Evaluating the impact of two decades of USAID interventions and projecting the effects of defunding on mortality up to 2030: a retrospective impact evaluation and forecasting analysis

Source: Lancet
Type: news-reporting
Author: Daniella Medeiros Cavalcanti, PhD ∙ Lucas de Oliveira Ferreira de Sales, PhD ∙ Andrea Ferreira da Silva, PhD ∙ Elisa Landin Basterra, MSc ∙ Daiana Pena, MSc ∙ Caterina Monti, MA ∙ Gonzalo Barreix, MSc ∙ Natanael J Silva, MSc ∙ Paula Vaz, PhD ∙ Francisco Saute, PhD ∙ Gonzalo Fanjul, MPA ∙ Prof Quique Bassat, PhD ∙ Prof Denise Naniche, PhD ∙ Prof James Macinko, PhD ∙ Prof Davide Rasella, PhD

Source Text

Excerpt only. The source text contains graphs and charts that I can’t reproduce here. Please visit the actual source for more information:

Summary

Background

The US Agency for International Development (USAID) is the largest funding agency for humanitarian and development aid worldwide. The aim of this study is to comprehensively evaluate the effect of all USAID funding on adult and child mortality over the past two decades and forecast the future effect of its defunding.

Methods

In this retrospective impact evaluation integrated with forecasting analysis, we used panel data from 133 countries and territories— including all low-income and middle-income countries (LMICs)—with USAID support ranging from none to very high. First, we used fixed-effects multivariable Poisson models with robust SEs adjusted for demographic, socioeconomic, and health-care factors to estimate the impact of USAID funding on all-age and all-cause mortality from 2001 to 2021. Second, we evaluated its effects by age-specific, sex-specific, and cause-specific groups. Third, we did several sensitivity and triangulation analyses. Lastly, we integrated the retrospective evaluation with validated dynamic microsimulation models to estimate effects up to 2030.

Findings

Higher levels of USAID funding—primarily directed toward LMICs, particularly African countries—were associated with a 15% reduction in age-standardised all-cause mortality (risk ratio [RR] 0·85, 95% CI 0·78–0·93) and a 32% reduction in under-five mortality (RR 0·68, 0·57–0·80). This finding indicates that 91 839 663 (95% CI 85 690 135–98 291 626) all-age deaths, including 30 391 980 (26 023 132–35 482 636) in children younger than 5 years, were prevented by USAID funding over the 21-year study period. USAID funding was associated with a 65% reduction (RR 0·35, 0·29-0·42) in mortality from HIV/AIDS (representing 25·5 million deaths), 51% (RR 0·49, 0·39–0·61) from malaria (8·0 million deaths), and 50% (RR 0·50, 0·40–0·62) from neglected tropical diseases (8·9 million deaths). Significant decreases were also observed in mortality from tuberculosis, nutritional deficiencies, diarrhoeal diseases, lower respiratory infections, and maternal and perinatal conditions. Forecasting models predicted that the current steep funding cuts could result in more than 14 051 750 (uncertainty interval 8 475 990–19 662 191) additional all-age deaths, including 4 537 157 (3 124 796–5 910 791) in children younger than age 5 years, by 2030.

Interpretation

USAID funding has significantly contributed to the reduction in adult and child mortality across low-income and middle-income countries over the past two decades. Our estimates show that, unless the abrupt funding cuts announced and implemented in the first half of 2025 are reversed, a staggering number of avoidable deaths could occur by 2030.

Funding

The Spanish Ministry of Science and Innovation, UK Medical Research Council, and EU Horizon Europe.

