Harvard Study Finds AI Model Beats Doctors in ER Triage Diagnosis

A collaborative study by Harvard Medical School, Stanford University and Beth Israel Deaconess Medical Center evaluated OpenAI’s “o1” large‑language model on 76 real emergency‑room cases from a Boston hospital. The AI correctly identified exact or near‑correct diagnoses in 67 % of triage cases, outperforming physicians who scored between 50 % and 55 %. When additional patient information was supplied, the model’s accuracy rose to 82 %, still modestly ahead of doctors. Researchers said the system excels in early triage when clinicians must decide quickly with limited data, but emphasized it is intended as a decision‑support tool rather than a replacement for physicians. The findings suggest AI could serve as a second‑opinion aid to improve diagnostic accuracy and treatment planning in emergency settings.