tangail District.” J Health Popul Nutr 24: 519–529 Berg, LB. 2004. Qualitative research methods for the social science (5th ed.). Boston: Pearson Education. Bhutta, ZA. 2002. “Children of war: the real casualties of the Afghan conflict.” BMJ 324: 349. Bhutta, ZA, AK Yousafzai, and A Zipursky. 2010. “Pediatrics, War, and Children.” Current Problems in Pediatric and Adolescent Health Care 40 (2): 20–35. http://doi.org/10.1016/j.cppeds.2009.12.004. Bhutta, ZA, WJ Keenan, and S Bennett. 2016. “Children of war: urgent action is needed to save a gener- ation.” Lancet 388: 1275–1276. Boerma, JT, J Bryce, Y Kinfu, H Axelson, and CG Victora. 2008. “Mind the gap: equity and trends in coverage of maternal, newborn, and child health services in 54 Countdown countries.” Lancet 371 (9620): 1259–1267. Brazier, A. 2015. Afghan Community Engagement Project – Building community capacity to better understand and access primary health care services. Melbourne, Australia: South Eastern Melbourne Medicare Local. http://www.enliven.org.au/sites/default/files/Afghan%20Community%20Engagement% 20%26%20Health%20Literacy%20Project%20Final%20Report%20-%20SEMML%202015.pdf. Centre for Research on the Epidemiology of Disasters CRED. 2013. People Affected by Conflict: Human- itarian needs in numbers. Charmaz, K. 2006. Constructing grounded theory: A practical guide through qualitative analysis. Thousand Oaks, CA: Sage. Chowdhurym, AM, M Aminul Alam, and J Ahmed. 2006. “Development knowledge and experience: from Bangladesh to Afghanistan and beyond.” Bulletin of the World Health Organization 84 (8): 677–81. Coleman, I, and GT Lemmon. 2011. “Maternal Health in Afghanistan: Improving Health and Strength- ening Society.” Working Paper, New York, NY: Council on Foreign Relations. Cooper, PJ, M Tomlinson, L Swartz, M Landman, C Molteno, A Stein, K McPherson, and L Murray. 2009. “Improving quality of mother-infant relationship and infant attachment in socioeconomically deprived community in South Africa: randomised controlled trial.” BMJ (Online) 338: 997. Cottingham, J, C García-Moreno, and C Reis. 2008. “Sexual and reproductive health in conflict areas: The imperative to address violence against women.” BJOG: An International Journal of Obstetrics and Gynaecology 115 (3): 301–303. http://doi.org/10.1111/j.1471-0528.2007.01605.x. Creswell, JW. 2007. Qualitative inquiry and research design: Choosing among five traditions (2nd ed.). Thousand Oaks, CA: Sage. Creswell, JW. 2013. Qualitative inquiry and research design: Choosing among five approaches. Los Angeles, CA: Sage. Denzin, NK, and YS Lincoln, eds. 2005. The Sage handbook of qualitative research. Thousand Oaks, CA: Sage. Edward, A, C Branchini, I Aitken, M Roach, K Osei-Bonsu, and SH Arwal. 2015. “Toward universal coverage in Afghanistan: a multi-stakeholder assessment of capacity investments in the community health worker system.” Social Science & Medicine 145: 173–83. Engineer, CY, E Dale, A Agarwal, A Agarwal, O Alonge, A Edward, S Gupta, HB Schuh, G Burnham, and DH Peters. 2016. “Effectiveness of a pay-for-performance intervention to improve maternal and 60 PROGRESS IN THE FACE OF INSECURITY child health services in Afghanistan: a cluster-randomized trial.” International Journal of Epidemiology 45 (2): 451–459. European Union. 2005. The National Risk and Vulnerability Assessment. https://www.wfp.org/content/ afghanistan-national-risk-and-vulnerability-assessment-2005–2007. ———. 2009. The National Risk and Vulnerability Assessment 2007–08. http://documents.wfp.org/ stellent/groups/public/documents/ena/wfp213398.pdf?iframe. ———. 2014. The National Risk and Vulnerability Assessment 2011–12. http://www.af.undp.org/ content/dam/afghanistan/docs/MDGs/NRVA%20REPORT-rev-5%202013.pdf. Ghobarah, HA, P Huth, and B Russet. 2004. “The post-war public health effects of civil conflict.” Soc Sci Med 59 (4): 869–884. Gilmore, B, and E McAuliffe. 