Large Language Model Automated Extraction of Clinical Signs and Symptoms From Emergency Department Reports for Machine Learning Prediction Models: Development and Validation Study

1

Automated feature extraction using large language models (LLMs) can improve data usability from emergency department (ED) reports.
2

This study evaluates a small multilingual LLM's ability to extract clinical features from Dutch ED reports for acute abdominal pain.
3

The research focuses on comparing LLM-extracted features with physician annotations to assess predictive model performance.
4

Two 0-shot prompting strategies were developed to optimize the LLM for extracting clinically relevant features.
5

The study aims to establish a scalable, privacy-preserving workflow for decision support systems in the ED.

Jmir Medical Informatics

by Anoeska Schipper, Peter Belgers, Rory David O'Connor, Lieke van de Wouw, Luc Builtjes, Joeran S Bosma, Ron Kusters, Steef Kurstjens, Matthieu Rutten, Bram van Ginneken
April 30, 2026

1