Human or LLM as Standardized Patients? A Comparative Study for Medical Education

TL;DR

This study introduces EasyMED, a multi-agent framework for virtual standardized patients that closely mimics human SP behavior and enhances medical training outcomes, offering a scalable and cost-effective alternative.

Contribution

The paper presents EasyMED, a novel multi-agent VSP system with a new benchmark, SPBench, and demonstrates its effectiveness in matching human SP behavior and improving medical education.

Findings

EasyMED outperforms existing VSPs in behavior consistency.

Training with EasyMED yields learning outcomes comparable to human SPs.

EasyMED offers improved flexibility, safety, and cost efficiency in medical training.

Abstract

Standardized patients (SPs) are indispensable for clinical skills training but remain expensive and difficult to scale. Although large language model (LLM)-based virtual standardized patients (VSPs) have been proposed as an alternative, their behavior remains unstable and lacks rigorous comparison with human standardized patients. We propose EasyMED, a multi-agent VSP framework that separates case-grounded information disclosure from response generation to support stable, inquiry-conditioned patient behavior. We also introduce SPBench, a human-grounded benchmark with eight expert-defined criteria for interaction-level evaluation. Experiments show that EasyMED more closely matches human SP behavior than existing VSPs, particularly in case consistency and controlled disclosure. A four-week controlled study further demonstrates learning outcomes comparable to human SP training, with stronger early gains for novice learners and improved flexibility, psychological safety, and cost efficiency.