2013. “Effectiveness of Community health workers delivering preven- tive interventions for maternal and child health in low- and middle-income countries: a systematic review.” BMC Public Health 13: 847. doi: 10.1186/1471-2458-13-847. Glesne, C. 1999. Becoming qualitative researchers: An introduction (2nd ed.). New York, NY: Longman. Grittner, MA. 2013. “Results-based Financing: Evidence from performance-based financing in the health sector.” Discussion paper, Deutsches Institut für Entwicklungspolitik. Guba, EG and YS Lincoln. 1989. Fourth generation evaluation. Thousand Oaks, CA: Sage. ———. 1994. “Competing paradigms in qualitative research.” In Qualitative research in nursing (3rd ed.) edited by I Holloway and S Wheeler. Oxford, UK: Blackwell Science. Hall, S. 2015. “Improving cash-based interventions Multipurpose cash grants and protection: Enhanced Response Capacity Project 2014–2015. The Impact of Cash Transfer Programmes on Protection Outcomes in Afghanistan.” Report commissioned by Norwegian Refugee Council (NRC) and UN Refugee Agency. ———. 2015. “A practical guide for practitioners of cash-based interventions in Afghanistan.” Report commissioned by Norwegian Refugee Council (NRC), UN Refugee Agency and Humanitarian Aid and Civil Protection. Haver, J, W Brieger, J Zoungrana, N Ansari, and J Kagoma. 2015. “Experiences engaging community health workers to provide maternal and newborn health services: Implementation of four programs.” International Journal of Gynecology & Obstetrics. 130: S32–39. Health Policy Project. 2015. A Health Insurance Feasibility Study in Afghanistan: Learning from Other Countries, a Legal Assessment, and a Stakeholder Analysis. Washington, DC: Futures Group, Health Policy Project. Hill, Z, M Dumbaugh, L Benton, K Källander, D Strachan, A ten Asbroek, J Tibenderana, B Kirkwood, and S Meek. 2014. “Supervising community health workers in low-income countries–a review of impact and implementation issues.” Glob Health Action 7 (1). doi:10.3402/gha.v7.24085. Huber, D, N Saeedi, and AK Samadi. 2010. “Achieving success with family planning in rural Afghan- istan.” Bulletin of the World Health Organization 88 (3): 227–31. Islamic Transitional Government of Afghanistan Ministry of Health. 2003. Policy on Community Based Health Worker (CHW). Jabbour, S. 2014. “Health and contemporary change in the Arab world.” Lancet 383 (9915): 477–479. http://doi.org/10.1016/S0140-6736(13)62567-2. Johns Hopkins School of Public Health. 2004–2016. BSC reports. Baltimore, MD: Johns Hopkins, School of Public Health. IMPROVING HEALTH OUTCOMES IN AFGHANISTAN 61 ———. Afghanistan Health survey 2012–13. Baltimore, MD: Johns Hopkins School of Public Health. Kruk, ME, LP Freedman, GA Anglin, and RJ Waldman. 2009. “Rebuilding health systems to improve health and promote statebuilding in post-conflict countries: a theoretical framework and research agenda.” Soc Sci Med 70 (1): 89–97. doi: 10.1016/j.socscimed.2009.09.042. Kvale, S. 1996. Interviews: An introduction to qualitative research interviewing. London, UK: Sage. KIT Royal Tropical Institute. 2015. An Impact Evaluation of the Results-based Financing Intervention in Afghanistan. FINAL REPORT 2015. Amsterdam, the Netherlands: KIT Royal Tropical Institute. http://www.kit.nl/health/wp-content/uploads/sites/4/2016/09/RBF-IE-2015-Final-Report.pdf. Lagarde, M, and N Palmer. 2009. “The impact of contracting out on health outcomes and use of health services in low and middle-income countries.” Cochrane Database of Systematic Reviews 4 (CD008133). doi: 10.1002/14651858.CD008133. Lin, A, and AS Salehi. 2013. “Stimulating demand: effects of a conditional cash transfer programme on increasing maternal and child health-service utilisation in Afghanistan, a quasi-experimental study.” Lancet 381: S84. Liu, L, S Oza, D Hogan, et al. 2016. “Global, regional, and national causes of under-5 mortality in 2000–15: an updated systematic analysis with implications for the Sustainable Development Goals.” Lancet 388 (10063): 3027–3035. http://dx.doi. org/10.1016/S0140-6736(16)31593-8. Lincoln, YS, and EG Guba. 2000. Paradigmatic controversies, contradictions, and emerging confluences. In Handbook of qualitative research, edited by NK Denzin and Y.S Lincoln, 163–188. Thousand Oaks, CA: Sage. Loevinsohn, B, and A Harding. 