Comparison of Clinical Risk Stratification for Predicting Stroke and Thromboembolism in Atrial Fibrillation

Background and Purpose Several accepted algorithms exist to characterize the risk of thromboembolism in atrial fibrillation. We performed a comparative analysis to assess the predictive value of 9 such schemes. Methods In a longitudinal community-based cohort study from Olmsted County, Minnesota, 2720 residents with atrial fibrillation were followed up for 4.43.6 yearsSD from 1990 to 2004. Risk factors were identified using a diagnostic index integrated with the electronic medical record. Thromboembolism and cardiovascular event data were collected and analyzed. Results We identified 350 validated thromboembolic events in our cohort. Multivariable analysis identified age >75 years (odds ratio, 2.08; P<0.0001), female sex (odds ratio, 1.45; P=0.0015), history of hypertension (odds ratio, 3.07; P<0.0001), diabetes mellitus (odds ratio, 1.58; P=0.0003), and history of heart failure (odds ratio, 1.50; P=0.0102) as significant predictors of clinical thromboembolism. The Stroke Prevention in Atrial Fibrillation (SPAF; hazard ratio, 2.75; c=0.659), CHADS(2)-revised (hazard ratio, 3.48; c=0.654), and CHADS(2)-classical (hazard ratio, 2.90; c=0.653) risk schemes were most accurate in risk stratification. The low-risk cohort within the CHA(2)DS(2)-VASc scheme had the lowest event rate among all low-risk cohorts (0.11 per 100 person-years). Conclusions A direct comparison of 9 risk schemes reveals no profound differences in risk stratification accuracy for high-risk patients. Accurate prediction of low-risk patients is perhaps more valuable in determining those unlikely to benefit from oral anticoagulation therapy. Among our cohort, CHA(2)DS(2)-VASc performed best in this purpose.