Figment Learning Labs Virtual Patients Outperform Written Cases in New Randomized Trial

Talking to an AI Patient Helped Health Students Learn More Than Reading About One, Study Finds

A randomized trial at the Medical University of South Carolina, published today in JMIR Nursing, compared interactive virtual patients with traditional written case studies.

For generations, students in medicine, nursing, pharmacy, and other health professions have learned patient care partly through written case studies: a page describing a fictional patient, followed by questions about what to do next. It's a reliable teaching tool, but real patients don't hand you a tidy summary. They talk, they worry, they leave things out, and the important details come out only if you ask the right questions.

A new study published today in JMIR Nursing tested whether letting students actually interview a patient, rather than read about one, makes a difference in how well they learn. The early answer is encouraging.

What the researchers did

The team, led by Dr. Jeff Borckardt of the Medical University of South Carolina (MUSC), enrolled 74 students from across the health professions, including allied health, medicine, pharmacy, nursing, and dental medicine. Fifty-eight completed every part of the study.

Everyone started with the same 12-minute video on pain management, covering types of pain, evidence-based treatments, the risks of opioid prescribing, disparities in how pain is treated, and how to talk with patients about pain and opioids. Students were then randomly assigned to one of two groups.

One group read three written case studies. The other group met the same three patients as interactive virtual patients in a simulated telehealth visit on the Figment Learning Labs Virtual Clinic. Using voice or text, students interviewed an animated, AI-powered patient, reviewed the patient's medical record, and ordered things like vital signs, lab results, and imaging, just as they would in a real clinic. The patient's facial expressions even changed with their emotional state during the conversation.

The key design choice: the case content was identical in both groups. Same patients, same histories, same test results. The only difference was whether students read the information or had to draw it out through conversation.

"We wanted a fair test," Borckardt said. "Same content, same learning goals. The only real difference was whether the patient was on a page or in a conversation."

What they found

Both groups learned, but students who worked with virtual patients showed greater gains on a knowledge test taken before and after the exercise.

The gap was clearer when students had to apply what they'd learned. After each case, students answered questions about the patient: what kind of pain they had, whether it was acute or chronic, and whether they were a good candidate for opioid therapy. Students in the virtual patient group outperformed the written case group by roughly 25%.

Students noticed the difference too. The virtual patient group rated their experience significantly higher on 13 of 17 survey items. Compared with students who read written cases, they were more likely to say the activity helped them practice clinical decision-making, helped them apply classroom knowledge to a practical situation, and taught them techniques that will help them in the real world. They also rated the experience as more engaging, more interesting and fun, and more satisfying, said they were more motivated to use activities like this in the future, and were more likely to recommend it to their peers.

"What struck me was the applied piece," Borckardt said. "Knowing a fact and using it with a person in front of you are two different skills. When students have to gather the information themselves, the way they would in a real exam room, it seems to stick differently."

Keeping the AI accurate

AI language models can sometimes produce information that sounds convincing but isn't correct. The researchers addressed that by having instructors, not the AI, supply every clinical fact in each case: the history, the medications, the lab results, and the correct answers. The AI's role was to portray the patient, with instructions to stay in character and not introduce information outside the case. No student-identifying information was shared with the AI provider.

"The AI is the actor, not the author," Borckardt said. "Faculty build each case and provide every medical fact. The AI brings the patient to life so students can practice the conversation."

Why it matters

Health professions programs face real constraints. Clinical placements are limited, and programs that use trained actors as "standardized patients" are valuable but expensive and hard to scale. Virtual patients could give students many more chances to practice difficult conversations, on their own schedule, with no risk to a real person.

"Nobody is suggesting this replaces real clinical experience or human standardized patients," Borckardt said. "Think of it like a flight simulator. Pilots still fly real planes, but you'd want them to have had plenty of safe practice first. We want students walking into their first difficult conversation having already had a few."

The fine print

The authors are careful to call these results preliminary. The study was conducted at a single university with a modest number of students, and it measured learning only right after the activity, not months later. About one in five students who signed up dropped out before reaching the case studies, for reasons that weren't recorded. The researchers also didn't track how much time each student spent, so it's possible the interactive group simply spent longer with the material.

Most importantly, the study can't say whether the benefit came from the AI itself or from interactivity in general. The authors describe it as evidence for the interactive virtual patient format as a whole, not proof of an AI-specific advantage.

"The next step is for other people to test this too," Borckardt said. "At other schools, with larger groups, following students over time, and comparing against interactive formats that don't use AI."

The paper discloses that Borckardt and co-author Dr. Debra Henninger-Borckardt are founders of Figment Learning Labs, the company that commercializes the platform, which was originally developed at MUSC with NIH funding.

The study was supported by the National Institute on Drug Abuse at the National Institutes of Health through a Small Business Technology Transfer (STTR) award, and the trial is registered at ClinicalTrials.gov(NCT06624124).

Read the full study: Borckardt JJ, Henninger-Borckardt D, Barth K, et al. "Knowledge Acquisition, Case Discussion, and Engagement in Health Professional Students Using Interactive Virtual Patient Cases Versus Written Case Studies: Randomized Controlled Trial." JMIR Nursing, 2026;9:e98700. nursing.jmir.org/2026/1/e98700