CONTEXT: Warfarin has been shown to be highly efficacious for preventing thromboembolism in atrial fibrillation in randomized trials, but its effectiveness and safety in clinical practice is less clear. OBJECTIVE: To evaluate the effect of warfarin on risk of thromboembolism, hemorrhage, and death in atrial fibrillation within a usual care setting. DESIGN: Cohort study assembled between July 1, 1996, and December 31, 1997, and followed up through August 31, 1999. SETTING: Large integrated health care system in Northern California. PATIENTS: Of 13,559 adults with nonvalvular atrial fibrillation, 11,526 were studied, 43% of whom were women, mean age 71 years, with no known contraindications to anticoagulation at baseline. MAIN OUTCOMES: Ischemic stroke, peripheral embolism, hemorrhage, and death according to warfarin use and comorbidity status, as determined by automated databases, review of medical records, and state mortality files. RESULTS: Among 11,526 patients, 397 incident thromboembolic events (372 ischemic strokes, 25 peripheral embolism) occurred during 25,341 person-years of follow-up, and warfarin therapy was associated with a 51% (95% confidence interval [CI], 39%-60%) lower risk of thromboembolism compared with no warfarin therapy (either no antithrombotic therapy or aspirin) after adjusting for potential confounders and likelihood of receiving warfarin. Warfarin was effective in reducing thromboembolic risk in the presence or absence of risk factors for stroke. A nested case-control analysis estimated a 64% reduction in odds of thromboembolism with warfarin compared with no antithrombotic therapy. Warfarin was also associated with a reduced risk of all-cause mortality (adjusted hazard ratio, 0.69; 95% CI, 0.61-0.77). Intracranial hemorrhage was uncommon, but the rate was moderately higher among those taking vs those not taking warfarin (0.46 vs 0.23 per 100 person-years, respectively; P =.003, adjusted hazard ratio, 1.97; 95% CI, 1.24-3.13). However, warfarin therapy was not associated with an increased adjusted risk of nonintracranial major hemorrhage. The effects of warfarin were similar when patients with contraindications at baseline were analyzed separately or combined with those without contraindications (total cohort of 13,559). CONCLUSIONS: Warfarin is very effective for preventing ischemic stroke in patients with atrial fibrillation in clinical practice while the absolute increase in the risk of intracranial hemorrhage is small. Results of randomized trials of anticoagulation translate well into clinical care for patients with atrial fibrillation.