Aid withdrawal: an event study of mortality, vaccine coverage and DALY following transitioning from Gavi support Authors: Kit Ming Isabel Yan, Jinfeng Liu, Ishwara Hegde, Eran Bendavid Published: 2026 ABSTRACT AbstractThis study investigates the health impacts of transitioning from Gavi, the Vaccine Alliance, on vaccination coverage, infant and under-5 mortality, and disability-adjusted life years (DALYs) in low and middle-income countries between 2000 and 2021. As Gavi plays a critical role in supporting immunisation programmes for some of the world’s most vulnerable populations, understanding the consequences of funding withdrawal is vital for ensuring the sustainability of health gains. Using a quasi-experimental event study design, we leverage Gavi’s eligibility criteria and transition policy to compare countries phasing out of Gavi support with those continuing to receive assistance. Our analysis reveals significant short-term and medium-term negative effects of reduced Gavi support: the prevalence of children receiving no basic vaccines increased by 1.94 percentage points, infant mortality rose by 7.59 per 1000 live births, and under-5 mortality increased by 17.31 per 1000 live births. DALYs also rose by 1264 units per 100 000 population, with the most pronounced effects occurring during the accelerated transition phase. Although these adverse effects diminish several years after the transition, our findings provide robust evidence of the heightened vulnerability of health systems during the withdrawal of external aid. This study contributes to the literature by quantifying the health impacts of Gavi’s transition process and underscores the need for strengthening transition policies. Policymakers should consider a more gradual and tailored approach to aid withdrawal to ensure the continuity of critical health services and protect long-term health outcomes in transitioning countries.Keywords: Global Health, Health economics, Health policy WHAT IS ALREADY KNOWN ON THIS TOPIC WHAT THIS STUDY ADDS This study provides robust evidence that transitioning out of Gavi support leads to significant short-term and medium-term negative effects, including a 1.94 percentage point increase in children receiving no basic vaccines, rises in infant and under-5 mortality by 7.59 and 17.31 per 1000 live births, respectively, and an increase in disability-adjusted life years by 1264 units per 100 000 population, with the most pronounced impacts during the accelerated transition phase. This study provides robust evidence that transitioning out of Gavi support leads to significant short-term and medium-term negative effects, including a 1.94 percentage point increase in children receiving no basic vaccines, rises in infant and under-5 mortality by 7.59 and 17.31 per 1000 live births, respectively, and an increase in disability-adjusted life years by 1264 units per 100 000 population, with the most pronounced impacts during the accelerated transition phase. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY These findings highlight the vulnerability of health systems during aid withdrawal, urging policymakers to adopt more gradual and tailored transition strategies to sustain health gains, while encouraging further research into long-term impacts and optimal transition mechanisms to protect population health in transitioning countries. These findings highlight the vulnerability of health systems during aid withdrawal, urging policymakers to adopt more gradual and tailored transition strategies to sustain health gains, while encouraging further research into long-term impacts and optimal transition mechanisms to protect population health in transitioning countries. Introduction Development Assistance for Health (DAH) includes the financial and in-kind resources transferred by development agencies to low and middle-income countries for the primary purpose of maintaining or improving health. In 2023, total DAH was US$64.6 billion. Gavi, the vaccine alliance (Gavi) accounted for US$2.5 billion.1 Gavi was established in 2000 as a public–private partnership to pick up where the Expanded Program on Immunization left off—delivering basic vaccines to developing countries. By 2023, Gavi has reached over 1 billion children and enabled roughly 6 billion vaccinations around the world.2 The current public health literature provides an extensive account on the sources and disbursements of DAH over time, past trends and the future of DAH spending.3,6 However, the introduction of DAH to a country often coincides with economic growth, creating a confounding factor that complicates the evaluation of DAH’s benefits. This interplay can potentially lead to an inflation of the perceived positive impacts associated with DAH. Our