A recent large, international multicenter study has shown that impaired renal function is a strong predictor of stroke and systemic embolism in patients with non-valvular atrial fibrillation at moderate to high risk of stroke. The risk stratification for stroke risk in patients with non-valvular atrial fibrillation should include renal function. The study was published online December 3, 2012, in Circulation. The study was designed to assess factors associated with stroke occurrence and systemic embolism. A total of 14,264 patients with non-valvular atrial fibrillation and creatinine clearance (CrCl) ≥30 mL/min were randomized to receive rivaroxaban or adjusted-dose warfarin in the Rivaroxaban Stroke Prevention in Atrial Fibrillation (ROCKET AF) randomized double-blind study. A Cox proportional risk model was used to identify random independent correlates of the occurrence of stroke or non-central nervous system (CNS) embolism on the basis of an intention-to-treat analysis. Risk scores were derived in the ROCKET AF cohort and validated in the ATRIA cohort (a separate cohort of patients in the AF group). Results showed that the primary endpoint event occurred in 575 patients (4.0%) during a median follow-up time of 1.94 years. Reduced creatinine clearance was a strong independent predictor of stroke and systemic embolism, second only to a history of previous stroke or transient ischemic attack (TIA). Other factors associated with stroke and systemic embolism include elevated diastolic blood pressure and heart rate, and vascular disease of the heart and extremities (C-index of 0.635). A model that included creatinine clearance (R2CHADS2) was able to improve the net reclassification index (NRI) by 6.2% (compared to CHA2DS2VASc, C-statistic = 0.578), and 8.2% (compared to CHADS2, C-statistic = 0.575). A model incorporating an estimated glomerular filtration rate <60 and a history of prior stroke or transient ischemic attack (no other covariates) resulted in a C-statistic of 0.590. In an external, independent population, the effectiveness of R2CHADS2 increased NRI relative to CHADS2 by 17.4% (95% CI 12.1-22.5%).