ANALYSIS

A modelling study puts 9.4 million deaths on the mild aid-cut scenario, not the severe one

The Lancet Global Health paper that produced this month's headline number also produced a larger one: 22.6 million additional deaths by 2030 if defunding deepens. The distinction matters more than the figure.

A modelling study published online on 2 February in The Lancet Global Health projects that continuing the current downward trend in official development assistance would produce 9.4 million additional deaths across low-income and middle-income countries by 2030, with a 95% uncertainty interval of 6.2 million to 12.6 million [s1].

That figure has been widely repeated since the paper appeared. What has been repeated less often is which scenario it belongs to. The authors modelled three futures, and 9.4 million is the middle one — the scenario they define as a simple continuation of current trends, which they label mild [s1]. Their severe defunding scenario projects 22.6 million additional deaths (95% UI 16.3–29.3), including 5.4 million (4.1–6.8) among children younger than five [s1]. Under the mild scenario, the under-five figure is 2.5 million (1.8–3.2) [s1].

The gap between the two is roughly 13 million lives, and it is a gap the paper explicitly leaves open. The number that ends up being correct is a function of decisions not yet made.

How the estimate was built

The analysis has two halves, and the second depends entirely on the first.

The retrospective half used longitudinal panel data from 93 low-income and middle-income countries to estimate the association between per-capita ODA funding and mortality between 2002 and 2021 [s1]. The model was a two-way fixed-effects multivariable Poisson regression with robust standard errors, adjusted for demographic, socioeconomic and health-system covariates [s1]. The authors report age-specific and cause-specific effects and describe sensitivity and triangulation analyses intended to test whether the associations can bear a causal reading [s1].

Higher ODA funding levels were associated with a 23% reduction in age-standardised all-cause mortality (rate ratio 0.77; 95% CI 0.70–0.85) and a 39% reduction in under-five mortality (0.61; 0.49–0.75) [s1].

The disease-specific estimates are where the mechanism becomes legible. ODA funding was associated with a 70% reduction in HIV/AIDS mortality (RR 0.30; 95% CI 0.24–0.39), 56% for malaria (0.44; 0.35–0.56), 56% for nutritional deficiencies (0.44; 0.30–0.65) and 54% for neglected tropical diseases (0.46; 0.36–0.59) [s1]. Significant reductions were also reported for tuberculosis, diarrhoeal disease, lower respiratory infections, and maternal and perinatal causes [s1].

Those are the categories where donor money buys commodities and delivery — antiretrovirals, bednets, therapeutic feeding, mass drug administration. It is not surprising that the estimated effect concentrates there. It is worth noting that this pattern is also what a plausible confounding story would predict, since donor funding flows toward exactly the diseases where measurable gains are achievable.

The forecasting half fed those retrospective estimates into country-level microsimulation models to project mortality to 2030 under the three scenarios [s1].

What the numbers can and cannot carry

This is a projection, not a count. Its central estimates inherit every assumption in the retrospective model, and the uncertainty intervals reported are statistical — they describe sampling variation in the estimates, not uncertainty about whether the underlying relationship will hold as funding contracts.

That last point deserves emphasis. The model assumes that removing a dollar of assistance reverses the mortality effect of adding one. Health systems do not necessarily behave symmetrically. Infrastructure built with donor money may persist after the money stops; conversely, an abrupt withdrawal may do more damage than a gradual one of the same total size, because programmes that lose continuity — treatment cohorts, cold chains, surveillance networks — can fail faster than they were built.

The study was funded by RF Catalytic Capital and the Spanish Ministry of Science and Innovation [s1].

The financing picture the study is modelling

WHO's own numbers, published a day after the paper, describe the same contraction from the operational side.

The agency launched its 2026 health emergency appeal on 3 February, seeking nearly US$1 billion to respond to 36 emergencies worldwide, 14 of them Grade 3 — the highest level of organisational response [s2]. In 2025, WHO reported, humanitarian funding fell below 2016 levels, and WHO and partners were able to reach only one-third of the 81 million people originally targeted for humanitarian health assistance [s2].

What that third bought is itemised: 30 million people supported, 5.3 million children vaccinated, 53 million health consultations, more than 8,000 health facilities supported and 1,370 mobile clinics deployed [s2]. WHO says it coordinates more than 1,500 partners across 24 crisis settings [s2].

WHO Director-General Tedros Adhanom Ghebreyesus framed the appeal as "not charity" but "a strategic investment in health and security" [s2].

What to watch

The paper's own logic points at the variable to track: not the headline projection but which trajectory actual disbursements follow over the next two to three years. The mild and severe scenarios diverge because donor budgets diverge, and budget decisions in 2026 and 2027 will determine which projection the decade is measured against.

Two other things are worth watching. Whether the retrospective effect estimates are replicated by independent groups using different model specifications — a single modelling paper, however careful, is one specification among many. And whether cause-specific mortality data from the countries most exposed to cuts begin to move in the direction the model predicts, which is the only test that will eventually settle the question.

Sources

Sources

  1. Impact of two decades of humanitarian and development assistance and the projected mortality consequences of current defunding to 2030: retrospective evaluation and forecasting analysisThe Lancet Global Health , February 2, 2026
  2. WHO launches 2026 appeal to help millions of people in health emergencies and crisis settingsWorld Health Organization , February 3, 2026

More on

Related coverage