study is motivated by a unique approach, focusing on the DAH funding mechanism of Gavi to investigate the impact of withdrawal of DAH on health outcomes. In contrast to many health studies within the DAH literature that typically assess the effects of providing DAH, which often coincides with the introduction of DAH funding, we concentrate on the repercussions of DAH withdrawal. This perspective offers a less convoluted examination, as it steers clear of the frequently observed positive correlation between DAH introduction and a country’s economic growth. This analysis is uniquely timely in the context of recent efforts by the US administration to freeze many development assistance programmes, including DAH. Like Gavi withdrawal, major aid cessations could significantly disrupt health services in low and middle-income countries. This paper analyses the impact of transition out of Gavi support and sheds light on the consequences of aid cessation. More specifically, Gavi provides time-limited support to low-income countries, with eligibility determined by a country’s Gross National Income (GNI) per capita.7 As countries’ economies grow, they transit through a series of phases, gradually increasing their self-financing for vaccines until they fully graduate from Gavi support. Details on Gavi’s eligibility criteria and the transition process are provided in the Methods section. A few recent papers have examined the effect of Gavi by measuring the changes in key health indicators using difference-in-difference (DiD) frameworks. Jaupart et al and Shastry and Tortorice (forthcoming) both provide evidence demonstrating that countries’ eligibility to receive Gavi support significantly increased the immunisation coverage and reduced the child mortality rate in several lower income countries.8 9 Another study by Debellut et al found that middle-income countries (MICs) that are not eligible for funding from Gavi had been slow to adopt rotavirus vaccines.10 Given their findings of positive effects from Gavi assistance, it has been a concern that the transition out of Gavi funding when countries graduate out of eligibility could lead to a significant reversal of the positive impacts. The actual effect resulting from the graduation out of Gavi is an important question that has not yet been thoroughly studied. While the prevailing literature on DAH leans towards a positive framework that examines how increased aid implementations enhance health outcomes,11,13 our study introduces a fresh perspective by investigating the effects of aid discontinuation. We posit that this approach mitigates potential endogeneity concerns caused by the coincidence between the introduction of DAH and other economic development policies that can lead to an overestimation of the impact of DAH. This study develops an event study to compare the dynamics of health outcome trends in countries transitioning out of Gavi support with those continuing to receive assistance. Specifically, our paper employs the eligibility criteria as an instrument to establish a quasi-experiment for investigating the impact of Gavi assistance. We undertake an econometric analysis to explore the impacts of discontinuing assistance for vaccines on a country’s health outcomes. The discontinuation process is initiated when a country no longer meets the aid criteria, determined by an income threshold specified by the World Bank’s International Development Association. This discontinuation serves as the foundation for a quasi-experimental design, allowing us to investigate the effects of health aid on a country’s health outcomes. Our findings revealed that reduced Gavi support has significantly negative impacts on key health indicators, such as vaccination rates, mortality rates and Disability-Adjusted Life Years (DALY). By quantifying these impacts and highlighting the temporal patterns of deterioration and recovery, this study offers valuable insights into the consequences of reducing Gavi support on health outcomes. The findings of this study, combined with existing evidence, underscore the critical importance of sustained support for vaccination programmes in countries transitioning out of Gavi assistance. The significant short-term and medium-term negative effects observed following reductions in Gavi support emphasise the necessity of crucial importance of managing the transition process to mitigate adverse health outcomes. These results advocate for enhanced strategies and interventions to ensure the continuity of essential health services and vaccine coverage post-Gavi support, guiding policymakers in re-evaluating and strengthening the transition mechanisms for countries transiting out from Gavi assistance. Our research distinguishes itself from past studies