Web-based Real-Time Case Finding for the Population Health Management of Patients With Diabetes Mellitus: A Prospective Validation of the Natural Language Processing–Based Algorithm With Statewide Electronic Medical Records

• DOI: 10.2196/medinform.6328
• Publication Date: 2016-11-11T09:00:13Z
• Authors: Le Zheng; Yue Wang; Shiying Hao; Andrew Y Shin; Bo Jin; Anh D Ngo; Medina S Jackson-...; Daniel J Feller; Tianyun Fu; Karena Zhang; Xin Zhou; Chunqing Zhu; Dorothy Dai; Yunxian Yu; Gang Zheng; Yu-Ming Li; Doff B McElhinney; Devore S Culver; Shaun T Alfreds; Frank Stearns; Karl G Sylvester; Eric Widen; Xuefeng Bruce Ling
• Keywords: Medical record; Diagnosis code; Electronic medical record; Electronic health record

Diabetes case finding based on structured medical records does not fully identify diabetic patients whose histories related to diabetes are available in the form of free text. Manual chart reviews have been used but involve high labor costs and long latency.This study developed tested a Web-based algorithm using both unstructured electronic (EMRs).This was health information exchange (HIE) EMR database that covers almost all facilities state Maine, United States. Using narrative clinical notes, natural language processing (NLP) retrospectively (July 1, 2012, June 30, 2013) with random subset HIE-associated facilities, which then blind remaining facilities. The NLP-based subsequently integrated into HIE validated prospectively 2013, 2014).Of 935,891 prospective cohort, 64,168 cases were identified diagnosis codes alone. Our found an additional 5756 uncodified (5756/64,168, 8.97% increase) positive predictive value .90. Of 21,720 by methods, 6616 (6616/21,720, 30.46%) before noted (mean time difference = 48 days).The online NLP effective identifying real time, leading significant improvement finding. successful integration Maine indicates strong potential for application this novel method achieve more complete ascertainment diagnoses mellitus.