The first hard numbers on what the PEPFAR stop-work orders did to HIV services

Most of what has been written about the dismantling of US global health assistance this year has been projection: models of what will happen if funding does not return. Three papers published in the past two weeks change the register. One of them reports what actually happened, week by week, in 165 clinics [s1].

Beginning in late January 2025, stop-work orders and contract cancellations disrupted HIV programmes supported by the President's Emergency Plan for AIDS Relief [s1]. Researchers analysed weekly aggregate service-delivery data from a convenience sample of 165 CDC-funded, ICAP-supported facilities — 22 in Angola, 75 in the Democratic Republic of the Congo, 20 in South Sudan and 48 in Zambia [s1]. They compared three phases: pre-stop-work (7 October 2024 to 23 January 2025), stop-work (24 January to 11 February 2025), and post-resumption (12 February to 31 March 2025) [s1].

What the facility data shows

The stop-work window itself lasted under three weeks. The declines it produced did not.

In Angola, the DRC and Zambia, the number of HIV-positive test results fell significantly across the three phases — by 58%, 34% and 17% respectively — and antiretroviral therapy initiations fell by 16%, 32% and 17% [s1]. Recovery was limited: positive tests in Zambia and ART initiations in Angola did not return to their pre-order levels by the end of March [s1].

HIV testing overall declined significantly in the DRC and Zambia, by 33% and 35% [s1]. Index testing — the tracing of partners and contacts of people newly diagnosed, which is how programmes find undiagnosed infections efficiently — fell by 37% in the DRC and 72% in Zambia [s1]. Testing of pregnant women fell 28% in the DRC [s1]. Early infant diagnosis testing declined 12% in Angola and 18% in Zambia, with limited recovery [s1].

Some measures did rebound. In Angola, HIV testing moved 2,476 → 2,205 → 2,519 across the three phases, and testing of pregnant women 280 → 233 → 287 [s1]. In the DRC, early infant diagnosis recovered, 6.5 → 6.3 → 7.9 [s1]. No reductions were observed at all in South Sudan [s1].

One finding deserves attention because it is easy to misread. Testing yield — the proportion of tests returning positive — rose in Zambia, from 2.8% to 3.1% to 4.0% [s1]. A rising yield alongside a falling test count is not an efficiency gain. It is what happens when a programme stops testing broadly and the remaining tests are concentrated among people already more likely to be infected.

The authors describe the disruptions as producing "substantial short-term reductions" in testing and treatment delivery, and conclude that long-term funding disruptions require careful planning, realistic timelines and investment in cost-effective service models [s1].

What the models add — and where they should be discounted

Two modelling papers published within days of the facility data attempt to scale the consequences forward.

A Zambia-specific agent-based transmission model, calibrated to provincial HIV data and using disruption assumptions supplied by the Zambian HIV programme's own leadership, simulated disruptions lasting 3 months, 1 year, 4 years or unabated, against a counterfactual of no disruption, over 2025–2060 [s2]. Zambia is a useful test case because 84% of its HIV programme funding came from PEPFAR at the start of 2025 [s2].

Unabated disruption, in that model, adds 3.3 million HIV acquisitions and 1.6 million HIV deaths over the period — 8.8 times and 5.3 times the no-disruption counterfactual — with the largest absolute burden among women (1.5 million acquisitions, 790,933 deaths) and the largest proportional increase among children (21.6 times acquisitions, 20.8 times deaths) [s2]. Restoration of services within three months limits the additional acquisitions to 54,863 (+13.1%) and additional deaths to 32,550 (+8.7%) [s2].

A separate analysis modelled the tuberculosis consequences across 26 high-burden countries under three recovery scenarios [s3]. Additional TB cases between 2025 and 2030 come to 0.63 million (CI 0.45–0.81) if services recover within three months, 1.66 million (CI 1.2–2.1) if within a year, and 10.67 million (CI 7.85–13.19) in the worst case of long-term service reduction [s3]. Projected additional TB deaths are 99,900, 268,600 and 2,243,700 respectively [s3].

These figures should be handled as what they are. A 35-year projection compounds every assumption in the model; the difference between the three-month and unabated scenarios in the Zambia model spans two orders of magnitude, which is a fair statement of how little the modelling constrains the answer once you leave the near term. The TB estimates extrapolate from representative countries to all 26 high-burden countries by dependency category [s3], a step that assumes countries within a category behave alike.

What is actually established

The facility data is the firmer ground, and it establishes three things. A stop-work order of under three weeks produced service declines measured in tens of percent. Those declines did not fully reverse within six weeks of resumption. And the effect was highly uneven — severe in Angola, the DRC and Zambia, absent in South Sudan [s1] — which suggests that programme structure and local funding mix determine how hard a given system is hit, rather than any uniform dose-response to the funding interruption.

The convenience sample is a real limit: 165 facilities supported by one implementing partner are not a random draw from PEPFAR-supported sites, and the direction of that bias is not obvious.

What to watch: whether service volumes in the affected countries have returned to baseline by the end of 2025, and whether the near-term restoration scenarios in the models remain the relevant ones.

This article is informational and is not medical advice.

Sources

• [s1] "Effects of Stop-Work orders on HIV testing, treatment and programmes for prevention of vertical transmission in four sub-Saharan African countries," Journal of the International AIDS Society, September 2025. https://doi.org/10.1002/jia2.70034
• [s2] "Impacts of US Bilateral Aid Disruptions on HIV Resurgence in Zambia: A Mathematical Modeling Study," Open Forum Infectious Diseases, 11 September 2025. https://doi.org/10.1093/ofid/ofaf511
• [s3] "A deadly equation: The global toll of US TB funding cuts," PLOS Global Public Health, 10 September 2025. https://doi.org/10.1371/journal.pgph.0004899