Introduction

For the past 20 years, the USA has been the leading government donor to humanitarian response plans, development aid, and multilateral development banks, mainly through the US Agency for International Development (USAID). USAID was established in 1961 as an independent agency in the executive branch under the direct authority and guidance of the Secretary of State. 1 The agency’s aim was twofold: the first was to provide humanitarian assistance, and the second to also assist and support economic growth and self-resilience of developing countries, especially those deemed strategic for the US economic and geopolitical impact. 2 In each mission and country where USAID has been operational, the agency engaged with diverse partners, such as central governments, private entities, local organisations, and international and national non-governmental organisations operating mostly bilaterally. Over the years, and despite a relatively modest effort in relation to its national wealth, the USA’s importance as a donor for development and humanitarian aid has overshadowed any other donor. Although the USA has indeed been the largest donor in absolute terms—providing more than US$55 billion in official development assistance (ODA) in 2023 and accounting for approximately 30% of total Development Assistance Committee (DAC) countries’ ODA—it ranked only 25th out of 30 DAC members in terms of ODA relative to national income, allocating just 0·24% of its gross national income. By contrast, countries such as Norway (1·09%) and Luxembourg (1·00%) exceeded the 0·7% target of the UN, reflecting a substantially higher proportional commitment to international development.

Research in context

Evidence before this study

Despite the US Agency for International Development (USAID) being the world’s leading donor for humanitarian and development aid, there is scarce evidence in the literature assessing its impact on global health. Few evaluations have attempted to estimate the effects of USAID funding on maternal and child mortality in selected low-income and middle-income countries (LMICs), and some reports have offered only approximate estimates for specific diseases. Since the onset of substantial funding cuts, few studies have evaluated the consequences of reductions in USAID-funded interventions. In a few cases, websites have provided rough estimates on the cumulative effect of these cuts on tuberculosis, child mortality, and other health diseases or conditions.

Added value of this study

To our knowledge, this study is the first comprehensive analysis to assess the impact of total USAID funding—including support for health care, nutrition, humanitarian aid, development, education, and related sectors—on mortality in LMICs over the past two decades. Our study also disaggregates effects by age group and cause of death. Importantly, it is the only study to integrate retrospective evaluations with forecasting models that project the effects of current and proposed funding cuts on child and all-age mortality to 2030. We estimate that over the past two decades, USAID-funded programmes have helped prevent more than 91 million deaths globally, including 30 million deaths among children. Projections suggest that ongoing deep funding cuts—combined with the potential dismantling of the agency—could result in more than 14 million additional deaths by 2030, including 4·5 million deaths among children younger than 5 years. These results provide essential evidence for policy makers, planners, and advocates navigating the future of US global health engagement.

Implications of all the available evidence

USAID funding has had a crucial role in improving global health, particularly by reducing mortality from poverty-related diseases and saving the lives of millions of adults and children. Current and proposed US aid cuts—along with the probable ripple effects on other international donors—threaten to abruptly halt and reverse one of the most important periods of progress in human development. For many LMICs, the resulting shock would be similar in scale to a global pandemic or a major armed conflict. Unlike those events, however, this crisis would stem from a conscious and avoidable policy choice—one whose burden would fall disproportionately on children and younger populations, and whose consequences could reverberate for decades.

In 2023, the USA accounted for 43% of all government funding donated by countries to the humanitarian system, up from about 39% a decade earlier. USAID has been estimated in 2024 to have managed more than 35 billion. [1](https://www.thelancet.com/journals/lancet/article/PIIS0140-6736\(25\)01186-9/fulltext#) Although the USA engages in a wide range of sectors, the biggest sectors being funded were humanitarian assistance (9·9 billion) and health (9·5 billion), and the largest benefiting region was sub-Saharan Africa (12·3 billion). However, throughout the years, and according to USAID reports, the agency also funded education projects by investing in teacher training, curriculum development, and school infrastructure, supporting, only between 2011 and 2015, more than 52 million children in 45 countries. 1

