The future of ChatGPT and other AI applications in clinical environments is still in question for numerous reasons (PDF), including liability and privacy concerns, but these tools have already earned special roles in undergraduate medical education by helping educators tailor content to learners’ needs.
A plenary session at the AMA ChangeMedEd® 2023 conference featured a presentation about how educators at New York University (NYU) Grossman School of Medicine are using ChatGPT to further their precision education efforts, from boosting clinical knowledge to achieving educational goals.
Learn more with the AMA about precision education, which allows educators and learners to leverage data and technology to improve the personalization of education and the efficiency of learning.
“We have a series of precision medical education tools that are now in everyday use at our medical school,” said Marc M. Triola, MD, associate dean of education informatics and director of the Institute for Innovations in Medical Education at NYU Grossman School of Medicine.
One of them is DX Mentor, which gives physicians real-time access to Epic data on medical students and resident physicians. This lets them know the diagnoses that medical students are seeing as they're writing history and physical examination notes and other admission notes.
“Through a complicated series of informatics, we map those diagnoses to a series of medical education resources,” Dr. Triola said.
Each medical student then gets a personalized 8:30 a.m. email with AI-generated practice questions and relevant review and guideline articles from Amboss, Osmosis, UpToDate and PubMed based on their new cases.
ChatGPT takes it a step further by, at the click of a button, providing summaries of all the suggested articles’ abstracts.
“This is not meant to change the way that they learn, but to make it seamless and frictionless,” Dr. Triola said.
The chatbot operates behind the scenes too. NYU Grossman’s newest tool, StudyBuddy, can take any set of clinical texts—say, a patient note or a lecture transcript—and map it to a catalog of medical education resources, including USMLE-style self-study questions.
That data “is now freely available because companies like OpenAI packaged it all to train the large language models for ChatGPT,” Dr. Triola said.
NYU Grossman has also integrated ChatGPT into its medical student coaching application. As students enter their goals, ChatGPT gives suggestions for how to accomplish them.
“We have a button there where they can say: Hey, ChatGPT … give me some advice on this goal,” Dr. Triola said, noting that ChatGPT “gives remarkably good advice to our medical students.”
In one case, a medical student asked for a plan that would expand the student’s knowledge of physiatry, including by reading a particular textbook.
ChatGPT suggested a five-point plan that included joining academic clubs, setting up a weekly schedule to review relevant journal articles, reading and discussing with classmates or mentors one chapter or section of the textbook per week, participating in clinical rotations or observing physiatrists, and attending physiatry-related webinars and workshops. Four of the five even included an explanation of why that activity would be helpful.
“Not only does it understand what that book was, it knows what the right journals were in that particular specialty,” Dr. Triola said.
ChatGPT even recommended conducting some of the activities at specific clinics in the NYU hospital system.
“We didn't tell it any of this stuff,” he said. “It already knows everything about all of us because it's read every webpage that we've got.”
Learn more with the AMA about augmented intelligence (AI) in medicine.