2005. “Buying results? Contracting for health service delivery in developing countries.” Lancet 366: 676–681. Mayhew, M, P Ickx, H Stanekzai, T Mashal, and W Newbrander. 2014. “Improving nutrition in Afghan- istan through a community-based growth monitoring and promotion programme: A pre–post eval- uation in five districts.” Global public health 9 (sup1): S58–75. Merriam, S. 2009. Qualitative research: A guide to design and implementation. San Francisco, California: Jossey-Bass. Ministry of Public Health Afghanistan. 2011. Afghanistan National Health Workforce Plan 2012–16. Kabul, Afghanistan: Ministry of Public Health Afghanistan. Mowafi, H. 2011. “Conflict, displacement and health in the Middle East.” Global Public Health 6 (5): 472–487. http://doi.org/10.1080/17441692.2011.570358. Najafizada, SA, R Labonté, and IL Bourgeault. 2014. “Community health workers of Afghanistan: a qualitative study of a national program.” Conflict and Health 8 (1): 1. Nyqvist, M, D de Walque, and J Svensson. 2017. “Experimental evidence on the Long-run Impact of Community-Based Monitoring.” American Economic Journal: Applied Economics 9 (1): 33–69. Patton, MQ. 2002. Qualitative research and evaluation methods (3rd ed). Thousand Oaks, CA: Sage Publication. Perry, H, R Zulliger, K Scott, D Javadi, J Gergen, and K Shelley. 2014. Case studies of large-scale commu- nity health worker programs: examples from Afghanistan, Bangladesh, Brazil, Ethiopia, India, Indone- sia, Iran, Nepal, Pakistan, Rwanda, Zambia and Zimbabwe. Developing and strengthening community health worker programs at scale: a reference guide and case studies for program managers and policymakers. Washington, DC: Maternal and Child Health Integrated Program (MCHIP). 62 PROGRESS IN THE FACE OF INSECURITY Perry, HB, R Zulliger, and MM Rogers. 2014. “Community health workers in low-, middle-, and high-income countries: an overview of their history, recent evolution, and current effectiveness.” Annual review of public health 18 (35): 399–421. Rao, KD, H Waters, L Steinhardt, S Alam, P Hansen, and AJ Naeem. 2009. “An experiment with community health funds in Afghanistan.” Health Policy and Planning 24 (4): 301–311. Reichenbach, L, and SN Shimul. 2011. “Sustaining health: the role of BRAC’s community health volunteers in Bangladesh, Afghanistan and Uganda.” Dhaka, Bangladesh: BRAC Research and Evaluation Division. Richards, L. 2000. Using NVivo in qualitative research (2nd ed.). Melbourne, Australia: QSR International. Robertson, L, P Mushati, JW Eaton, et al. 2013. “Effects of unconditional and conditional cash transfers on child health and development in Zimbabwe: a cluster-randomised trial.” Lancet 381: 1283–1292. Samuel Hall Consulting. 2014. “Humanitarian Assistance through Mobile Cash Transfer in Northern Afghanistan: An Evaluation of a DFID Pilot Project in Faryab, Jawzjan, and Samangan.” London, UK: Department for International Development. Save the Children. 2014. State of the World’s Mothers 2014: Saving Mothers and Children in Humanitarian Crises. London, UK: Save the Children. Retrieved from www.savethechildren.org. Sheth, M, and M Obrah. 2004. “Diarrhea prevention through food safety education.” Indian J Pediatr 71: 879–882. doi: 10.1007/BF02830824. Stake, RE. 1995. The Art of Case Study Research. Thousand Oaks, CA: Sage Publications. Standing, G. “How cash transfers boost work and economic security.” Trabajo preliminar, 58. Strauss, A, and J Corbin. 1990. Basics of qualitative research: Grounded theory procedures and techniques. Newbury Park, CA: Sage. UNFPA. Mid term review of FHH model Afghanistan. New York, NY: United Nations Population Fund. ———. 2014. State of Afghanistan’s Midwifery 2014. New York, NY: United Nations Population Fund. ———. 