Development of a large language model clinical variable extraction pipeline for the study of lymphedema.
8 affiliations
- 1Stanford Division of Vascular Surgery, Department of Surgery, Stanford University, Stanford, CA.
- 2Stanford Center for Artificial Intelligence in Medicine and Imaging, Stanford University, Stanford, CA.
- 3Stanford Department of Radiology, Stanford University, Stanford, CA.
- 4Technical University of Munich, Munich, DE.
- 5Stanford Department of Biomedical Data Science, Stanford University, Stanford, CA.
- 6Stanford Cardiovascular Institute, Stanford, CA.
- 7Weill Cancer Hub West, San Francisco, CA.
- 8VA Palo Alto Healthcare System, Palo Alto, CA.
Abstract
For many disease processes, clinical information is often captured in unstructured clinical notes that require manual data extraction for research. Large language models (LLMs) have the potential to automate the data extraction process, enabling study for diseases with fewer tabular metrics available. Here, we report the development and validation of an LLM-based clinical note data extraction approach for a retrospective cohort study of lymphedema etiologies and treatments. After using International Classification of Diseases, 10th Revision codes to identify potential lymphedema patients and text-matching to find potentially relevant clinical notes, an LLM was used to further filter notes, summarize for relevant content, and extract per-patient variables concerning etiology, treatments and efficacies, clinical stage, and wound status. Prompts were optimized for accuracy with an iterative review strategy, starting with zero-shot approaches and modified to few-shot approaches where necessary, validated by subset human review. A Protected Health Information-compliant Application Programming Interface version of OpenAI's GPT-5 was used for all LLM queries. Human review of LLM-extracted final variables revealed high accuracy when compared against both an intermediate summarization step (99% accurate) and against original clinical notes (98% accurate) across 2858 patients/32,288 notes. LLMs are a valuable tool for conducting clinical research at scale using unstructured text data.
Related
- DICD-IE: an expert-annotated benchmark for evaluating pretrained biomedical language models in dental implant literature and clinical knowledge extraction
- Large Language Model Framework for Structured Biomarker Extraction and Triple-Negative Breast Cancer Identification From Pathology Reports: Development and Evaluation Study.
- Large Language Models for Structured Information Extraction from German Histopathology Reports - Hepatocellular Carcinoma
- Variable clinical course of retinitis pigmentosa patients with RPE65 gene mutations
- Common and variable clinical, histological, and imaging findings of recessive RYR1-related centronuclear myopathy patients
BibTeX
@article{fisher2026development,
title = {{Development of a large language model clinical variable extraction pipeline for the study of lymphedema.}},
author = {Andrea T Fisher and Anoushka Lakshmi and Benjamin C Liu and Jenny Woo and Richard Gaus and Akshay S Chaudhari and Shipra Arya and Katherine Hekman},
journal = {JVS-vascular science},
year = {2026},
month = sep,
doi = {10.1016/j.jvssci.2026.100448},
url = {https://europepmc.org/article/MED/42921181},
note = {CC BY},
}
RIS
TY - JOUR
TI - Development of a large language model clinical variable extraction pipeline for the study of lymphedema.
AU - Andrea T Fisher
AU - Anoushka Lakshmi
AU - Benjamin C Liu
AU - Jenny Woo
AU - Richard Gaus
AU - Akshay S Chaudhari
AU - Shipra Arya
AU - Katherine Hekman
PY - 2026
DA - 2026/09/21
JO - JVS-vascular science
DO - 10.1016/j.jvssci.2026.100448
UR - https://europepmc.org/article/MED/42921181
AB - For many disease processes, clinical information is often captured in unstructured clinical notes that require manual data extraction for research. Large language models (LLMs) have the potential to automate the data extraction process, enabling study for diseases with fewer tabular metrics available. Here, we report the development and validation of an LLM-based clinical note data extraction approach for a retrospective cohort study of lymphedema etiologies and treatments. After using International Classification of Diseases, 10th Revision codes to identify potential lymphedema patients and text-matching to find potentially relevant clinical notes, an LLM was used to further filter notes, summarize for relevant content, and extract per-patient variables concerning etiology, treatments and efficacies, clinical stage, and wound status. Prompts were optimized for accuracy with an iterative review strategy, starting with zero-shot approaches and modified to few-shot approaches where necessary, validated by subset human review. A Protected Health Information-compliant Application Programming Interface version of OpenAI's GPT-5 was used for all LLM queries. Human review of LLM-extracted final variables revealed high accuracy when compared against both an intermediate summarization step (99% accurate) and against original clinical notes (98% accurate) across 2858 patients/32,288 notes. LLMs are a valuable tool for conducting clinical research at scale using unstructured text data.
SN - 2666-3503
ER -