To state the obvious, artificial intelligence (AI) is everywhere – and use for health information is no exception. People are increasingly seeking health information and advice from AI.
Physicians are also embracing AI – an American Medical Association survey showed nearly three-quarters of physicians are now incorporating AI into their practices. With use becoming increasingly common, central Ohioans may have questions about the role of AI in their health.
Why are patients turning to AI for their health questions?
There are many reasons patients are turning to AI. Some of it is convenience – according to a Kaiser Family Foundation survey, the speed of information was a major reason for most respondents seeking health information via AI. At the same time, many respondents cited the cost of seeing a health care
provider and difficulty getting appointments. I acknowledge that AI may be filling a real gap for many users.
Should I use AI agents for health questions?
At this point, use of AI for health-related questions is unavoidable. Physicians want patients to be well-informed, and self-education often lives online. About half of Google internet searches include AI overviews – so users don’t have to be looking for AI-generated content to find it.
There are still great online resources on symptoms and conditions that don’t use AI – including familydoctor.org from the American Academy of Family Physicians and healthychildren.org from the American Academy of Pediatrics. But we know our patients are going to be interfacing with AI and should plan accordingly.
What concerns might your doctor have about using AI for health questions?
As a family physician, I take the responsibility for coordinating a patient’s care seriously. A difficulty in counseling patients on AI use is that their AI agents are not accountable to a health care provider (unless they directly connect a user to a clinician). Physicians have no way of vetting information patients receive from AI unless patients share it.
There are the headline-grabbing, sometimes tragic stories of users sharing plans for suicide only to AI chatbots or receiving inaccurate health information from AI output, but I also worry about the smaller-scale day-to-day fragmentation of patients’ health advice. Checking AI’s sources is also critical when possible. Even AI designed for use by health care professionals makes mistakes.
The chatbots may also experience an information gap. Their guidance depends on input from users and may lack the context of a user’s full medical and social history. Tech companies have created health chatbots specifically designed to connect data from electronic health records and wearables, but this raises the issue of privacy. If patients are sharing their medical records with an AI agent, they should read the fine print to understand exactly how that information could be used or disclosed.
How should I use AI agents for health care questions?
When it comes to navigating the logistics of health care (finding a nearby location, identifying covered providers), it is hard to see much personal risk to use AI beyond the risk of using a search engine, especially if a user is not providing sensitive health information.
Things get more complicated when the questions are related to health. A physician-researcher has developed a framework for assessing risk in these scenarios, which may be found at news.harvard.edu. The lowest-risk questions are general health questions that don’t rely on the interpretation of a professional (Example: “What are examples of high-fiber foods?”) – the kinds of questions for which AI is acting as an advanced search engine.
Unsafe questions involve asking AI to make or evaluate management decisions about a condition. Even health-oriented agents typically state they do not replace medical care from a human health care professional.
With caution, patients can use AI to prepare for a visit or to help understand their discussions with a medical professional. Critically, that professional is still in the loop directing care. If you’re spending more time discussing AI output than your symptoms or history with your physician, AI is probably not serving you well. But if it helps you save time and reach a shared understanding together, AI has the potential to improve your health care experience.
Andrew Detty, MD, is a board certified family medicine physician and a member of the Board of Directors for the Columbus Medical Association. This column is brought to you by the Columbus Medical Association, an organization uniting physicians and patients to build a healthier community for all. Submit health-related questions at columbusmedicalassociation.org/askthedoc.
This column is for informational and educational purposes only and does not constitute medical advice, diagnosis, or treatment. Always consult your physician or other qualified health care provider with questions or concerns about your health.
This article originally appeared on The Columbus Dispatch: How to safely use AI for health advice
