The US President's Malaria Initiative and under-5 child mortality in sub-Saharan Africa: A difference-in-differences analysis

Abstract

Background: Despite substantial financial contributions by the United States President's Malaria Initiative (PMI) since 2006, no studies have carefully assessed how this program may have affected important population-level health outcomes. We utilized multiple publicly available data sources to evaluate the association between introduction of PMI and child mortality rates in sub-Saharan Africa (SSA).

Methods and findings: We used difference-in-differences analyses to compare trends in the primary outcome of under-5 mortality rates and secondary outcomes reflecting population coverage of malaria interventions in 19 PMI-recipient and 13 non-recipient countries between 1995 and 2014. The analyses controlled for presence and intensity of other large funding sources, individual and household characteristics, and country and year fixed effects. PMI program implementation was associated with a significant reduction in the annual risk of under-5 child mortality (adjusted risk ratio [RR] 0.84, 95% CI 0.74-0.96). Each dollar of per-capita PMI expenditures in a country, a measure of PMI intensity, was also associated with a reduction in child mortality (RR 0.86, 95% CI 0.78-0.93). We estimated that the under-5 mortality rate in PMI countries was reduced from 28.9 to 24.3 per 1,000 person-years. Population coverage of insecticide-treated nets increased by 8.34 percentage points (95% CI 0.86-15.83) and coverage of indoor residual spraying increased by 6.63 percentage points (95% CI 0.79-12.47) after PMI implementation. Per-capita PMI spending was also associated with a modest increase in artemisinin-based combination therapy coverage (3.56 percentage point increase, 95% CI -0.07-7.19), though this association was only marginally significant (p = 0.054). Our results were robust to several sensitivity analyses. Because our study design leaves open the possibility of unmeasured confounding, we cannot definitively interpret these results as causal.

Conclusions: PMI may have significantly contributed to reducing the burden of malaria in SSA and reducing the number of child deaths in the region. Introduction of PMI was associated with increased coverage of malaria prevention technologies, which are important mechanisms through which child mortality can be reduced. To our knowledge, this study is the first to assess the association between PMI and all-cause child mortality in SSA with the use of appropriate comparison groups and adjustments for regional trends in child mortality.

Fig 1. Time trends in development assistance for malaria and coverage of malaria interventions in sub-Saharan Africa.

Total assistance for malaria, in 2014 US dollars, was divided into 3 categories: (1) US bilateral aid for malaria, a proxy for PMI disbursements; (2) Global Fund, limited here to malaria disbursements; (3) All other malaria aid includes total malaria aid minus US bilateral aid minus GF malaria aid. Data Sources: Development Assistance for Health (DAH) data from 1995–2012 were obtained from Institute for Health Metrics and Evaluation. Population coverage of ITNs, ACTs, and IRS from 2000–2015 were obtained from the Malaria Atlas Project. Abbreviations: ACTs, estimated proportion of cases of fever in under-5 year olds that were treated with artemisinin combination therapy; IRS, estimated proportion of the population protected by indoor residual spraying of insecticides; ITNs, estimated proportion of people who slept under an insecticide-treated bednet on any given night; PMI, President’s Malaria Initiative.

Fig 2. Adjusted risk ratios of child mortality and adjusted percentage changes in population coverage of malaria interventions as a function of year of PMI program implementation.

Risk ratios of child mortality were estimated using modified Poisson regression model controlling for a set of indicators of the year of PEPFAR implementation, a set of indicators of the year of Global Fund implementation, individual-level covariates, country and year fixed effects (Model 9 in S1 Appendix). Standard errors were clustered at the country level. Error bars represent 95% confidence intervals. Changes in ITN, ACT, and IRS coverage were obtained using OLS regression models, controlling for a set of indicators of the year of PEPFAR implementation, a set of indicators of the year of Global Fund implementation, individual-level covariates, country and year fixed effects (Model 10) and robust standard errors. PMI year 9 was omitted from the figure because too few observations were available for the calculations (full list of coefficients and confidence intervals is listed in Table G in S1 Appendix). Data Sources: Demographic and Health Surveys from 1995–2014; Malaria Atlas Project from 2000–2014. Abbreviations: ACTs, estimated proportion of cases of fever in under-5 year olds that were treated with Artemisinin Combination Therapy; IRS, estimated proportion of the population protected; ITNs, estimated proportion of people who slept under an insecticide-treated bednet on any given night; PMI, President’s Malaria Initiative; RR, adjusted risk ratio.