HealthPulse AI Interpreter — Uganda's Blinded Pilot for AI-Assisted Malaria Test Reading

Malaria rapid diagnostic tests (RDTs) are read visually by community health workers, and misreadings — especially faint positive lines — are a persistent source of misdiagnosis and surveillance error in high-burden districts. Uganda's Ministry of Health, together with the global digital-health nonprofit Medic and health-AI company Audere, integrated Audere's HealthPulse AI Interpreter — a computer-vision tool that photographs and automatically classifies RDT results — into the national electronic Community Health Information System (eCHIS). The pilot runs in four districts (Mpigi, Moyo, Buikwe and Lira), with support from the Gates Foundation and a parallel deployment with Kenya's Ministry of Health.

Phase I, running roughly four months to August 2026, deliberately withheld the AI's interpretation from Village Health Teams so their manual reading could be compared against the AI's result without bias. Phase II, a three-month follow-on, will for the first time show Village Health Teams the AI's interpretation alongside their own, testing whether visible AI assistance improves decisions or simply displaces human judgement.

A feedback and synthesis workshop following Phase I reported 'very high agreement' between the AI and Village Health Team readings and growing user confidence, but no agreement percentage, sample size or case count has been made public.