that have laid some foundational insights on aid discontinuation.14,20 Unlike our analysis, Galiani focuses on the impacts of economic aid discontinuation on economic growth rather than on health outcomes,21 and Ojal et al focus on the effect of transitioning out of Gavi on one specific country—Kenya.22 Another study by Kolesar analyses the effect on the vaccination coverage and postneonatal mortality of eight countries but concludes that there is no uniform result across countries and health metrics, and there is only limited impact on the postneonatal mortality.23 Mao et al further hypothesise that bilateral donors might follow Gavi’s lead in health aid transitions, except in countries where they have strategic interests.24 Our paper complements these works by conducting a quasi-experiment focused on Gavi, specifically studying the impact of transitioning out of health aid on health outcomes such as infant mortality, under-5 mortality, vaccine coverage and DALY at different time horizons. It provides substantiated evidence via a broader sample encompassing all countries, which have transited out of Gavi. This helps us to have a more comprehensive understanding of the impact of DAH on a wider set of countries. Methods Gavi eligibility and transition policy Gavi’s primary focus centres on providing support to low-income countries, with time-limited assistance contingent on the country’s GNI per capita, determined by the GNI per capita Atlas method.25 To be eligible for Gavi support, a country must maintain a 3-year average GNI per capita below or equal to the Gavi’s eligibility threshold (US$1580 in 2018 and US$1730 in 2023) when submitting their applications. Overall, Gavi defines four phases of its transition policy as follows26: Phase 1: initial self-financing phase—a Gavi-eligible country is considered to be in the initial self-financing phase and enjoy standard Gavi support if its GNI p.c. is equal to or below the World Bank’s low-income threshold ($1085 or less in 2023). Eligible countries cofinance a flat rate of $0.20 per dose of any Gavi-supported vaccines. A country remains in this phase until its GNI per capita exceeds the World Bank’s low-income threshold. Phase 2: preparatory transition phase—a Gavi-eligible country is considered to be in the preparatory transition phase if its GNI p.c. is above the World Bank’s low-income threshold. This phase typically spans 4 years, during which cofinancing commitments increase by 15% annually. Phase 3: accelerated transition phase—triggered when a country’s 3-year average GNI per capita, as well as its most recent GNI per capita, exceeds Gavi’s eligibility threshold. During this phase, the country must cofinance a minimum of 35% of its vaccine costs. The accelerated transition phase originally lasted for 5 years but was extended to 8 years starting from January 2023. Phase 4: fully self-financing (graduated) phase—after completing phase 3, the country fully self-finances the cost of vaccines. Gavi ensures continued access to UNICEF tenders for Gavi-approved vaccines for an additional 5 years’ post-graduation.27 To facilitate a country’s transition out of Gavi support, a designated transition plan is established, starting with the mostly Gavi-financed initial self-financing phase (phase 1) or the preparatory transition phase (phase 2) and progressing towards a phase with minimal Gavi financing (phase 3) before reaching the fully self-financing, post-Gavi graduated phase (phase 4). Under this transition plan, Gavi gradually reduces its funding support for vaccination as the recipient country moves through the phases, ultimately achieving full self-financing for vaccination. The Gavi Transition Policy, outlined in figure 1, illustrates the transition process and self-financing proportion during those four phases. Figure 2 presents the list of Gavi-eligible countries during the sample period (2000–2021) and their transition status. Gavi transition policy. GNI, Gross National Income. Gavi-eligible countries and their transitions. Congo was once graduated in 2018, but regained its eligibility in January 2019. The entry into the preparatory transition phase marks the initiation of the reduction of Gavi’s financial support for vaccination. In this paper, we employ a panel DiD analysis, designating the beginning of the preparatory transition phase as the treatment event. Approach Our study exploits Gavi’s transition process to study the relationship between phasing out from Gavi support and subsequent changes in a country’s vaccine coverage, mortality rates and DALY during 2000–2021. Employing a staggered DiD methodology, we focus on the pivotal event of entry into Gavi’s preparatory transition phase (phase 2) as the primary event in our quasi-experimental