King Holmes, MD, PhD. University Consortium for Global Health. Sept. 15, 2009.
Understanding the Effect of the GAVI Initiative on Reported Vaccination Coverage Levels
1. Accountability, transparency and corruption in global health: the critical role of health metrics and evaluation Stephen S Lim Assistant Professor of Global Health
2. 2 Outline Context Two examples Tracking immunization coverage Conditional cash transfers to women for delivering in a health facility
3. 3 Outline Context Two examples Tracking immunization coverage Conditional cash transfers to women for delivering in a health facility
5. 5 Global goals, funders and initiatives Goals 1978: Alma-Ata: Primary Health Care and Health for All 1984: Universal childhood immunization by 1990 2000: Millennium Development Goals 2003: 3 by 5 – 3 million people on antiretroviral drugs by 2005 2008: Malaria Elimination Private philanthropic organizations 2000: Bill and Melinda Gates Foundation 2006: Warren Buffet pledged additional US$30 to BMGF Global Health Initiatives 2000: Global Alliance for Vaccines and Immunizations (GAVI) 2002: Global Fund for Aids, Tuberculosis and Malaria (GFATM) 2003: US President’s Emergency Plan for AIDS Relief (PEPFAR) 2005: US President’s Malaria Initiative (PMI) 2007: International Health Partnership (IHP+)
6. 6 National health sector reforms and programs 2001: Thailand Universal Coverage 30 Baht Scheme 2003/04: Mexico Health Sector Reforms System of Social Protection in Health including Seguro Popular 2005: Indian National Rural Health Mission Conditional cash transfers for women to give birth in health facilities
7. Critical role of health metrics and evaluation High-quality measurement of health indicators and evaluation of programs is central to issues of transparency and accountability Are resources being used effectively? Have initiatives and reforms led to improvements in health system delivery and population health? Mounting pressure from funders, civil society, etc to document this Economic crisis has led to rising fears that development assistance budgets will be cut 7
8. 8 Outline Context Two examples Tracking immunization coverage Conditional cash transfers to women for delivering in a health facility
9. 9 Tracking childhood immunization coverage Substantial resources have been invested in delivering immunization services Basic vaccines, e.g. three-dose diptheria, pertussis and tetanus (DTP3) as well as new vaccines, e.g. HiB, rotavirus Global initiatives 1984: Universal Childhood Immunization (UCI) by 1990 initiative, defined as 80% immunization coverage 1999: Global Alliance on Vaccines and Immunizations (GAVI) Monitoring the extent of immunization delivery is critical for evaluating how effective these investments and initiatives have been
10. 10 Tracking childhood immunization coverage GAVI’s Immunization Services Support (ISS) is the funding that aims to increase coverage of basic vaccines such as three-dose diptheria, tetanus and pertussis (DTP3) vaccination. ISS payments are performance-based with funds disbursed in proportion to the number of additional children targeted or reported to receive DTP3. Number of additional children receiving DTP3 is based on official reports from countries to WHO and UNICEF.
11. 11 Tracking childhood immunization coverage Two main questions: What is the trend in the fraction of children receiving three-dose diptheria, tetanus and pertussis vaccination (DTP3 coverage) over the period 1986 to 2006? Do target-oriented initiatives such as universal childhood immunization (UCI) and results-based financing initiatives such as GAVI’s Immunization Services Support (ISS) lead to over-reporting of DTP3 immunization coverage?
12. 12 Data sources Micro-data from standardized multi-country surveys DHS, MICS, CDC Crude coverage: three DTP vaccinations by maternal recall or card documented, irrespective of vaccine schedule Estimated for each birth cohort (up to 5 years prior to the survey) 225 surveys Survey reports and WHO/UNICEF database: 78 multi-country surveys 142 country-specific surveys with sample size reported 145 country-specific surveys without sample size reported Administrative data estimates based on health service provider registries Reported routinely to WHO and UNICEF since 1990 Officially reported estimates since 1980
13. 13 Quick review of Immunization Services Support (ISS) Performance-based payment Number of additional children reported by countries to have received DTP3 Reports largely based on administrative data Baseline is the year prior to approval of the proposal US$20 is paid once per additional child Data quality audit (DQA) of administrative data system before reward payments commence (from Year 3)
18. 18 Estimating missing survey coverage Two purposes: Generate plausible estimates over time to allow monitoring of indicators Reduce compositional bias in in causal inference that can stem from missing data We use validated statistical approaches that are Objective Replicable Characterizes uncertainty
20. 20 Global trends in DTP3 coverage Survey-based global coverage of DTP3 (black) with 95% uncertainty estimates compared to countries’ officially reported (red) and WHO and UNICEF estimates (blue), 1986 to 2006.
