Evaluation of a natural language processing approach to identify diagnostic errors and analysis of safety learning system case review data: retrospective cohort study.

Artificial intelligence (AI) presents a wide range of opportunities to potentially improve patient safety. This study investigated the use of machine learning (ML) and natural language processing (NLP) to improve diagnostic safety surveillance in hospitals. The study compared diagnostic errors identified by case reviewers (7.4% of the 1,704 patients included in the study) against 3 NLP models to identify which model performed best in predicting diagnostic errors. The researchers concluded that NLP can be a useful tool to efficiently identify diagnostic error cases, reducing the burden of case review and improving diagnostic safety in hospitals.