I happened to be listening to EMRAP’s UC Max recently. They discussed a case where necrotizing fasciitis was overlooked, partially attributed to anchoring bias (ie the famous Pat Croskerry insights).
What if we had AI-enhanced triage notes that flagged potential Croskerry biases in free text?
For instance, instead of the triage note for the EMRAP case reading:
“18 yo female, feels like she may pass out. She is tearful. States this is the first time she has been away from home.”
The EHR could append to “Potential anchoring bias detected” on the fly, so it reads:
“18 yo female, feels like she may pass out. She is tearful. States this is the first time she has been away from home. [Potential anchoring bias detected.]”
I don’t mean to permanently save this in the note, but add it on the fly when relevant.
With the capabilities of GPT-4's function calling and speed, it's certainly seems within reach.