22. 22 Does ISS lead to over-reporting? Statistical analysis of over-reporting (officially reported coverage minus survey coverage) by years since the GAVI ISS baseline
23. 23 GAVI Immunization Services Support (ISS) Number of additional children vaccinated in 51 countries receiving ISS funding up to the year 2006 : Based on official reports: 13.9 million Survey-based: 7.3 (5.5 to 9.2) million ISS payments Based on official reports: US$289 million Survey-based: US$148 million Over-reporting is not uniform 4 countries that reported increases, number of additional children did not increase 6 overestimated by > 4x 10 overestimated by > 2x but ≤ 4x 23 overestimated by > 1x but ≤ 2x 8 countries underestimated
24. 24 Implications At the global level, survey-based immunization coverage has increased continuously and gradually over the last 20 years Reflects time and investment needed to expand health services Improvements more pronounced in some regions (e.g. Central, West sub-Saharan Africa) and countries (e.g. Cameroon) during recent time period Targets and payments for performance such as GAVI’s ISS can incentivize improvements but also lead to over-reporting Over-reporting likely reflects pressures throughout the reporting system to meet targets Monitoring and evaluation systems need to be based on independent, rigorous, empirical measurements that are robust to these effects
25. 25 Outline Context Two examples Tracking immunization coverage Conditional cash transfers to women for delivering in a health facility
26. JananiSurakshaYojana – “Safe motherhood scheme” Launched in 2005; 100% centrally funded Goal: reduce maternal and neonatal mortality Works by incentivizing women to deliver in a health facility Implemented through Accredited Social Health Activists (ASHAs) ASHAs also receive a cash benefit Budget allocation of US$342 million in 2009-10 26 Accredited Social Health Activist (ASHA) Madhya Pradesh, India Photo: Department for International Development, UKAID
27. National guidelines, Eligibility In 10 high-focus states All pregnant women delivering in government facility or accredited private institutions Other states & home deliveries Below the Poverty Line >19 years of age First two live births Targeted to women from scheduled caste or tribe 27
29. Questions What is the level of implementation of JSY at district and state-levels? Is JSY reaching its intended beneficiaries? Does receipt of financial assistance under JSY lead to increased antenatal care and in-facility delivery and reduced perinatal, neonatal and maternal mortality? 29
30. Data India District-level Household Surveys (DLHS) DLHS-2: ~1,000 households from 593 districts, 2002 to 2004 DLHS-3: 1,000 to 1,500 households from 611 districts, late 2007 to early 2009 Ever-married women aged 15 to 44, for most recent pregnancy Antenatal care (no. of visits) Delivery care (type of provider, location) Outcome (live birth, still birth, spontaneous or induced abortion) Survival of the child in the case of a live birth Receipt of financial assistance under JSY (DLHS-3) Individual and household characteristics, e.g. asset-based wealth, caste, education, location of residence and distance to facility 30
34. Evaluating impact of JSY on coverage and outcomes Exact matching Match births receiving JSY to those not receiving JSY in DLHS-3 Matching covariates: urban/rural residence, BPL card ownership, wealth quintile, caste, education, parity, and maternal age Logistic regression on matched data allows more precise control for confounders With-vs-without Logistic regression, comparing births receiving JSY to births that did not receive JSY in DLHS-3 and all births in DLHS-2 District-level differences-in-differences Compare districts by level of JSY uptake, controlling for baseline differences (DLHS-2) 580 district aggregates from DLHS-2 to DLHS-3. 34
35. Outcomes Intervention coverage Antenatal care with at least three visits In-facility birth Skilled birth attendance (in-facility birth or birth outside of a facility with a skilled attendant) Mortality Perinatal death (stillbirth or death up to and including 7 days after a live birth) Neonatal death (death up to and including 1 month after a live birth) Maternal mortality* (death of women aged 15 to 49 during pregnancy or up to 6 weeks after birth or termination) * In district-level analysis only 35
36. Potential confounders Controlled for: maternal age; number of live births; birth interval; single or multiple birth; maternal education; household wealth based on asset ownership; caste/tribe; religion; and location of residence with respect to distance to the nearest health facility Varied using district, state-level fixed and random effects 36
37. JSY and intervention coverage, national level Change in probability of receiving intervention: JSY vs no JSY
38. Impact on mortality, national level Change in probability of death: JSY vs no JSY
41. Implications Varied uptake of JSY across states; not reaching the very poor Increases in ANC coverage and intra-partum care coverage Likely reductions in perinatal and stillbirth/neonatal mortality But potential quality of care issues in high-focus states indicates Alternative monitoring approach needed for maternal mortality Continued monitoring and evaluating the program is critical 41
42. 42 Summary Substantial resources are being directed towards improving population health Need to track in a valid, reliable and comparable way health indicators and evaluate the impact of programs Ensure that increased resources for health are being utilized intended purpose and are making a difference to the health of populations Increasing relevance during a time of global financial crisis Independent and empirically-based monitoring of health indicators and evaluation of programs
Notes de l'éditeur
Skilled birth attendance (SBA) coverage improved from 33% in 1990 to 48% in 2005Maternal mortality ratio declined from around 500 per 100,000 live births in 1990 to around 260 in 2005Neonatal mortality rate decreased from 54 per 1,000 live births in 1990 to 29 in 2005
Consistent results across three analytical approachesTranslation: For every 10 women receiving JSY, 1 additional woman will complete 3 ANC visits, 4 to 5 additional women will give birth in-facility, 3 to 4 women will give birth in a facility or with a skilled attendant
Less consistent results across three analytical approaches for mortality impacts. Individual level analysis suggest reduction of around 3 to 6 perinatal deaths per 1,000 births for every woman receiving JSY; similar magnitude for still birth or neonatal mortality. District-level results suggest larger effects but are not significantly different from zero change for perinatal, stillbirth/neonatal mortality. We also see no effect on maternal mortality. Two possible explanations:Small numbers problem in the district-level numbers. District analysis removes bias associated with selective uptake of JSY that the individual-level analyses are not controlling for. Given the consistent results across the three analytical approaches for intervention coverage that is not prone to small numbers problems we think the former is more likely but cannot say definitively. Maternal mortality a big problem – need an alternative approach to measure impact
Much larger effects for in-facility birth/SBA in high-focus states, compared to north-east states compared to other states
Examining differences by high-focus states vs other states shows smaller effects in high-focus states compared to other states. Two possible explanations:Different targeting of women in high-focus compared to non-high-focus states. In high-focus states all women are eligible for JSY, whereas in non-high-focus states only women living below the poverty line are eligible. As a result, the former group may include lower risk women where the benefit of the program may be smaller. Another possible explanation is that high-focus states have lower quality of obstetric care or are less likely to be able to cope with the increased workloads that have resulted from the large increases of in-facility deliveries due to JSY.