Between 2017 and 2020, the agency responded to more than 240 natural disasters and crises worldwide; in 2016 alone, the organisation provided food assistance to more than 53 million people across 47 countries. 3 Moreover, USAID has been a supporter of the Global Alliance for Vaccines and Immunization (GAVI) and pledged 1·16 billion over 2020–23 to support the organisation. USAID has also been involved in combating malaria through the President's Malaria Initiative (PMI). USAID is also one of the seven agencies involved in the direct implementation of the President's Emergency Plan for AIDS Relief (PEPFAR), launched in 2003, and investing an accumulated amount of over 100 billion in the global HIV/AIDS response. 3 In 2023, 60% of PEPFAR’s bilateral HIV assistance was obligated and implemented by USAID. It is important to recognise that the USAID funding model— like that of other US federal agencies—is largely driven by political negotiations within Congress and the Senate. Consequently, ensuring the long-term sustainability of funding is inherently challenging, particularly in the absence of strong support from the executive branch. 1,2,4 For these reasons, a few USAID sectors have been elaborating plans to promote the scalability and durability of their interventions.

On Jan 20, 2025, the Trump administration released the Executive Order 14169, Reevaluating and Realigning United States Foreign Aid, which suspended existing foreign aid programmes, except for emergency food assistance and military aid. 5 On March 10, 2025, it was announced that 83% of the programmes run by USAID would be cancelled. 6 These cuts are already being challenged in court, and the outcome of the process is uncertain, at least for the current fiscal year. Assuming the cancellations stand, this could include a potential 88% cut in support to maternal and child health aid, 87% to epidemics and emerging diseases surveillance, and 94% cuts to programming for family planning and reproductive health. 7

Suspended US contracts for the PMI have halted hundreds of millions of dollars annually to countries such as Nigeria and Uganda, threatening an increase of nearly 15 million additional cases and 107 000 additional deaths globally in just 1 year of a disrupted malaria-control supply chain .8,9 The UN World Food Programme has closed its southern Africa office, placing 27 million people at risk of hunger amidst the country’s worst drought in decades. 10 Moreover, a recent survey has estimated that 79 million people previously targeted for assistance are no longer being reached because of USAID programme cuts, and that the local capacity of national non-governmental organisations has been profoundly affected. 11 If these substantial cuts continue, the vast majority of USAID-funded activities will be affected, with most likely to be terminated.

To date, only a few assessments, most of them not peer-reviewed and focused on health-related USAID programmes, have tried to estimate the impact of USAID and the effects of the current funding cuts on specific diseases. 12–14 This study aims to provide a comprehensive evaluation of the impact of total USAID funding—including support for health care, nutrition, development, humanitarian relief, education, and other sectors—on mortality in low-income and middle-income countries (LMICs) over the past two decades. The analysis includes disaggregated estimates by age group and cause of death and employs forecasting models to project the potential effects of current funding cuts on all-age and child mortality to 2030.

Methods

Study design

Our study combined a retrospective impact evaluation (ex-post, covering the period 2001–21) with forecasting analyses (ex-ante, projecting from 2022 to 2030). Both components relied on a shared underlying data structure, study design, and analytical approach.

The impact evaluation had a longitudinal ecological design, whereby countries (unit of analysis) were observed and projected over time. This longitudinal dataset combined aggregated demographic, socioeconomic, health, and USAID data from several sources (all data used in this study are publicly available from the sources listed in the appendix p 10.) during 2001–23. From all countries and territories worldwide, we selected 133 low-income, lower-middle-income, and upper-middle-income countries with consolidated data available up until 2023. Models including all countries, and weighting by population size were also estimated (appendix p 21).

Study variables

The main dependent variable in the analysis was age-standardised all-cause mortality (ASMR) per 1000 inhabitants. In addition, mortality was examined across different age groups, including child mortality (children younger than 5 years per 1000 live births). Given that USAID interventions are expected to have the greatest effect on children, we further disaggregated under-five mortality into specific age categories: infancy (0–1 year), preschool (2–4 years), child (younger than 5 years), and school age (5–9 years). To assess cause-specific associations, we also selected a set of mortality categories linked to USAID priorities and poverty-related conditions on the basis of existing literature. 12–19 These associations were defined using the International Classification of Diseases, 10th revision (ICD-10), and included tuberculosis (A15–A19 and B90), HIV/AIDS (B20–B24), maternal causes (O00–O99), lower respiratory infections (J09–J22, P23, and U04), malnutrition (E00–E02, E40–E46, E50, D50–D53, D64.9, and E51–E64), diarrhoeal diseases (A00, A01, A03, A04, and A06–A09), malaria (B50–B54, P37.3, and P37.4), and neglected tropical diseases (A66, A67, A69.1, A71, A77, A78, A79, B55–B56, B57, B65, B66, B73–B74, B76–B77, B79, B83, B88.0, B88.1, and B88). Lastly, specific injury-related mortality (V01–Y89, excluding X41–X42, X44–X45, and U12.9) was included as a negative control, as in previous studies. 20

