Maintenance Notice

Due to necessary scheduled maintenance, the JMIR Publications website will be unavailable from Wednesday, July 01, 2020 at 8:00 PM to 10:00 PM EST. We apologize in advance for any inconvenience this may cause you.

Who will be affected?

Currently submitted to: JMIR Medical Education

Date Submitted: Feb 20, 2026

Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.

The Effects of Generative AI Virtual Patient in Serious Illness Communication Skills: Randomized Controlled Trial

  • Kurtis G. Haut; 
  • Masum Hasan; 
  • Thomas Carroll; 
  • Ronald M. Epstein; 
  • Taylan Sen; 
  • Ehsan Hoque

ABSTRACT

Background:

Serious illness communication (SIC) is a critical skill in medical education, especially in end-of-life care, yet clinicians often lack sufficient opportunities for deliberate practice. Standardized patients (SPs) remain the gold standard for SIC training, but they are limited by cost, availability, and consistency. To support wider access to SIC training, scalable alternatives are needed.

Objective:

To evaluate whether an AI-powered virtual patient providing real-time dialogue and automated feedback (SOPHIE) improves SIC skills among clinicians and trainees.

Methods:

We conducted a single-blind parallel-group randomized controlled trial from June to December 2024 via videoconferencing. Participants were randomized 1:1 to intervention (SOPHIE training) or control (reading module). Blinded standardized patient actors and independent raters evaluated outcomes pre- and post-intervention. The intervention group completed three interactive SOPHIE modules (Empathize, Be Explicit, Empower) for 30 minutes, combining dialogue with an interactive virtual patient and automated feedback. Control participants reviewed reading materials covering the same framework for the same duration. The primary outcome was the change in communication skills (Empathize, Be Explicit, Empower) assessed by standardized patients and blinded raters using a validated rubric. Secondary outcomes included self-reported confidence in communication skills.

Results:

Fifty-one medical, nursing, and physician assistant students and practicing clinicians (mean age, 30.6 years; 78.4% women) from academic and clinical settings. Participants were stratified by training level and randomized; all completed the study. Compared with controls, SOPHIE participants demonstrated significantly greater improvement: Empower (Δ = 17% vs 6%; P = .004), Be Explicit (Δ = 13% vs 5%; P = .003), and Empathize (Δ = 14% vs 7%; P = .04). Here, Δ indicates mean improvement on the aggregate rubric score across 5 evaluators. The SOPHIE group demonstrated greater improvement, with effect sizes ranging from 0.59 to 0.92. SOPHIE participants also reported higher confidence in their skills (mean score: SOPHIE 4.31 vs Control 3.83 on a 5-point Likert scale; P = .009).

Conclusions:

Training with an AI-powered virtual patient significantly improved serious illness communication skills and confidence compared with reading modules. Generative AI-based simulations may provide a scalable, accessible, and effective adjunct or alternative to traditional standardized patient training in clinical education. Clinical Trial: Retrospectively registered. ClinicalTrials.gov NCT07409233; http://clinicaltrials.gov/ct2/show/NCT07409233


 Citation

Please cite as:

Haut KG, Hasan M, Carroll T, Epstein RM, Sen T, Hoque E

The Effects of Generative AI Virtual Patient in Serious Illness Communication Skills: Randomized Controlled Trial

JMIR Preprints. 20/02/2026:93034

DOI: 10.2196/preprints.93034

URL: https://preprints.jmir.org/preprint/93034

Download PDF


Request queued. Please wait while the file is being generated. It may take some time.

© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.