Skip to main content
. 2024 Jan 18;4(1):e0002467. doi: 10.1371/journal.pgph.0002467

Table 3. Estimates of PEPFAR impact on all-cause mortality (deaths per 1,000), cumulative 5-year estimates.

All-Cause Mortality Rate All PEPFAR COP PEPFAR
2004–2008 2004–2013 2004–2018 2004–2008 2004–2013 2004–2018
Model 1. DID model with no covariates, 1990–2018 -0.945** -1.567*** -2.090*** -0.996* -2.016*** -2.883***
Model 2. DID model with non-financial BL covariates only, 1990–2018 a -1.081*** -1.677*** -2.206*** -1.205** -2.207*** -3.086***
Model 3. DID model with non-financial and financial BL covariates, 1990–2018 b -1.027*** -1.629*** -2.157*** -1.152*** -2.157*** -3.036***
Model 4. DID model with non-financial BL covariates and yearly other donor spending on health covariate, 2002–2018 -0.721* -1.294*** -1.814*** -0.961* -1.939*** -2.809***
Model 5. DID model with non-financial BL covariates and yearly other donor spending and domestic spending on health covariates, 2002–2016 -0.221 -0.691* -1.072*** -0.340 -1.215** -1.854***
Approximate % reduction in ACM c 2.1–10.3%
6.6–16.0%
10.9%
10.2–21.0%
2.7–9.6%
9.6–17.5%
14.7–24.5%
Incremental average d 7.6% 5.5% 4.7% 7.4% 7.7% 6.6%

Notes: ACM = All-cause mortality (rate of deaths per 1,000); BL = baseline (2004); COP = country operating plans; DID = difference-in-difference; PEPFAR = President’s Emergency Plan for AIDS Relief.

a Baseline (2004) non-financial covariates include HIV prevalence rate (% of population ages 15–49), GDP per capita (constant USD), population size, percent of urban population, secondary school enrollment (% gross), life expectancy at birth, fertility rate (births per woman), whether the country was a low-income or a middle-income country (dummy), and whether the country received HIV aid from the U.S. prior to 2004 (dummy).

b Baseline (2004) financial covariates include other donor health spending and domestic health spending.

c For each model the percent is calculated by dividing the coefficient by the mean of ACM in 2004 (10.5 for ALL, and 12.6 for COP). The range is formed by taking the lowest and highest percentages across the five models

d The incremental impact for the first period is the average impact across the 5 models for 2004–2008. The incremental impact for the second 5-year period is computed by subtracting the average impact across models in the first period (2004–2008) from the estimated average impact across models in the second period (2004–2013). The third period incremental impact is computed by subtracting the average impact across models in the second period (2004–2013) from the estimated average impact across models in the third period (2004–2018)

***p < 0.001

**p < 0.01

* p< 0.05.