Natural language processing with deep learning for medical adverse event detection from free-text medical narratives: A case study of detecting total hip replacement dislocation

Accurate and timely detection of medical adverse events (AEs) from free-text narratives can be challenging. Natural language processing (NLP) with deep learning has already shown great potential for analyzing data, but its application AE been limited. In this study, we developed based NLP (DL-NLP) models efficient accurate hip dislocation following primary total replacement standard (radiology notes) non-standard (follow-up telephone narratives. We benchmarked these proposed traditional machine (ML-NLP) models, also assessed the accuracy International Classification Diseases (ICD) Current Procedural Terminology (CPT) codes in capturing AEs a multi-center orthopaedic registry. All DL-NLP outperformed all ML-NLP convolutional neural network (CNN) model achieving best overall performance (Kappa = 0.97 radiology notes, Kappa 1.00 follow-up notes). On other hand, ICD/CPT patients who sustained were only 75.24% accurate. demonstrated that used largescale registries AEs. The study was data most frequently electronic record (EMR) system U.S., Epic. This could potentially implemented Epic-based EMR systems to improve detection, consequently, quality care patient outcomes.