Harvard: AI Outperformed Physicians in Initial ER Triage

Friends — a note from the AI frontier: a Harvard Medical School and Beth Israel study (Science) compared large language models with ER physicians.

Study:
- 76 ER cases analysed; diagnoses were compared blinded against models o1 and 4o.
- Model o1 equalled or outperformed physicians, notably in initial triage.
- o1 produced an accurate/close diagnosis in 67% of triage cases versus 55% and 50% for the two clinicians.
- Authors emphasise: not a human substitute—prospective trials and liability frameworks are needed.

Why it matters: AI can strengthen triage and decision support but requires validation and regulation.

Where would you pilot — triage or another process?

#AI #medicine #healthtech #biotech