Integrating Large Language Models Into Trauma Education for Medical Students: Randomized Controlled Pilot Trial.

Area of Science:

• Medical Education
• Artificial Intelligence in Healthcare
• Clinical Simulation

Background:

• Medical knowledge is rapidly expanding, creating challenges for applying information in clinical settings.
• Large language models (LLMs) are increasingly used for information retrieval, but their educational impact in high-pressure clinical environments is unclear.

Purpose of the Study:

• To evaluate if unstructured access to a large language model (LLM) improves medical students' decision-making, teamwork, and confidence during trauma education.
• To assess the impact of LLM assistance on knowledge retention and performance in simulated trauma scenarios.

Main Methods:

• A randomized controlled pilot study involving 41 final-year medical students in trauma simulation.
• Teams were randomized to either LLM-assisted (ChatGPT-4o mini) or control groups for 18 video-based trauma scenarios.
• Outcomes included decision accuracy, response times, teamwork ratings, confidence, and knowledge retention.

Main Results:

• Confidence in trauma management increased in both groups, with greater gains in the non-LLM group.
• LLM use did not improve decision accuracy or speed and was linked to longer response times in complex cases.
• Non-LLM teams showed better teamwork and discussion; knowledge retention was similar across groups.

Conclusions:

• Unstructured LLM use in trauma education did not enhance student performance and may have hindered group reasoning.
• Non-English prompting potentially reduced LLM effectiveness, emphasizing the need for language alignment.
• Structured AI integration and AI literacy training are crucial for optimizing LLMs in medical education.