Ching-Lan Cheng1, Hsu-Chih Chien2, Cheng-Han Lee3, Swu-Jane Lin4, Yea-Huei Kao Yang5. 1. School of Pharmacy and Institute of Clinical Pharmacy and Pharmaceutical Sciences, College of Medicine, National Cheng Kung University, Tainan, Taiwan; Health Outcome Research Center, National Cheng Kung University, Tainan, Taiwan. 2. School of Pharmacy and Institute of Clinical Pharmacy and Pharmaceutical Sciences, College of Medicine, National Cheng Kung University, Tainan, Taiwan. 3. Department of Internal Medicine, National Cheng Kung University Hospital, Tainan, Taiwan. 4. University of Illinois at Chicago, Chicago, IL, USA. 5. School of Pharmacy and Institute of Clinical Pharmacy and Pharmaceutical Sciences, College of Medicine, National Cheng Kung University, Tainan, Taiwan; Health Outcome Research Center, National Cheng Kung University, Tainan, Taiwan. Electronic address: yhkao@mail.ncku.edu.tw.
Abstract
OBJECTIVE: The aim of this study was to determine the validity of in-hospital mortality records in the National Health Insurance Research Database (NHIRD) by cross-comparing with death records from the electronic medical records (EMR) of a medical center in southern Taiwan. METHODS: Data on patients admitted to the medical center for acute myocardial infarction (AMI) or stroke during the years 2005 to 2010 were extracted from the two databases and cross-linkages with patients' characteristics (birth date, gender, admission date, and discharge date). While the death record was available in the catastrophic illness registry data files (CIRD), we also estimated the insurance status and death record in the CIRD subset using confirmed death cases. Additionally, agreement in comorbidities between records from the two databases was evaluated. RESULTS: A total of 6197 cases were successfully linked, with a linkage rate of 96.56% of cases in the NHIRD when linked to those from the EMR. Among the linked population, 538 of 682 patients retrieved as expired in the NHIRD were also so recorded in the EMR. This yielded a positive predictive value of 0.79 when the EMR was used as the gold standard. Patients having death records in both the CIRD subset and the EMR totaled 364, which yielded a percentage positive agreement rate of 76%. The consistency in comorbidity diagnoses between the two databases was more than 90% among matched cases. CONCLUSIONS: The accuracy of death records in the NHIRD was high, and appears to be a valid resource for population research in cardiovascular diseases.
OBJECTIVE: The aim of this study was to determine the validity of in-hospital mortality records in the National Health Insurance Research Database (NHIRD) by cross-comparing with death records from the electronic medical records (EMR) of a medical center in southern Taiwan. METHODS: Data on patients admitted to the medical center for acute myocardial infarction (AMI) or stroke during the years 2005 to 2010 were extracted from the two databases and cross-linkages with patients' characteristics (birth date, gender, admission date, and discharge date). While the death record was available in the catastrophic illness registry data files (CIRD), we also estimated the insurance status and death record in the CIRD subset using confirmed death cases. Additionally, agreement in comorbidities between records from the two databases was evaluated. RESULTS: A total of 6197 cases were successfully linked, with a linkage rate of 96.56% of cases in the NHIRD when linked to those from the EMR. Among the linked population, 538 of 682 patients retrieved as expired in the NHIRD were also so recorded in the EMR. This yielded a positive predictive value of 0.79 when the EMR was used as the gold standard. Patients having death records in both the CIRD subset and the EMR totaled 364, which yielded a percentage positive agreement rate of 76%. The consistency in comorbidity diagnoses between the two databases was more than 90% among matched cases. CONCLUSIONS: The accuracy of death records in the NHIRD was high, and appears to be a valid resource for population research in cardiovascular diseases.
Authors: Chang-Fu Kuo; Matthew J Grainge; Ana M Valdes; Lai-Chu See; Kuang-Hui Yu; S W Steven Shaw; Shue-Fen Luo; Weiya Zhang; Michael Doherty Journal: Rheumatology (Oxford) Date: 2017-06-01 Impact factor: 7.580