Our exposure variable was the financial assistance provided by USAID, in all sectors and areas of interventions, specifically measured as USAID funding per capita. The USAID funding per capita was calculated, similarly to previous studies, 12,15 by dividing the total disbursed amount (numerator, in monetary values) by the total population (denominator) for each of the 133 countries across each fiscal year from 2001 to 2023. As in previous studies, 20,21 we categorised the USAID funding per capita to estimate the non-linear dose–response relationship associated with increasing levels of intervention implementation. In the absence of established reference values in the literature—and in line with previous evaluations that classify interventions intensity into three levels (low, intermediate, and high), 20,21 and considering the countries where USAID mostly operates—USAID funding per capita was categorised using quartiles of its distribution in low-income countries: baseline, considered as not exposed (0–1·96 per capita per year); low (25th percentile, 1·97–3·96); intermediate (50th percentile or median, 3·97–7·09); and high (75th percentile, 7·10 and higher). This categorical approach was preferred over a continuous specification because the functional form of the dose–response relationship was unknown, and the exposure variable included potential outliers that could disproportionately influence effect estimates in continuous models. Nevertheless, we also tested models with continuous variables and several alternative categorisations (and the results remained robust across all different specifications; more details on these sensitivity analyses and alternative categorisations are available in the appendix pp 18–20).

All relevant time-variant demographic, socioeconomic, and health-care adjusting variables, according to the literature, 12,15,22,23 were included in the models: gross domestic product (GDP) per capita at purchasing power parity (GDP pc PPP); public expenditures on education, health, and the military (each on as a percentage of GDP); literacy rate; Gini index; the percentage of households with inadequate sanitation and with access to piped water; the number of doctors per 1000 population; and the number of hospital beds per 1000 population. Additional covariates and specifications where also tested in sensitivity analyses (appendix pp 16–33). As in previous studies, 20,21 to harmonise the model with the categorised exposure variables and to mitigate the influence of potential outliers, we dichotomised these covariates according to their median value over the period. Moreover, we included time dummy variables (for 2008–09, 2013–14, 2015–16, and 2020–21) to adjust for major economic and health shocks that occurred globally in the past two decades. 24

Data sources

Data on age-standardised mortality, per cause and age groups, were collected from the Global Burden of Disease Collaborative Network. 24 The data on USAID funding was collected from US Government agencies reporting foreign assistance. 2 Demographic, socioeconomic, and healthcare-related variables were obtained from the World Bank data and the WHO Global Health Observatory data. 25 The complete list of data sources and related detailed methods is presented in the appendix (p 10).

