This med school instructor is still learning.
To efficiently utilize artificial intelligence to improve medical education, Nan Du MD’16, MPH, has happily embraced the roles of student and teacher.
Du, a pediatric gastroenterologist at Boston Children’s Hospital and instructor at Harvard Medical School, is part of a team developing and testing a large language model, called MedTutor, to give students on-demand access to engage with medical cases—particularly addressing the diagnosis and treatment of celiac disease. The assessment by Du and her peers includes a comparison to previously created, open pediatric cases.
“My job is to build multiple sample cases for the AI model,” says Du, who earned her bachelor’s in biomedical engineering at Harvard. “Can we help accelerate learning by overcoming a limited number of instructors?” She has been writing pilot cases for the project, which is funded by a grant from the Celiac Disease Foundation.
Du, who was born in England, grew up in Barrington, RI. Her mother worked as a research technician in the labs of Brown medical school faculty.
“It was lovely to go to Brown and have a 15-minute drive to my apartment,” Du says. “It was great to learn and serve in the state that I grew up in. I volunteered at Hasbro [Children’s Hospital] in high school, gathering supplies and walking patients to tasks. Coming back as a med student, I knew the hospital very well.”
Pediatric gastroenterology offered Du the benefits of being able to work with kids while also working with her hands, doing endoscopies, colonoscopies, and other procedures. “What I love about GI is, at its most basic level, it is one tube carrying nutrients and calories. But it’s much more complicated than that,” she says. “I was lucky to have mentors who helped hone my interests.”
Theresa Fiorito F’17, MD, was a pediatric infectious diseases fellow at Brown when she collaborated on two research projects with Du. The first was connected to a 2015 outbreak of meningitis B at Providence College. “Nan helped me with research on adverse effects of the vaccine, which helped me write the paper,” which was published in The Pediatric Infectious Disease Journal in 2018, Fiorito says.
They then worked together on a poster that they presented at both the 2017 Yale Pediatric Research Forum and the American Society of Tropical Medicine and Hygiene, about malaria surveillance in Rhode Island. “Nan is diligent and always eager to learn more,” Fiorito says. Du’s current team effort is a chance for her to return to those student roots. “What is super fascinating about AI is the unlimited opportunity with it when you put it in a medical education lens,” she says. “You get overwhelmed with possibilities for teachers and learners. It allows us to get better insight on how students think.”
In addition to clinical reasoning, MedTutor could help medical students learn to communicate with patients about diagnoses and other difficult topics, she says. “It’s helpful to have practice sessions before you are talking with a patient about a lifechanging situation,” Du says. “That includes going to a lifelong gluten-free diet for celiac disease or new diagnosis of inflammatory bowel disease.”
AI can fill in the gaps in medical education that trainees may experience due to duty hours or their many conflicting responsibilities, she adds.
“It makes sure they are able to experience these important cases and they are able to slot it in their minds as something they experienced before they see it as an attending or fellow,” Du says. “Ideally it might allow us to train more physicians, as well as adjust to the changing demands on medical education.”