2016. Humanitarian Action 2016. New York, NY: United Nations Population Fund. Overview http://www.safebirthevenhere.org/sites/default/files/pub-pdf/16-150_UNFPA_Humanitarian_ 2016_Overview_Final_Sheet_Final_Web_version.pdf. UNICEF. 2010. MICS. New York, NY: United Nations International Children’s Emergency Fund. http://www.cso.gov.af/Content/files/AMICS-Jun24-2012-FINAL.pdf. ———. 2011. UNICEF Annual Report 2010. New York, NY: United Nations International Children’s Emergency Fund. ———. 2013. National Immunization Coverage Survey Afghanistan. New York, NY: United Nations Inter- national Children’s Emergency Fund.UNHCRC. 2012. An Introduction to Cash-Based Interventions in UNHCRC Operations. New York, NY: United Nations High Commissioner for Refugees. ———. 2013. National Nutrition Survey. Afghanistan 2013. New York, NY: United Nations International Children’s Emergency Fund. http://reliefweb.int/sites/reliefweb.int/files/resources/Report%20 NNS%20Afghanistan%202013%20( July%2026-14).pdf. ———. 2015. Assessing the Community Health Worker Role in the Polio Eradication Initiative’s Social Mobilisation Strategy. A report for the Afghanistan Ministry of the Public Health. New York, NY: United Nations International Children’s Emergency Fund. UNICEF. 2003. MICS. New York, NY: United Nations International Children’s Emergency Fund. http://catalog.ihsn.org/index.php/catalog/746. IMPROVING HEALTH OUTCOMES IN AFGHANISTAN 63 USAID. Afghanistan Demographic and Health Survey 2015. Central Statistics Organization. Min- istry of Public Health. Rockville, Maryland: The DHS Program, ICF International. https://www. dhsprogram.com/pubs/pdf/PR77/PR77.pdf. ———. 2012. World Vision Afghanistan. USAID Title II Multi-Year Assistance Program. Health and Livelihoods Initiative in Ghor. End of Project Evaluation Report. Waldman, R, L Strong, and A Wali. 2006. Afghanistan’s health system since 2001: condition improved, prognosis cautiously optimistic. Kabul, Afghanistan: Research and Evaluation Unit. Witvorapong, N, and AI Foshanji. 2016. “The impact of a conditional cash transfer program on the uti- lization of non-targeted services: Evidence from Afghanistan.” Social Science & Medicine 152: 87–95. World Bank. 2009. Taking stock: World Bank experience with results-based financing (RBF) for health. A World Bank Study. Washington, DC: World Bank. ———. 2010. Getting health results in Afghanistan. A World Bank Study. Washington, DC: World Bank. ———. 2011. Setting Targets in Health Nutrition and Population Projects. A World Bank Study. Washington, DC: World Bank. http://siteresources.worldbank.org/HEALTHNUTRITIONAND POPULATION/Resources/281627-1095698140167/TargetSettinginHNPProjects.pdf. ———. 2013. Using Results-Based Financing to Achieve Maternal & Child Health. A World Bank Study. Washington, DC: World Bank. ———. 2014. Critical Administrative Constraints to Service Delivery: Improving Public Services in Afghan- istan’s Transformational Decade. A World Bank Study. Washington, DC: World Bank. ———. 2015. Verification of Performance in Results-Based Financing (RBF): The Case of Afghanistan. A World Bank Study. Washington, DC: World Bank. ———. 2017. Social Service Delivery in Violent Contexts: Achieving results against the Odds. A World Bank Study. Washington, DC: World Bank. World Health Organization. 2015. Humanitarian Health Action. Geneva, Switzerland: World Health Organization. http://www.who.int/hac/crises/ssd/releases/who_suistainable_development_ goals/en/. Yin, RK. 2003. Case study research: Design and methods (3rd ed). Thousand Oaks, CA: Sage. ———. 2009. Case study research: Design and methods. Applied social research methods series. (4th ed). Thousand Oaks, CA: Sage You, D, et al. 2015. “Global, regional, and national levels and trends in under-5 mortality between 1990 and 2015, with scenario-based projections to 2030: a systematic analysis by the UN Inter-agency Group for Child Mortality Estimation.” Lancet 386 (10010): 2275–2286. 64 PROGRESS IN THE FACE OF INSECURITY