The landscape of modern clinics has changed significantly over the past decade, especially with the advent of artificial intelligence. With large language models like ChatGPT and Bing chatbots helping create personalized, comprehensive answers to everyday medical questions, it's becoming increasingly rare to rely on Google for medical questions. . As a family physician, I ask myself how this tool can help not only in the exam room, but also for patients once they leave the office. Previously, when a patient came in with her Google findings, much of the consultation consisted of pulling out pearls of reliable information from a slurry of online misinformation. Patients went home with handouts filled with relevant but overly complex information, and usually never looked at them again. Recent research suggests that artificial intelligence may play a key role in bridging the gap between what patients take home and what doctors think they know.
Artificial intelligence is revolutionary in the sense that it can build a foundation of knowledge from the web and then over time create its own connections and form new narratives. But how accurate are these stories? Considering all the misinformation that exists online, if a medical question is posed to a language model like ChatGPT, can it provide reliable information to patients? Guidelines recommend that women undergo mammography screening until age 74, but recommendations for women over age 74 are nuanced and vary by organization. In a recent study, a team of six clinicians with expertise in general medicine, family medicine, geriatrics, population health, cancer control, and radiology looked at ChatGPT responses to questions about mammograms in people 74 and older. We evaluated how appropriate it was. This study found that: ChatGPT returned an appropriate response 64% of the time. We also found that answers were incorrect 18% of the time, and the remaining answers were unreliable or lacked true consensus.
For simpler medical questions, large language models seem to be much better. A study by a team of radiologists found that the Bing chatbot could easily handle questions about imaging tests, with 93 percent of answers completely correct and 7 percent mostly correct. The reliability of your answer can depend on several different factors. 1) avoid asking too many questions when asking AI; 2) some AI platforms have been shown to fabricate information, while others cite the sources from which they derive the information; 3) Field of Information In medicine, where there are no clear answers, AI may not provide reliable and appropriate responses.
Another place where AI has been shown to thrive is in patient education. Extensive language models can be used to make patient education materials more streamlined and easier to understand, and in some cases even have the ability to translate this information into another language. Medical jargon can overcomplicate patient instructions and cause miscommunication between patients and clinicians. The ability of an extensive language model to interpret and clarify patient education materials lowers communication barriers without prohibitive time and expense. As a primary care physician, I'm always looking for different places to connect my patients with trusted medical information when they want more information before or after a visit. I think that with some improvement in reliability, artificial intelligence could be the glue that connects my patients to more meaningful visits and addresses their health concerns together.
Olivia Hilal is a family medicine resident.