Australian general practitioners (GPs) are rapidly adopting Medical AI Scribe software. This is an artificial intelligence program that captures consultations in real-time by recording in the background and generating detailed notes ready to go into a patient's medical record.
While it may initially seem like an advanced dictation tool, as a GP, I'm finding that it's transforming clinical practice in ways that go far beyond basic transcription.
In my medical career spanning over three decades, medical records have evolved from illegible scribbles on cards to electronic records -- and AI scribes represent the next big leap.
I think of it as the latest evolution of clinical consultations, profoundly shaping the way doctors document and provide care.
AI as a conscientious listener
One surprising advantage is that when I review the notes generated by the software, I often find small but significant details that I may have missed in real-time while focusing on my next question or the management plan.
For example, I might have overlooked a comment about the patient's lack of sleep or a past reaction to medication.
It's like having an extra set of ears in the room, quietly catching additional symptoms or bits of social context that a patient mentions in passing. This extra layer of listening helps ensure important comments don't slip through the cracks, helping me catch things I might otherwise miss.
That said, as with all AI, doctors have to be cautious.
AI can "confabulate," acting more as a storyteller than a documenter. This can create very plausible-sounding facts that can easily slip past a vigilant clinician.
Every clinician should never forget that, ultimately, the buck stops with them.
AI as a trusted clinical assistant
As with any groundbreaking technology, what starts as optional quickly becomes essential.
Just as we rely on Google Maps to navigate or Uber Eats to order a late-night pizza without thinking about the mechanics, AI scribe software soon feels indispensable, seamlessly integrating into clinical workflows.
Before long, it's hard to imagine practicing without it.
It saves time and, like any technology, fades into the background, allowing more cognitive space to focus on important aspects of the consultation.
For a technology that's going to be so deeply embedded in our health care system, we must ensure in Australia that we are well-prepared -- developing our own national AI expertise rather than simply relying on overseas technologies for such an important role in our health system.
The structure of information influences our thinking, and this is especially true in clinical reasoning.
With medical AI software, the process of organizing notes isn't just about documentation -- it also shapes my clinical understanding of the patient. For example, when the software organizes information into clinical issues with the various relevant facts and symptoms grouped under each theme, this shapes my thought processes.
In this way, AI is not merely serving as a record-keeper but is already an active part of the clinical reasoning process.
AI as a virtual GP supervisor
Using AI scribe software feels a bit like having a GP supervisor or a medical student in the consulting room. It reminds me of the Hawthorne effect -- where research participants modify their behavior when they know they're being observed.
I already sense that I'm subtly adjusting my consulting style.
Knowing that the consultation is being recorded and transcribed for my own reviewing adds a layer of accountability and encourages me to be more thorough.
Even in straightforward cases, like a respiratory infection, I'm more likely to ask all the classic systems review questions as if in a medical student examination. I'll talk through my findings in detail, and even announce aloud negative findings like "chest clear, no wheezing, no crepitations."
I'd be interested in future research on whether having AI software in the consulting room impacts consulting style or if patients notice any differences. They may end up with more questions: "Why is my doctor saying 'no crepitations'? What is a crepitation?"
The future potential for AI to help tailor medical education also seems extraordinary.
I was simultaneously complimented and shocked when the software recently advised me after a long mental health consultation that it was impressed with my empathetic approach.
Cautions on data security and privacy
While the benefits of medical AI software are clear, as with any new technology, we're in the honeymoon period, and some risks will likely become more apparent.
It's essential to consider patient consent and data privacy. I make it a point to explain to patients if I'm using the software and document their consent.
So far, I am finding almost all patients are comfortable with it, but they must be informed about what's being recorded, how it's used and their right to opt out if they prefer.
Data security is another critical factor, as these tools handle sensitive medical information that must be securely stored and transmitted to prevent unauthorized access.
As we become familiar with new tools, we develop routines and workflows that maximize their strengths while being mindful of their limitations.
Working with AI is no different.
For the next generation of clinicians, AI will be a core skill, underscoring the need for its integration into medical training and ongoing professional development.
If we can establish appropriate regulatory frameworks and learn to use AI scribes as an ally, it has the potential to deepen clinical reasoning, help structure our thinking and improve patient care.
