Uganda's AI Malaria-Diagnosis Pilot Shows Strong Results, Enters Harder Second Phase

A pilot programme that uses AI to help Uganda's community health workers interpret malaria rapid diagnostic tests has recorded very high agreement between the AI's readings and workers' own assessments after its first four-month phase, and is now moving into a tougher test: whether showing workers the AI's verdict in real time actually changes what they do.

Built Into the National Health System

The tool, developed by global health-tech nonprofit Medic together with AI diagnostics group Audere and Uganda's Ministry of Health, is embedded directly in the country's electronic Community Health Information System (eCHIS), which already guides roughly 16,297 Village Health Teams nationwide. The pilot ran across four districts — Mpigi, Moyo, Buikwe, and Lira — using smartphone computer vision to help health workers read malaria rapid diagnostic tests and assess febrile illness in children, a task where misreads can delay correct treatment.

From Blind Comparison to Live Guidance

In Phase 1, health workers made their own assessment first, and the AI's interpretation was compared afterward without being shown to them in real time — a deliberate design to test the AI's accuracy without influencing worker behaviour. That phase found very high agreement between the AI and worker assessments, alongside a feedback workshop showing growing worker confidence in AI-assisted interpretation, though implementers also flagged lower-than-projected testing volumes, supply shortages, device malfunctions, and connectivity gaps. Phase 2 changes the design: workers will now see the AI's interpretation alongside their own in real time, testing whether a visible AI recommendation changes clinical decisions rather than simply agreeing with them after the fact.

Part of a Wider Regional Push

The programme, backed by the Gates Foundation, was announced in April 2025 alongside a parallel deployment in Kenya, where the same eCHIS architecture guides roughly 106,320 community health workers nationally. "By partnering with the Gates Foundation, Audere, and Ministries of Health with bold agendas, we're bringing Audere's AI-powered diagnostics to the last meter," said Medic interim CEO Dykki Settle at the time. Audere CEO Dino Rech called it an opportunity to advance AI-powered diagnostics "through collaboration with esteemed partners like Medic."

Why it matters

Malaria case counts have been rising across Africa since 2020, and community health workers are often the first and only point of contact for a sick child in rural areas. Uganda and Kenya's willingness to embed AI directly into government-run health-worker systems, rather than pilot it in isolated clinics, points to a model where AI diagnostics could scale to reach well over 100,000 frontline health workers if Phase 2 proves the tool changes outcomes and not just agreement rates.