JMIR Publications

ABSTRACT

Background:

Effective pain and opioid management education remains a persistent challenge in health professions training. Static case studies and lecture-based instruction are widely used but may be insufficient to develop the clinical reasoning and communication skills required in practice. Interactive virtual patient simulations, including those powered by large language models (LLMs), offer scalable alternatives, but head-to-head comparisons of interactive and static case formats using identical content in interprofessional learner populations remain limited.

Objective:

This randomized controlled trial examined whether learner knowledge, case discussion quiz performance, and engagement differed when educational case studies were delivered through an interactive, AI-enabled virtual patient format versus a static written format. Because the formats differ in interactivity as well as in their use of AI, the trial was designed to test the effect of an “interactive virtual patient format” rather than to isolate a distinct AI-specific effect.

Methods:

Health professional students from medicine, nursing, dentistry, pharmacy, and allied health programs at a US academic health sciences university viewed a 12-minute video on pain management and were then randomized 1:1 to complete either (1) 3 static written case studies or (2) 3 interactive virtual patient case studies using the Figment Learning Labs Virtual Clinic. Case content was identical across conditions; only the delivery format was manipulated. The virtual patient system used an AI-enabled conversational interface (OpenAI GPT-4o) delivering facilitator-authored case material, including a viewable medical record (history, medications, allergies, and orderable assessments). Knowledge was assessed pre- and post-intervention using an 8-item test. Case discussion quiz accuracy and learner engagement (17 items) were also measured. Of 74 enrolled participants, 58 (78%) completed all study components and were included in this completer-based analysis. Analyses included a 2×2 mixed analysis of variance (ANOVA) and independent-samples t tests (α=.05).

Results:

A significant time (pre-post) × group (interactive vs written) interaction was observed (F1,56=4.39, P=.04; partial η²=0.07), indicating greater knowledge gains in the interactive condition. Participants in the virtual patient group demonstrated higher case discussion quiz accuracy (mean 0.79, SD 0.15) than those in the written case group (mean 0.64, SD 0.11; t56=4.48, P<.001). Agreement scores were significantly higher in the interactive group on 13 of 17 engagement items (P<.05), including items addressing realism, enjoyment, and perceived clinical relevance; because these were uncorrected item-level comparisons, they should be read as supportive rather than definitive.

Conclusions:

In this preliminary, hypothesis-generating trial, exposure to interactive virtual patient cases was associated with higher engagement and higher short-term knowledge test scores than static written cases. The design cannot isolate the independent contribution of AI, conversational interactivity, or time-on-task; the findings therefore support the value of an interactive virtual patient “format” rather than a distinct AI-specific advantage. Because the analysis was completer-based with unexplained attrition and 2 authors are founders of the company commercializing the platform, independent replication in larger, multi-institutional, and longitudinal samples is needed. Clinical Trial: ClinicalTrials.gov: NCT06624124; Reviewed and approved by the Medical University of South Carolina Institutional Review Board for the Protection of Human Subjects.