framework. This event serves as the cornerstone for comparing outcomes between a control group comprising countries eligible for Gavi support that have not entered the preparatory transition phase and a treatment group comprising countries eligible for Gavi support that have started this transition in varying years. Countries enter the panel in the year they become Gavi-eligible (2000 for most countries) and are observed until 2021. The DiD analysis enables us to scrutinise changes in health outcomes between the groups that differ in both groups before and after the transition. Our application of the DiDs methodology aligns with established research that has analysed significant interventions in public health and social phenomena. For instance, Venkataramani et al employed this approach to investigate how automotive assembly plant closures affected opioid overdose mortality, demonstrating the flexibility of DiD across diverse policy contexts.28 Similarly, the effectiveness of DiD in assessing nutritional outcomes in sub-Saharan Africa further underscores its value in health policy analysis.29 Their findings validate the use of DiD for analysing the impacts of policy changes on health. Gavi funding assumes a pivotal role in our analysis of international DAH, owing to its transparent and well-defined transition policy. Gavi’s funding phases and transition policy provide clarity to our event timing. This transition not just increased the cost of vaccines to countries but also affected their overall desirability to acquire vaccines.30 Moreover, Gavi’s extensive operational history, in comparison to other forms of DAH, facilitates a longitudinal study over a substantial time period. This extended timeframe proves essential in assessing the enduring efficacy of health aid on health indicators. These attributes collectively position Gavi as a good instrument for conducting our event study. Data Mortality and vaccination coverage data were derived from the DHS.31 DHS relies on nationally representative household surveys with substantial sample sizes for data collection. Household surveys are preferred over nationally reported figures, as the former provides a more reliable measure of coverage over time.32 The DALY data were obtained from the Institute for Health Metrics and Evaluation—The Global Burden of Disease database, which provides measurement of health problems from 204 countries and territories, 369 diseases and injuries and 87 risk factors.33 Covariates such as real gross domestic product per capita, domestic health expenditures per capita (in 2017 USD) and the percentage of population living in the urban area are extracted from the World Bank.34 To ensure data completeness, linear interpolation was applied where necessary, and we also utilise data without interpolation in the Sensitivity analysis section. A statistical summary table and a sources of the variables table are provided in online supplemental appendix SA1 and SA2 respectively. In our investigation, we scrutinise the relationship between Gavi’s disbursements for vaccines and the percentage of children with no coverage of the basic vaccines outlined by the WHO.35 The basic vaccines include BCG, three doses of DPT-containing vaccine, three doses of polio vaccine (excluding polio vaccine given at birth) and one dose of measles-containing vaccines (MCV). As illustrated in figure 3A, we observe an increase in the percentage of children who received no basic vaccines within the treatment group on entering the preparatory transition phase. Raw trends during transition. (A) Percentage with no basic vaccines; (B) infant mortality rate; (C) under-5 mortality rate; (D) disability-adjusted life years (DALYs, related causes). Year 0 marks the beginning of the preparatory transition phase (ie, the treatment). All the series use demeaned value. The graphs are presented with a balanced set of countries, ie, countries that have complete data points for all periods depicted in the graphs, spanning from t = −2 to 9. Our focus centres on two crucial mortality indicators: infant mortality and under-5 mortality. To illustrate the trends in mortality rates over time, we present figures below (figure 3B, C), with time 0 denoting our event of interest—entry into the preparatory transition phase. A noticeable increase in mortality rates for the treatment group compared with the control group at time 3 is evident in both cases. The findings from our event study graphs shed light on the impact of Gavi support on infant and under-5 mortality rates. From the DALY trends over time for both the treatment and control groups in figure 3D, we observe an upward trajectory in DALY for the treatment group in comparison to the control group on entering the