Project Details

Performance and usability of an AI digital stethoscope for diagnosing childhood pneumonia in Africa

Eric McCollum, MD

Summary

BACKGROUND: Each year, pneumonia and other lower respiratory infections kill roughly 700,000 children under five worldwide, with sub-Saharan Africa bearing about half of these deaths. The World Health Organization’s Integrated Management of Childhood Illness (IMCI) guidelines diagnose pneumonia using a simplified clinical algorithm that is only about 60% accurate, leading to misdiagnosis, preventable deaths, and antibiotic overuse that fuels antimicrobial resistance. GAP: IMCI excludes lung auscultation because conventional stethoscopes are subjective, vary between listeners, and are difficult for frontline health workers to use. No study has prospectively tested whether a commercially available, AI-enabled digital stethoscope can overcome these limitations and improve pneumonia diagnosis for children in Africa. HYPOTHESIS: We hypothesize that IMCI augmented by an AI digital stethoscope system will achieve at least 80% accuracy for diagnosing pneumonia and will be acceptable to healthcare workers in Africa. METHODS: In a prospective observational study nested within a multi-country randomized trial (ToolCAP), we will enroll 1,100 children aged 2 months to 10 years with IMCI-defined pneumonia across four sites in South Africa, Tanzania, and Senegal, comparing AI-classified lung sounds against lung ultrasound-confirmed pneumonia and assessing device usability with healthcare workers. RESULTS: Pending. IMPACT: If validated, AI-enabled auscultation could turn pneumonia diagnosis into an objective, scalable tool that reduces unnecessary antibiotic use, helping curb antimicrobial resistance and improve survival for children in low-resource settings worldwide.