Study of AI triaging in ED

Public eHealth News
1

I found this article quite interesting. AI used to triage in EDs (USA).

  • Patients had a reduced wait by 2 hours in going to critical care.
  • Reduced ED waiting times.
  • Improved triage accuracy.
  • Staff who aligned with AI recommendations tended to perform better when used as a clinical decision support tool.

What if AI ran ER triage? Here’s how it sped up patient care in real-world tests | ZDNET

It would be interesting to see more research done in this space to back the need in public hospitals.

2

I haven’t found the full article but have read the abstract. I’m quite sceptical of the the ZDNet writeup of it and the claim that using the AI triage system sped up transfer time to a critical care unit 1) that’s what EDs are best at already, and 2) the ZDNet article also claims that the AI triage system resulted in reduction of time to discharge “by as much as 82 minutes on average” (not sure how you can use “as much as” and “on average” in one sentence), but the abstract mentions a reduction of average time to discharge from 311 minutes to 292 minutes, a savings of 19 minutes. I suspect that the ZDNet writer didn’t really understand the paper.

There is already a fair amount of literature about AI in triage. And a few years ago a system for doing it was developed in Nelson-Marlborough District that was trained on their local data. Their audits showed increased accuracy of triage, but it didn’t receive any funding to continue there and wasn’t picked up by any of the other districts that it was shown to.

3

Thanks for your thoughts. It did cross my mind when they mentioned the reduced time to transfer a patient to critical care.
I wasn’t aware that this was triaged in Nelson-Marlborough area. That’s a huge shame that it wasn’t picked up as it would definitely help with the general uptake of AI and becoming more digitalised.

1 Like