Comparison of 12 risk stratification schemes to predict stroke in patients with nonvalvular atrial fibrillation

Background and Purpose-More than a dozen schemes for stratifying stroke risk in patients with nonvalvular atrial fibrillation have been published. Differences among these schemes lead to inconsistent stroke risk estimates for many atrial fibrillation patients, resulting in confusion among clinicians and nonuniform use of anticoagulation. Methods-Twelve published schemes stratifying stroke risk in patients with nonvalvular atrial fibrillation are analyzed, and observed stroke rates in independent test cohorts are compared with predicted risk status. Results-Seven schemes were based directly on event-rate analyses, whereas 5 resulted from expert consensus. Four considered only clinical features, whereas 7 schemes included echocardiographic variables. The number of variables per scheme ranged from 4 to 8 (median, 6). The most frequently included features were previous stroke/TIA (100% of schemes), patient age (83%), hypertension (83%), and diabetes (83%), and 8 additional variables were included in >= 1 schemes. Based on published test cohorts, all 8 tested schemes stratified stroke risk, but the absolute stroke rates varied widely. Observed rates for those categorized as low risk ranged from 0% to 2.3% per year and those categorized as high risk ranged from 2.5% to 7.9% per year. When applied to the same cohorts, the fractions of patients categorized by the different schemes as low risk varied from 9% to 49% and those categorized by the different schemes as high-risk varied from 11% to 77%. Conclusions-There are substantial, clinically relevant differences among published schemes designed to stratify stroke risk in patients with atrial fibrillation. Additional research to identify an optimum scheme for primary prevention and subsequent standardization of recommendations may lead to more uniform selection of patients for anticoagulant prophylaxsis.