Harvard

Hāfó

哈佛

study / research

yánjiū

研究

to find / to discover

fāxiàn

发现

OpenAI (AI company)

OpenAI

OpenAI

model

móxíng

模型

in

zài

在

emergency room / urgent care

jízhěn

急诊

diagnosis / to diagnose

zhěnduàn

诊断

in / among

zhōng

中

to outperform / to surpass

chāoyuè

超越

doctors / physicians

yīshēng

医生

。

New Words

News in English

Harvard Study Finds OpenAI Model Outperforms ER Doctors in Diagnosis

A Harvard Medical School study published in Science found that OpenAI reasoning model o1 outperformed experienced emergency room physicians in text-based triage diagnoses, achieving accuracy of 67% compared to physician rates of 50-55%.

Researchers from Harvard Medical School and Beth Israel Deaconess Medical Center published a study in the journal Science comparing OpenAI's o1 reasoning model with the diagnostic abilities of internal medicine physicians in emergency settings.

The results showed that o1 provided accurate or nearly correct diagnoses in 67% of cases, whereas the two physicians had accuracy rates of 55% and 50% respectively.

The researchers emphasized that both the AI and the doctors had access to the exact same electronic medical record information without any preprocessing. However, the study also has clear limitations: the AI relied only on written text, while clinicians also need to reference medical images, sounds, and nonverbal patient cues.

The authors did not advocate for using AI directly in clinical decision-making but rather called for conducting prospective clinical trials.

This study is among the largest artificial intelligence emergency room diagnostic comparison studies to date, attracting widespread attention in the healthcare AI field.

More in Science & Technology