HIV Funding Impacts Global health funding · LMIC Download the panel

Phase 1.1  ·  Southeast Asia

What happens to HIV services when the funding is withdrawn

The United States has restructured its global health assistance under the America First Global Health Strategy. This project brings together the published record of what follows, including donor funding, epidemic estimates, treatment coverage, and what the services themselves report. Our project aim is to measure the effect on HIV programmes in low- and middle-income countries (LMICs). Southeast Asia is covered first, and we will expand the data coverage to the rest of the developing world.

   

3,937
observations, each with its source and the date it was read
11
sources, from international estimates to national programme registers
31
indicators, on one common vocabulary
2000–2026
years covered
29
of 34 bilateral agreements signed so far cover HIV

Regions

One of eight published

Southeast Asia is the pilot region of our project and, so far, is the only one with a published view. The other seven global regions are grouped and listed here and their regional views will be available in due course. Point at any region to see the countries it covers.

Southeast Asia 9 Published Southeast Asia Nine countries. Four national programmes are read at source; the remainder are covered by the aggregator layer.
  • CambodiaAT SOURCE
  • IndonesiaAT SOURCE
  • Laos
  • Malaysia
  • Myanmar
  • PhilippinesAT SOURCE
  • ThailandAT SOURCE
  • Timor-Leste
  • Vietnam
9 countries 3,937 observations 31 indicators 2000–2026 Open the region page →

What our panel holds

3,937 observations  ·  2000–2026

Every row carries the source it came from, the URL it was read from and the date it was retrieved. Where two sources report the same figure, both are kept. Nothing is averaged away.

Financing

What donors report disbursing, and what countries report spending. The OECD record and the United States self-reported record are kept separately because they are the panel's one independent pair.

768 rows
4 indicators
OECD Creditor Reporting SystemForeignAssistance.govUNAIDS AIDSinfo

Treatment cascade

Diagnosis, treatment and viral suppression. Modelled coverage estimates sit beside programme counts from national registers; the two measure different things and are not merged.

990 rows
11 indicators
UNAIDS AIDSinfoWHO Global Health ObservatoryPEPFAR Panorama Spotlight

Epidemic estimates

People living with HIV, new infections, deaths and prevalence. Almost all of this originates in one modelling exercise, whichever body publishes it.

1,862 rows
5 indicators
UNAIDS AIDSinfoWHO Global Health ObservatoryWorld Bank Open Data

Prevention and outreach

PrEP provision and key-population outreach. Outreach figures are period flows summed to a year, not stocks.

53 rows
2 indicators
UNAIDS AIDSinfoCambodia NCHADSThailand HIV Info Hub

Service access

What service users report about stockouts, interruptions, staffing and cost. A sentinel signal from a small number of facilities, not a representative sample.

30 rows
8 indicators
CLM Asia

Population denominators

Total population, used as the denominator for coverage and rates.

234 rows
1 indicators
World Bank Open Data

Reconciliation

Most agreement between sources is republication, not corroboration.

UNAIDS, the WHO Global Health Observatory and the World Bank largely publish the same modelled estimate. When they agree, that agreement is arithmetic. Treating it as independent confirmation would manufacture confidence the data does not support.

Of 1,233 pairwise comparisons across the panel, 84% fall between sources that share an underlying estimate. Only 143 comparisons, on two source pairs, are between separate measurement systems, where the data differences carry meaningful information.

Every classification is inferred from the data and written out with the evidence behind it (e.g., the number of overlapping cells, the median relative difference, the agreement share and the direction consistency). Each one can be checked or overridden.

Read the reconciliation

Comparisons by relationship

  • 1,035 Derived — the same estimate, republished
  • 49 Subset — different scope by construction
  • 143 Independent — separate measurement systems
  • 6 Insufficient — too few overlapping cells to classify
52
material divergences on comparable cells
33
cells set aside as immaterial or non-comparable
70
rows carrying a data-quality flag

How the panel is built

Phase 2 feasibility

Structured sources only

Figures come from machine-readable feeds and from published tables. Prose is never parsed for numbers and no language model is used anywhere in the pipeline. Where a country does not publish in a usable form, the site says so instead of estimating. Such an approach is designed to maximise data accuracy and avoid data contamination.

Read at source where possible

National programme reporting is the primary record; the international aggregators are the cross-check. Four Southeast Asian programmes are now read directly. The rest of the region is covered by the aggregator layer alone, and each country page states which applies.

Uncertainty is kept

Modelled estimates carry confidence bounds and those bounds are stored and drawn. Reporting periods that are still open are recorded, because a partial year plotted as a complete one results in measurement inaccuracy, as only provisional but not full funding in a given year is measured.

Thailand
Read at source
Vietnam
Aggregators only
Indonesia
Community monitoring
Philippines
Read at source
Myanmar
Access to be sought
Cambodia
Read at source
Laos
Access to be sought
Malaysia
Aggregators only
Timor-Leste
Access to be sought
  • At source — the national programme publishes in a form the pipeline can read.
  • Lead — a source exists but is not yet accessible in a usable form.
  • Aggregator only — the country appears in the panel through UNAIDS, WHO, the World Bank and OECD reporting.
  • The full assessment is in Methods.