Related Experiment Video
Updated: Jan 12, 2026
05:04
Author Spotlight: Evaluating Clinicians' Adoption of Ultrasound-Guided Vascular Cannulation Through Simulation Training
Published on: August 9, 2024
1.4K
LLM-based pedagogical agent for ICU simulation instructor training: A quasi-experimental study
Jingbang Liu1, Ting Chen1, Shan Li1
1Nursing Department, Sir Run Run Shaw Hospital, Zhejiang University School of Medicine, Hangzhou, 310000, Zhejiang Province, China.
Nurse Education Today
|November 6, 2025
Summary
Integrating large language models (LLMs) into Intensive Care Unit (ICU) nursing simulation training improved perceived simulation design and self-efficacy. Usability was comparable to traditional methods, suggesting feasibility for enhanced instructor development.
Area of Science:
- Medical Education Technology
- Artificial Intelligence in Healthcare
- Nursing Simulation
Background:
- Intensive Care Unit (ICU) nursing requires advanced skills, with simulation-based training being crucial but limited by cost, time, and faculty capacity.
- Large language models (LLMs) offer potential to enhance ICU nursing education through rapid scenario generation and on-demand assistance.
- Limited real-world data exists on the effectiveness and usability of LLM-based agents in ICU instructor training.
Purpose of the Study:
- To assess the feasibility of incorporating an LLM-based pedagogical agent into simulation instructor training for ICU nurses.
- To evaluate the impact of LLM integration on learner-perceived simulation design quality and online learning self-efficacy.
Main Methods:
- An exploratory quasi-experimental study involved 40 ICU nurses, randomly assigned to an LLM agent group (n=20) or traditional blended learning group (n=20).
- Training effectiveness was measured using the Jeffries Simulation Design Scale (SDS), System Usability Scale (SUS), and Adult Online Learning Self-Efficacy Scale.
- Data analysis employed Wilcoxon rank-sum tests and t-tests to compare outcomes between groups.
Main Results:
- The LLM group demonstrated significantly higher scores in case authenticity, scenario complexity, feedback mechanisms, interactivity, and teaching objectives on the SDS.
- Participants using the LLM agent reported greater self-efficacy in learning ability and learning technology compared to the control group.
- Teaching satisfaction was high in both groups, with no significant difference in System Usability Scale scores.
Conclusions:
- Embedding LLM-based pedagogical agents in ICU simulation instructor training is feasible and enhances perceived simulation design and online learning self-efficacy.
- Usability of LLM agents was comparable to traditional blended learning methods.
- Further multi-center randomized controlled trials are needed to confirm efficacy and the independent contribution of LLMs.
