The American Journal of Cardiology, 2019. 123(11): pp. 1828-1834.
Avgil Tsadok M, Berliner Senderey A, Reges O, Leibowitz M, Leventer-Roberts M, Hoshen M, Haim M.
Several stroke risk stratification scores have been developed to guide clinical decision-making in patients with nonvalvular atrial fibrillation (AF). The aim of this study was to compare the performance of the CHADS2, CHA2DS2-VASc and R2CHADS2 risk scores to predict stroke. This retrospective cohort study was based on electronic medical records from Clalit Health Services (CHS), the largest payer provider healthcare organization in Israel. Data of CHS members with AF diagnosis between 2004 and 2015 were extracted. Demographic and co-morbidity data were used to calculate the 3 risk scores, and the performance of the scores to predict stroke were compared using area under the curve and net reclassification index. Of the 89,213 CHS members with AF, 52.3% were women and median age was 76 years. The proportions of patients at high risk were 66.2%, 86.7%, and 71.1% in the CHADS2, CHA2DS2-VASc, and R2CHADS2, respectively, with stroke incidence rates of 2.91, 2.35, and 2.80 per 100 person-years, respectively. Area under the curves for stroke prediction were 0.61 for both CHADS2 and CHA2DS2-VASc and 0.59 for R2CHADS2. Net reclassification index analysis demonstrated a net improvement of 0.089 in the index when CHA2DS2-VASc was compared with CHADS2 and a net reduction of 0.083 when R2CHADS2 was compared with CHADS2. In conclusion, current stroke stratification scores have comparable but limited ability to predict stroke in patients with AF. Stroke prevention strategies may vary depending on the applied stratification. There is a need for a better stroke risk stratification score for patients with AF.