preparatory transition phase. This observation amplifies our motivation to delve deeper into the impact of Gavi funding on a country’s DALY. We incorporate covariates to account for significant potential confounding factors in the relationship between our event of interest (entry into the preparatory transition phase) and the dependent variables of interest (mortality rates, vaccination coverage and DALY). These covariates encompass log-transformed real GDP per capita (PPP adjusted), log-transformed health expenditure per capita (in 2010 USD) and the percentage of the population residing in urban areas. Log-transformed real GDP per capita represents the level of economic development, which has potential influence on mortality rates and DALY. This establishes the real GDP per capita as a relevant correlate in our analysis. Similarly, log-transformed health expenditure per capita is included, considering its impact on health indicators. Additionally, the percentage of the population living in urban areas is part of our covariates, given existing evidence suggesting disparities in infant mortality rates between rural and urban areas.36 By incorporating these covariates, we aim to refine our analysis by accounting for these crucial factors and ensuring a more comprehensive understanding of the relationships under investigation. Statistical approach In our primary analysis, the dependent variables comprised the percentage of children with no basic vaccine coverage, the infant mortality, under-5 mortality and DALY. The core regression model can be represented as follows: Here i indexes the country and t the calendar year. The variable yit represents the outcome of interest. The subscript p refers to event years, which are annual intervals relative to the calendar year of a country enters into the (preparatory) transition phase. For example, period 2 refers to 2 years before a country enters into the transition phase. Event periods 4 or more years prior to the transition were combined into a single bin, as were event periods 9 or more years after the entering of the transition phase. Diptransit takes the value 1 for countries that have entered the transition phase, and otherwise. This model analyses countries eligible for Gavi support that have entered the preparatory transition phase (phase 2) as the ‘treatment’ group, while countries eligible for Gavi support that have not entered the preparatory transition phase are the ‘control’ group. Xit is a vector of covariates capturing economic development, healthcare spending and urban-rural disparities, which influence health outcomes. Finally, μi indicates the country fixed effect, whereas θt captures the calendar year fixed effects. This model estimates the difference in outcomes for leads and lags of the entering year relative to a reference year (event time −1) and relative to the countries in control group, which joined the Gavi but have not progressed to the transition stage. Prior to transition, most health indicators show no statistically significant differences between treatment and control groups. These relative differences between the treatment and control groups are captured by the coefficients βp, estimates for which are reported in the regression output (table 1). Our specification takes an ‘event study specification’. By indexing the reference point to time since the event (as opposed to calendar time), this event study specification provides a more in-depth examination of the variation of dynamic treatment effects relative to the initial treatment time. Table 1Exhibition of main regression resultsDependent variablesNo vaccine coverageInfant mortalityUnder 5 mortalityDALY (related causes)Transit, −4 years and prior 1.74 −0.74 −2.92 −779** (−0.56 to 4.05)(−5.54 to 4.07)(−12.49 to 6.64)(−1508 to −50)p=0.15p=0.77p=0.55p=0.040Transit, −3 years 0.61 −0.53 −3.31 −412** (−0.34 to 1.55)(−3.67 to 2.62)(−8.66 to 2.04)(−726 to −98)p=0.21p=0.74p=0.23p=0.012Transit, −2 years 0.23 0.53 0.069 −279*** (−0.30 to 0.76)(−1.25 to 2.32)(−2.99 to 3.12)(−444 to −115)p=0.40p=0.56p=0.97p=0.0014Transit, −1 years(ref)(ref)(ref)(ref)Transit, +0 years 0.94 5.55*** 10.87*** 686*** (−0.29 to 2.18)(1.85 to 9.24)(4.42 to 17.32)(209 to 1163)p=0.14p=0.0048p=0.0017p=0.0063Transit, +1 years 1.17 7.01*** 14.39*** 1024*** (−0.27 to 2.62)(2.27 to 11.75)(6.48 to 22.3)(384 to 1665)p=0.12p=0.0054p<0.001p=0.025Transit, +2 year 1.94** 7.59** 17.31*** 1264*** (0.43 to 3.45)(0.95 to 14.24)(6.68 to 27.94)(490 to 2038)p=0.015p=0.029p=0.0023p=0.0021Transit, +3 years 2.43*** 8.89* 20.42*** 1474*** (0.74 to 4.11)(0.06 to 17.72)(6.86 to 33.98)(610 to 2338)p=0.0066p=0.053p=0.0046p=0.0013Transit, +4 years 2.57*** 9.96* 22.78*** 1555*** (0.78 to 4.36)(−0.43 to 20.35)(6.89 to 38.67)(547 to 