Developing a Conversational Virtual Standardized Patient to Enable Students to Practice History-Taking Skills

Area of Science:

• Medical Education
• Artificial Intelligence in Healthcare
• Clinical Simulation

Background:

• Traditional virtual patient simulations often lack conversational dialogue.
• Virtual standardized patients (VSPs) integrate natural language processing for realistic patient interviews.
• VSPs allow students to practice history-taking and differential diagnosis similar to real patient encounters.

Purpose of the Study:

• To develop and evaluate a virtual standardized patient (VSP) system for medical education.
• To assess the VSP's ability to facilitate history-taking and differential diagnosis practice.
• To create a platform for analyzing student-patient interactions and information gathering.

Main Methods:

• Developed VSPs using a dialogue management system (ChatScript) and a game engine (Unity).
• Tested the VSP system with medical students at The Ohio State University College of Medicine.
• Recorded student-VSP interactions for analysis of questioning and documentation in an electronic medical record.

Main Results:

• VSP response accuracy ranged from 79% to 86%, varying with case complexity and student skill.
• Students successfully developed appropriate differential diagnoses based on VSP interactions.
• The system captured comprehensive data on student information-gathering approaches.

Conclusions:

• VSPs provide a valuable tool for medical students to practice clinical reasoning and history-taking skills.
• The VSP system enhances preparation for encounters with standardized or actual patients.
• Future work will focus on automated assessment and feedback mechanisms for VSP sessions.