An Automated Approach for Clinical Quantitative Information Extraction from Chinese Electronic Medical Records

Clinical quantitative information commonly exists in electronic medical records (EMRs) and is essential for recording patients' lab test or other characteristics in clinical notes. This study proposes an automated approach for extracting quantitative information from Chinese free-text EMR data including admission records, progress notes and ward-inspection records. The approach leverages pattern-learning combining with rule-based strategy to identify and extract clinical quantitative expressions. The experiments are based on 1,359 de-identified EMRs from the burn department of a domestic Grade-A Class-three hospital. The evaluation results present that our approach achieves a precision of 96.1%, a recall of 90.9%, and an F1-measure of 92.9%, demonstrating its effectiveness in clinical quantitative information extraction from EMR text.