SynthEHR-Eviction: Enhancing Eviction SDoH Detection with LLM-Augmented Synthetic EHR Data

TL;DR

This paper presents SynthEHR-Eviction, a scalable pipeline that uses large language models and human-in-the-loop methods to extract eviction-related social determinants of health from clinical notes, creating a large annotated dataset and improving detection accuracy.

Contribution

The paper introduces a novel pipeline combining LLMs, human annotation, and prompt optimization for eviction SDoH detection, resulting in the largest public eviction dataset and improved model performance.

Findings

Fine-tuned LLMs achieved high Macro-F1 scores of 88.8% and 90.3%.

The pipeline reduces annotation effort by over 80%.

Models outperform existing baselines like GPT-4o-APO and BioBERT.

Abstract

Eviction is a significant yet understudied social determinants of health (SDoH), linked to housing instability, unemployment, and mental health. While eviction appears in unstructured electronic health records (EHRs), it is rarely coded in structured fields, limiting downstream applications. We introduce SynthEHR-Eviction, a scalable pipeline combining LLMs, human-in-the-loop annotation, and automated prompt optimization (APO) to extract eviction statuses from clinical notes. Using this pipeline, we created the largest public eviction-related SDoH dataset to date, comprising 14 fine-grained categories. Fine-tuned LLMs (e.g., Qwen2.5, LLaMA3) trained on SynthEHR-Eviction achieved Macro-F1 scores of 88.8% (eviction) and 90.3% (other SDoH) on human validated data, outperforming GPT-4o-APO (87.8%, 87.3%), GPT-4o-mini-APO (69.1%, 78.1%), and BioBERT (60.7%, 68.3%), while enabling cost-effective deployment across various model sizes. The pipeline reduces annotation effort by over 80%, accelerates dataset creation, enables scalable eviction detection, and generalizes to other information extraction tasks.