Extracting Social Determinants of Health From Electronic Health Records: Development and Comparison of Rule-Based and Large Language Model Methods

1

Social determinants of health (SDoH) significantly influence health outcomes, accounting for 30%-55% of health disparities.
2

Electronic health records (EHRs) often underdocument SDoH, which are typically found in unstructured clinical notes.
3

Natural language processing (NLP) methods, including rule-based and large language models (LLMs), are used to extract SDoH from clinical text.
4

This study develops a fine-grained classification system for 7 SDoH domains, emphasizing less-explored factors like social resources and health insurance.
5

The research evaluates NLP approaches for SDoH extraction, combining rule-based prescreening with LLM classification to enhance performance.

1