5 tips for clinicians on patients using ChatGPT

Your patients will turn to ChatGPT for questions about their health and there is nothing you can do to stop it. Here are five ways the healthcare system should adapt in response:For every story of ChatGPT helping patients failed by the healthcare system, there’s another of it giving terrible advice.Patients often overhype the former while conveniently ignoring the latter.But clinicians often sound the alarm on the latter while conveniently ignoring the former.The reality is that this clash between hype and alarm reflects an old tension around democratizing medical advice—only now it’s accelerated by AI.Before ChatGPT it was Google. And we had the same conversations about that too.But can you blame patients?Getting medical advice is often expensive or unavailable.Read: Dr. Joshua Liu: Dr. Artificial Intelligence More than that, if we are being honest, many of us “expert” clinicians often go to ChatGPT for expert advice.Many of us have asked ChatGPT for legal advice instead of paying for a lawyer. Or plumbing advice instead of paying for a plumber.I’m sure ChatGPT has told us some awful advice on those topics but as non-experts we often have no idea.And are any of us going to stop doing that?No . . . we will just believe that “the AI will only get better with time”.And certainly our patients will feel the same way about ChatGPT.So if we can’t and shouldn’t stop patients using ChatGPT . . . what should the healthcare system do to adapt?My five suggestions:Clinicians should educate patients on what questions ChatGPT is safe/effective for, and when they really need to see a clinician for help: e.g. “You can rely on it for low-risk issues but you really should see me for red-flag symptoms.”Clinicians should educate patients on effective ways to write prompts that minimize hallucinations and bias—e.g.: "Please give me information based only on well-established, evidence-based medical guidelines and peer-reviewed research. Summarize the most common explanations and treatments in plain language. Highlight any areas where the evidence is uncertain or controversial.”Clinicians should expect to face more external pressure to stay up to date with the latest evidence and tools. It’s not a good look if patients are asking you about changes in practice that you either don’t know about or you struggle to access relevant information quickly/easily because you’re unwilling to use the latest tools (e.g. OpenEvidence).Clinicians should be pleased to be freed up from phone calls and patient portal messages for simple things—and patients who otherwise didn’t do much self-management or even considered using LLMs are now more empowered.Health systems should embed conversational AI into their digital health tools in a manner that is safe/effective—e.g. how we are embedding conversational AI into SeamlessMD so patients can easily ask questions about their care plans that have already been vetted by their own care team.What’s your take?Dr. Joshua Liu is a Toronto physician and the co-founder/CEO of SeamlessMD. This item originally appeared on LinkedIn.

5 tips for clinicians on patients using ChatGPT

Register for free access or login.

Related Topics