ChatGPT Health performance in a structured test of triage recommendations
Summary
Nature Medicine evaluates ChatGPT Health's triage performance using 60 clinician-authored vignettes across 21 domains under 16 conditions (960 responses). The study finds an inverted U-shaped performance with major safety concerns, including under-triage of emergencies and bias in crisis messaging, underscoring the need for prospective validation before consumer AI deployment in healthcare.