I don't know how much compute you want to use but if it could automatically suggest intelligent questions based on what it hears in the visit that would be nice - e.g. if it hears I diagnosed migraine, little pop-ups could appear like "What is evidenced-based management for treating migraine for this patient?" or “What red flags would suggest secondary headache and what investigations would be appropriate?” or “Differentiate migraine vs tension-type vs cluster headache using this presentation.” Then you could click on those, maybe edit them before you submit, and get a response.

Alternatively, I could just type those in myself. Maybe a compromise would be a button to create suggestions, rather then having it continuously checking through the transcript?

I once saw a patient with a persistent cough who wasn't my regular patient. Only at the end of the encounter did I notice that it said on the EMR that they had asthma. It would be really cool if AI could bring important information to my awareness based on the discussion in the room and the EMR information - past medical history, recent ER visits, recent lab results, relevant recent encounters, etc. I'm not exactly sure how you would implement this in Heidi though. I imagine every 5 minutes or so it takes the transcript, compares it with the EMR data, and I get a little pop-up "This patient has asthma; they had an ER presentation for asthma twice in the last year."

The only really repetitive question I have is help when I am stumped with the diagnosis. Currently here is my process: I put this request into Heidi: "create a concise case summary, as if you were a medical resident summarizing the patient to your attending". Then I paste that into a custom version of Chat-GPT I have, that uses the following instructions in the attached text file.

Hey Matt,

I'm working on this literally right now, thanks for your comment.

A question, do you think it's the same queries every time? or would you like us to suggest some things based on what's happening in the visit? Also would you like evidence suggestions applied in your outputs and/or references added that back up your plan/patient explainer/letter etc?

Thanks your inputs will help me build it