0
OpenAI's o1 correctly diagnosed 67% of ER patients vs. 50-55% by triage doctors
(www.theguardian.com)
o1 ? That's from almost 2 years ago when reasoning was new.
this post was submitted on 03 May 2026
0 points (50.0% liked)
Hacker News
4988 readers
460 users here now
Posts from the RSS Feed of HackerNews.
The feed sometimes contains ads and posts that have been removed by the mod team at HN.
founded 2 years ago
MODERATORS
Turns out medical professional don’t look at basic data to make basic choices.
I’m married to a medical professional, and it sounds about right.
So 33% would be misdiagnosed and die. Cool.
Vs 40-45% by doctors.
So what we need are both working at the same time.
Not necessarily and this is something the article doesn't even bring up. While people are triaged before seeing the ER doc, the ER doc's job itself is to triage and determine whether the patient needs to be further worked up in the hospital or outpatient. Often times an ER doc may not have a clue about the diagnosis but the critical thing to be correct about is if the patient needs to stay to get worked up. They give an example about docs who think a clot is due to an anticoagulant when the AI thought it might be due to lupus. The ER doc's concern is that the patient be observed/admitted and the internal medicine team would likely make that catch (or the rheumatologist).