Usefulness of a healthcare database for epidemiological research in atrial fibrillation

Background Big data are gaining attention as a valuable resource for providing insights into a range of issues and questions in healthcare. We evaluated the capacity of a Japanese healthcare database to conduct epidemiological research in non-valvular atrial fibrillation (NVAF). Methods and results We examined data collected between April 2008 and September 2013 in a Japanese healthcare database. Prior to the risk factor analysis, we validated the criteria for defining the occurrence of a stroke, systemic embolic event (SEE), and intracranial bleeding event during the study period. The validity was considered appropriate based on the resulting high positive predictive values. The data of 18,998 NVAF patients demonstrated that the incidence rates of stroke, SEE, and any bleeding events were 2.2, 0.08, and 2.4 per 100 patient-years, respectively. In patients who had not been treated with an anticoagulant, incidence of stroke significantly increased in higher CHADS₂ or CHA₂DS₂-VASc score, 1.7 and 1.5 fold by 1 point increase, respectively. The use of a proton pump inhibitor (PPI) was also identified as an independent risk factor for stroke. In patients who had been treated with an anticoagulant, the independent risk factors for any bleeding events were hypertension, renal dysfunction, hepatic failure, medical history of stroke, older age (≥65 years), use of nonsteroidal anti-inflammatory drug, and PPIs. Conclusion The data obtained in this study were comparable with results obtained in prospective cohort studies conducted in Japan.