SynthEHR-Eviction: Enhancing Eviction SDoH Detection with LLM-Augmented Synthetic EHR Data
This work addresses the understudied problem of eviction detection in healthcare data for researchers and practitioners, enabling scalable applications with reduced annotation effort, though it is incremental as it builds on existing LLM and synthetic data techniques.
The paper tackled the problem of detecting eviction status from unstructured electronic health records, which is rarely coded in structured fields, by creating SynthEHR-Eviction, a pipeline combining LLMs, human-in-the-loop annotation, and automated prompt optimization to generate synthetic data. The result was fine-tuned LLMs achieving Macro-F1 scores of 88.8% for eviction detection and 90.3% for other social determinants of health, outperforming existing methods like GPT-4o-APO and BioBERT.
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.