A Method to Detect Errors in Electronic Discharge Summaries Based on Named Entity Recognition

Abstract

The hospital discharge summaryis an essential document, containing clinical and administrative information necessary for the continuity of care after the patients are discharged from hospital. The utilization of electronic discharge summaries has grown in popularity. However, many transcription errors and spelling mistakes exit, potentially reducing the medical quality of patient care. To solve this problem, this paper presents a novel approach to detect these errors automatically by using Named entity recognition (NER). The NER model was trained by 450 discharge summaries and rich features set was used to improve the recall and precision. Experiment on the independent test set validated the good performance of NER. The follow-up error detection using the trained NER discovered that the mistakes and ambiguous information that frequently occurred in discharge summaries.