2563)p=0.0067p=0.066p=0.0068p=0.0035Transit, +5 years 2.72** 10.85* 23.91*** 1741*** (0.71 to 4.73)(−0.76 to 22.46)(6.49 to 41.33)(535 to 2947)p=0.011p=0.072p=0.0095p=0.0061Transit, +6 years 2.15 13.68*** 28.07*** 1876*** (−0.40 to 4.71)(4.83 to 22.53)(14.28 to 41.86)(516 to 3236)p=0.11p=0.0037p<0.001p=0.0086Transit, +7 years 1.95 13.82*** 27.88*** 2002** (−0.73 to 4.62)(4.70 to 22.94)(13.37 to 42.39)(466 to 3538)p=0.16p=0.0044p<0.001p=0.013Transit, +8 years 1.23 12.74** 26.70*** 2058** (−1.68, 4.14)(3.19, 22.29)(10.80, 42.60)(159, 3957)p=0.41p=0.012p=0.0018p=0.037Transit, +9 years and onward 1.02 12.23** 26.38*** 1728 (−2.31 to 4.36)(2.25 to 22.21)(8.61 to 44.15)(−867 to 4323)p=0.55p=0.020p=0.0052p=0.20Time fixed effectYesYesYesYesCountry fixed effectYesYesYesYesN (county-years)6636666661196N (counties)54565671 Asterisks indicate statistical significance: *p<0.1, **p<0.05, and ***p<0.01 for two-tailed tests. The numbers inside the parentheses report the 95% CI. Exhibition of main regression results Asterisks indicate statistical significance: *p<0.1, **p<0.05, and ***p<0.01 for two-tailed tests. The numbers inside the parentheses report the 95% CI. Patient and public involvement Patients or the public were not involved in the design, or conduct, or reporting or dissemination plans of our research. Results Vaccination coverage For our treated group at time t-1 (1 year before countries enter into the preparatory transition stage), the percentage of children aged 12–23 months who had not received any vaccinations ranged from 0.3% to 19.5%, with an average of 5.19%. Employing a dependent variable of no basic vaccine coverage, our analysis reveals compelling evidence that the percentage of children receiving no basic vaccines significantly increases by 1.94 percentage points (95% CI 0.43 to 3.45; p=0.015) for the treatment group starting 2 years after the country enters the preparatory transition phase, in comparison to the control group. The increase reached its peak at 2.72 percentage point in the fifth year (95% CI 0.71 to 4.73; p=0.011), which corresponds to the second year after the countries entered the accelerated transition phase. This increase in the percentage of children not receiving basic vaccines in the treatment group indicates a concerning trend during the critical transitional phases of Gavi. The delayed emergence of significant effects on vaccine coverage may seem counterintuitive, as one might expect this proximal outcome to respond first given Gavi’s emphasis on immunisation support. This pattern likely stems from the cohort-based definition of the DHS ‘No Vaccine Coverage’ metric, which evaluates the percentage of children aged 12–23 (or 24–35) months who received no basic vaccinations (eg, BCG, DPT, polio and MCV). As such, the measure reflects vaccinations administered in the past and only captures effects once affected birth cohorts age into the survey window, introducing a natural lag of 1–2 years. In contrast, downstream outcomes like infant/under-5 mortality and DALYs can manifest more promptly, as they draw on estimates of recent health events potentially exacerbated by early disruptions in vaccine access or related health systems. Mortality outcomes For our treated group at time t −1, the infant mortality rates per 1000 live births ranged from 22.0 to 69.9, with an average of 41.5; while the under-5 mortality rates per 1000 live births ranged from 27.2 to 131.2, with an average of 57.1. Plotting mortality rates against time in the form of an event-study graph provides motivation to pursue the hypothesis that there is a reversal in the declining mortality trends on entering the preparatory transition phase (figure 3). However, it is to be noted that the change in trend does not seem instantaneous. Lagged transition dummies are used in the regressions, as was foreshadowed by the event-study graphs. In our analysis, we regress the dependent variables on contemporaneous and various lagged transition dummies to capture the impact of the initiation of the preparatory transition phase (time 0) on the dependent variables either contemporaneously or with lags, as foreshadowed by the event study graphs, which show that some effects occur with delays. Specifically, we assess the impact from contemporaneous treatment (effect at time 0) and effects with rolling lags. The resulting DiDs coefficients are detailed in table 1 and visually summarised in figure 4A–D. In the Figures, each point on the Y-axis reflects the difference in the health indicators between the treatment and control groups relative to the year before the transition (denoted as event time −1). Difference-in-difference estimates of vaccination and health outcomes in Gavi-eligible countries by year relative to transition. (A) No