PrecLLM: A Privacy-Preserving Framework for Efficient Clinical Annotation Extraction from Unstructured EHRs using...

Large Language Models (LLMs) have demonstrated remarkable proficiency in automated text annotation within natural language processing. However, their deployment in clinical settings is severely constrained by strict privacy regulations and the prohibitive computational cost of processing voluminous unstructured Electronic Health Records (EHRs). Unstructured EHRs typically include crucial information for clinical decisions in precision medicine, such as in cancer. To overcome this bottleneck and enable scalable annotation extraction from clinical notes, we developed a compact LLM framework featuring a novel EHR-specific resource-efficient PREproCessing technique that can be adopted in existing LLM procedures (PrecLLM). This technique is particularly useful for the smaller LLMs which are often more accuracy-challenged. PrecLLM has been optimized for local deployment in computational environments with stringent privacy requirements and restricted access to high-performance GPUs. Two alternative simple yet powerful procedures are provided in the preprocessing step: regular expressions (regex) and Retrieval-Augmented Generation (RAG), which extract and highlight key information from unstructured clinical notes. Pre-filtering long and unstructured texts enhanced the performance of smaller LLMs on EHR-related tasks. Evaluation was performed on two distinct cohorts: a locally curated private EHR dataset from the EPIC system for a Head and Neck Cancer (HNC) cohort, and the publicly available EHR dataset (MIMIC-IV). Using MIMIC-IV, we further compared PrecLLM against fine-tuned LLMs. Results demonstrated that PrecLLM substantially enhanced the performance of the original smaller LLMs in terms of sensitivity, specificity, and F1 scores, making it well-suited for privacy-sensitive and resource-constrained applications. This study offers optimized LLM performance for local, secure, and efficient healthcare applications, and provides practical guidance for clinical LLM deployment while addressing challenges related to privacy, computational feasibility, and clinical applicability.