Statistical analyses

In the retrospective impact evaluation, we estimated the effect of USAID per capita funding during the years 2001–21 on age-standardised mortality, overall and for age-specific and cause-specific groups, using Poisson multivariable regression models with robust SEs and fixed-effects specifications. Additional estimates included sex–age-specific mortality, further stratified by country groupings based on income level, GDP, Gini index, and Human Development Index (HDI). This is a consolidated methodological approach that, accompanied by extensive sensitivity and triangulation analyses, enables the estimation of the effects of interventions on mortality using panel data at the aggregate level. 20–22,26 The fixed-effects models include a term to control for unobserved characteristics of the unit of analysis that are approximately constant during the study period, such as some geographical, historical, infrastructural, or sociocultural aspects of each country, and could be associated both with the outcome and with the intervention implementation. 20–22 To evaluate the robustness of the estimates, we did several sensitivity analyses (appendix pp 13–33). First, to evaluate the influence of the exposure categorisation, we fitted the models by using continuous variables and by changing the number of categories and variable thresholds. Second, to assess the external validity of our estimates—that is, to verify that the findings hold true in a broader context—we fitted the models using all 204 countries and territories in the world. We also tested the effect of population weighting in the regression and assess the effect of excluding highly populated countries, such as China and India. Third, to investigate the influence and relevance of the time trends, we tested different sets of time variables. Fourth, to evaluate the stability of the results with alternative models, we fitted negative binomial regression models and compared their estimates with Poisson models. Fifth, to verify the specificity of the USAID funding-per-capita effects, we fitted the same models with injury-related mortality, used as a negative control. 20,27 Finally, to reach a higher degree of confidence in the causal inference and the overall impact evaluation, we did triangulation analyses 28 using difference-in-difference with propensity score matching, 21,26,29 evaluating the countries with low USAID coverage versus medium and high coverage in the years 2001 and 2021. We used Stata (version 17·0) for database processing and analysis.

For the forecasting analysis, we employed validated country-level microsimulation models to project the health effects of current USAID defunding and its progressive phase-out until 2030. Microsimulation is widely regarded as one of the most accurate forecasting methods, because it enables the incorporation of country-specific characteristics and their associated outcome probabilities into the modelling process. This characteristic is especially true when models are developed using projections on the basis of retrospective real-world cohorts, preserving the original distribution of variables, their intercorrelations, and country-specific trends. 30 Our modelling approach, based on previous studies, 21,26,31 was done in two stages: first, we created a synthetic cohort of all countries for the years 2024–30, extrapolating and modelling each country-level independent variable from the retrospective dataset; second, we predicted mortality using these independent variables as inputs in the same multivariate regression models employed in the retrospective analysis, incorporating the effect estimates derived from the ex-post evaluation.

In the first stage, we simulated two USAID scenarios: first, a business-as-usual scenario, keeping USAID funding at the levels of 2023; and second, the currently prospected 83% funding cuts of 2025, and the potential termination of USAID funding from 2026 to 2030 (appendix pp 39–40). In the second stage, for each outcome and each scenario, we did 1000 Monte Carlo simulations, allowing parameter values to vary in each simulation cycle according to their underlying distribution. All-age mortality and under-five mortality across scenarios were compared using mortality rate ratios and absolute differences in the number of deaths, including estimates of the cumulative number of excess deaths over the entire 2025–30 period.

Further details of the modelling process—done in accordance with international model reporting guidelines (ISPOR-SMDM) 32—are provided in the appendix (pp 39–44). These details also include model calibration, validation, parameter distributions for Monte Carlo simulations, and the model equations. All forecasting analyses were done using R (version 4.1.2). 33

Role of the funding source

The funder had no role in the study design, data collection, data analysis, data interpretation, or writing of the manuscript and the decision to submit for publication.

Results

We calculated the mean values and trends of all the variables in the selected countries over the study period (2001–21; table 1). Overall ASMR started in 2001 at 11·65 (SD 3·46) per 1000 population and decreased by 13% until 2021, whereas the under-five mortality started at 73·71 (46·31) and reduced by 49%. On average, USAID funding increased by 97% (from 1·38 to 2·71 capita), whereas the average funding per country increased by 68% (from 151 million to 253 million). Overall, socioeconomic conditions improved, both in terms of GDP per capita, adequate sanitation, primary education, and other indicators. Health expenditure as a percentage of GDP and the number of physicians also increased. The Gini Index, hospital beds per 1000 population, and a few other indicators showed mild deterioration.

Events Citing This Source

Event	Date	Category
USAID Dismantled	Jan 24, 2025	Abuse of Power

People Mentioned

Person	Role

Institutions Mentioned

Institution	Description

Related Sources

Source	Type	Publisher
Study - 14 million lives could be lost due to Trump aid cuts	news-reporting	NPR