basic vaccine coverage by year of transition; (B) infant mortality by year of transition; (C) under-five mortality by year of transition; (D) disability-adjusted life years (DALYs, related causes) by year of transition. Estimates from a staggered difference-in-difference (DiD) event study model, showing regression coefficients for the effect of transitioning out of Gavi support on each outcome variable. Year 0 represents the starting year of the preparatory transition phase. Dashed lines represent 95% CIs. The regression results indicate that the infant mortality rates in the treatment group increase by 5.55 per 1000 live births (95% CI 1.85 to 9.24; p=0.0048) at the starting year of the transition period relative to the control group, with the effect attaining significance immediately on entering the transition phase. Additionally, there is suggestive evidence indicating a notable deterioration in under-5 mortality rates after the beginning of the transition. There is an increase of 10.87 per 1000 live births (95% CI 4.42 to 17.32; p value=0.0017) in treated countries on entering the transition phase relative to untreated countries. Moving forward to the midpoint of the preparatory transition stage, which occurs approximately 2 years after entering this phase, the infant mortality rate of the treated countries exhibited a stronger increase of 7.59 deaths per 1000 compared with the control counties (95% CI 0.95 to 14.24; p=0.0290), while the under-5 mortality rates also showed a pronounced increase of 17.31 per 1000 (95% CI 6.68 to 27.94; p=0.0023). The peak of the effects was observed after countries entered the accelerated transition phase. The infant mortality rate peaked at 13.82 deaths per 1000 (95% CI 4.70 to 22.94; p=0.0044) in treated versus control counties 7 years after entering the preparatory transition phase (approximately 3 years after entering the accelerated transition phase). The under-5 mortality rate peaked at 28.07 per 1000 (95% CI 14.28 to 41.86; p<0.001) in the sixth year following the preparatory transition phase, which correspond to around 2 years into the accelerated transition. Disability-adjusted life year For our treated group at time t−1, the total years of ‘healthy’ life lost per 100 000 population due to related causes varied from 205 to 13 768, averaging at 2427. A positive correlation is identified between entry into Gavi’s preparatory transition phase and DALY. During the midpoint of the preparatory transition stage, the DALY exhibited a significant increase of 1264 units (95% CI 490 to 2038; p=0.0021) in treated countries compared with control counties, with each unit representing the cumulative loss of a ‘healthy’ year of life per 100 000 population. The peak of the effect occurred 4 years after entering the accelerated transition phase, with an increase of 2058 units (95% CI 159 to 3957; p=0.0370). Sensitivity analysis Our results maintain robustness even when utilising DHS data without interpolation. The related table and graphs for the sensitivity analysis are reported in online supplemental appendix SA3. Prior to transition, most health indicators show no statistically significant differences between treatment and control groups. An exception was infant mortality, which was significantly higher in the treatment group by 20.15 per 1000 live births 2 years prior to transition, with p=0.015, which is significant at the 5% significance level, but not at the 1% level. The immediate effects of entering the transition phase are marked by significant deteriorations in health outcomes, both infant and under-5 mortality rates exhibited an even more pronounced increase of 12.97 and 33.81 per 1000 live births (p=0.016 and 0.004, respectively) in the year the treated countries entered into the preparatory transition phase compared with the control group. The direction closely aligns with that estimated using interpolated data, with the only distinction being that the magnitude of effects is larger when using uninterpolated data. Notably, the impacts on infant and under-5 mortality become significant immediately after entering the preparatory transition phase. These findings, based on a robust sample of 54–71 countries and controlling for both time and country fixed effects, strongly suggest that the transition from Gavi support has substantial and enduring negative effects on population health outcomes, particularly in mortality rates and disease burden measures. Nevertheless, for the indicator, no basic vaccine coverage, no significant impact is detected based on the uninterpolated data as the non-interpolated data are too sparse. This leads to a low degree of freedom and diminishes the statistical power of the analysis. Conclusion and discussion In this analysis