Using Natural Language Processing to Improve Efficiency of Manual Chart Abstraction in Research: The Case of Breast Cancer Recurrence

摘要: The increasing availability of electronic health records (EHRs) creates opportunities for automated extraction information from clinical text. We hypothesized that natural language processing (NLP) could substantially reduce the burden manual abstraction in studies examining outcomes, like cancer recurrence, are documented unstructured text, such as progress notes, radiology reports, and pathology reports. developed an NLP-based system using open-source software to process notes 1995 2012 women with early-stage incident breast cancers identify whether when recurrences were diagnosed. evaluated 1,472 patients receiving EHR-documented care integrated Pacific Northwest. A separate study provided patient-level reference standard recurrence status date. correctly identified 92% estimated diagnosis dates within 30 days 88% these. Specificity was 96%. overlooked 5 65 recurrences, 4 because documents unavailable. other incorrectly classified nonrecurrent standard. If used similar cohorts, NLP by 90% number EHR charts abstracted confirmed cases at a rate comparable traditional abstraction.