• MalReynolds@slrpnk.net
    link
    fedilink
    English
    arrow-up
    3
    arrow-down
    1
    ·
    edit-2
    3 hours ago

    Not to mention malpractice insurance up the wazoo… As it be should for doctors.

    Just pay for perfectly functional, deterministic, local, voice transcribing like dragon dictate and actually read it afterwards for errors. Or hire a person with an NDA. Sending privileged information overseas to known data miners for transcription is pants on head bonkers given the alternatives.

    Need to have a chat with my doctor to see if he’s one of the 40% (there’s a fucking scary number from the RACGP, talk about burying the lede).

    • SuspiciousCarrot78@aussie.zone
      link
      fedilink
      English
      arrow-up
      1
      ·
      edit-2
      2 hours ago

      I’m with you on this I think.

      I have no problem with anyone using an AI scribe (though I would prefer one that was on device rather than cloud based). I am aware of things like Lyrebird Health that integrate with EHR management software - frankly, anything that allows the practitioner to focus more solely on the patient is a good thing. After all, they are meant to be treating the patient in front of them, not the computer screen.

      The prior point about legal liability is accurate IMHO. Medical health records are functionally a legal record, and should be treated as such. Responsibility for review, redress of inaccuracies etc cannot be waved away as “ChatGPT did it”. If the practitioner is willing to take the onus of that on, and treats the scribed document with the same fidelity, chain of provenance etc as other records, I’m probably ok with it.

      Requiring patients to consent to cloud-based AI scribing as a condition of access is where it gets uncomfortable, and your point about local alternatives is exactly why. If deterministic, on-device transcription exists and does the job, the justification for mandating a cloud pipeline through a psychiatric service gets pretty dicey, pretty fast.

      I think I can see a way to have Dragon Dictate record the audio, convert it to text and then have on device AI pull out relevant bits to populate a template. That doesn’t abrogate the need to actually LISTEN to the patient but it might fix that ‘capture’ part of the funnel.