of countries transitioning out of support from Gavi, we consistently find harmful effects of aid withdrawal on vaccine coverage, mortality among young children and DALYs. Our results show that the cessation of Gavi support leads to significant short-term and medium-term increases in the percentage of children with no vaccine coverage as well as elevated infant and under-5 mortality rates, and DALYs. The most salient features of our findings are the timing and duration of these negative effects. We observed that the adverse impacts were most pronounced during the accelerated transition phase (phase 3), where countries move towards minimal Gavi financing. The effects peaked within this phase, particularly around 2–3 years after entering the preparatory transition phase (phase 2), and began to diminish only several years after the countries completed the transition process. This pattern suggests that the accelerated reduction in Gavi support is a critical period during which health outcomes are most vulnerable. Our study expands the understanding of the relationship between DAH and health outcomes by examining a dimension that has been less explored in the literature. While most studies focus on the positive effects of DAH on health outcomes—demonstrating how increased aid can lead to better health metrics—our analysis uniquely contributes by investigating the consequences of DAH withdrawal. This negative framing offers a different perspective, highlighting the potential risks and adverse effects associated with the cessation of aid. By focusing on the withdrawal phase, our study provides valuable insights into the broader debate on whether DAH improves health outcomes, emphasising that the end of DAH can have significant repercussions. While our findings demonstrate robustness, it is essential to consider certain caveats. First, the utilisation of interpolated data for specific health metrics may lead to data smoothing, potentially reducing the likelihood of identifying significant results associated with the Gavi transition. To address this concern, we have also presented results based on non-interpolated data, although with a reduced sample size. These findings largely support our primary results. Second, the recent transition of some countries out of Gavi support results in a short post-transition period, posing a limitation on our ability to analyse longer term impacts comprehensively. When more data become available in the future, further evaluation can be conducted to deepen our understanding of the sustained effects following the conclusion of Gavi assistance. This study adds to the growing body of evidence that DAH is intricately linked to health outcomes, not only during the active phase of aid but also during the withdrawal phase. Our findings suggest that funders must be cognisant of the potential negative impacts on population health when transitioning countries out of aid programmes. Policymakers should consider the timing and pace of aid withdrawal, as a more gradual reduction in support might mitigate the adverse effects we observed. The recent efforts to dissolve United States Agency for International Development (USAID) could reflect a broader trend of reducing external aid, which could significantly disrupt health services in low and middle-income countries. As our findings indicate, the withdrawal of Gavi support has already led to substantial increases in vaccination deficits and mortality rates, highlighting the fragility of health systems during periods of diminished aid. This underscores the potential value for policymakers to adopt a more strategic and measured approach to aid transitions, ensuring that the lessons learnt from our analysis of Gavi’s support withdrawal are applied to any changes in USAID’s operations, thereby safeguarding the health gains achieved over decades of international assistance. Supplementary material Footnotes Funding: This work was supported by the National Institutes of Health (grant number R01HD104835); Hong Kong University Grant Committee’s General Research Fund (grant number 9043088); and City University of Hong Kong’s Strategic Research Grant (grant number 7006145). Handling editor: John Tayu Lee Patient consent for publication: Not applicable. Ethics approval: This study does not involve direct research on human subjects requiring ethics committee approval. Provenance and peer review: Not commissioned; externally peer-reviewed. Patient and public involvement: Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research. Data availability statement Data are available in a public, open access repository. References Associated Data Supplementary Materials Data Availability Statement Data are available in a public, open access repository.