Enhancing Medical Training with AI-Driven Scenario Generation

Medical errors are currently the third leading cause of death in the United States, with more than a quarter of a million patients dying each year from these mistakes [1]. These errors are often attributed to human factors such as lapses in attention. However, medical personnel can be trained to det...

Full description

Saved in:
Bibliographic Details
Published in2025 IEEE International Conference on Digital Health (ICDH) pp. 11 - 20
Main Authors Smith, Gordon, Pho, Katalina, Harrison, Ka Yan, Kass, Steven, Reichherzer, Thomas
Format Conference Proceeding
LanguageEnglish
Published IEEE 07.07.2025
Subjects
Accuracy
Electronic healthcare
Health Care
Heuristic algorithms
Human factors
Large language models
LLM
Personnel
Prompt engineering
Refining
Scenario generation
Situational Awareness
Training
User Study
Online AccessGet full text

Cover

More Information
Summary:Medical errors are currently the third leading cause of death in the United States, with more than a quarter of a million patients dying each year from these mistakes [1]. These errors are often attributed to human factors such as lapses in attention. However, medical personnel can be trained to detect errors early and take corrective action to prevent patient harm. Building training lessons and making them readily available as the need for additional training arises is a time-consuming task for trainers. In this paper, we present an algorithm to automatically build training lessons for nursing education using large language models (LLMs). Our algorithm is based on advanced prompt engineering techniques that incorporate storytelling techniques to create innovative training scenarios designed to enhance the situational awareness of nurses in dynamic clinical settings. The algorithm has been evaluated in a user study that requires nurse trainers to evaluate training scenarios built by subject-matter experts (SMEs) and our algorithm. The results showed that nurse trainers were not able to distinguish between human and AI-generated scenarios, rating the AI-generated scenarios as equally accurate and creative as human-generated scenarios.
DOI:10.1109/ICDH67620.2025.00012