Epidemiology of subtherapeutic anticoagulation in the United States.

BACKGROUND Low international normalized ratio (INR; <or=1.5) increases risk for thromboembolism. However, little is known about the epidemiology of low INR. METHODS AND RESULTS We prospectively collected data from 47 community-based clinics located throughout the United States from 2000 to 2002. We examined risk factors for low INR (<or=1.5), reasons given in the medical record for low INR, and proportion of thromboembolic events that occurred during periods of low INR. Of the 4489 patients in our database, 1540 (34%) had at least 1 low INR. Compared with men, women had an increased incidence of low INR (adjusted incidence rate ratio, 1.44; P<0.001). Compared with patients anticoagulated for atrial fibrillation, patients anticoagulated for venous thromboembolism had an increased incidence of low INR (adjusted incidence rate ratio, 1.48; P<0.001). The 5 most common reasons for low INR were nonadherence (17%), interruptions for procedures (16%), recent dose reductions (15%), no reason apparent after questioning (15%), and second or greater consecutive low INR (13%). A total of 21.8% of thromboembolic events (95% CI, 12.2 to 35.4%) occurred during periods of low INR; 58% of these events were related to an interruption of warfarin therapy. CONCLUSIONS In this cohort of patients receiving warfarin, more than 1 in 5 thromboembolic events occurred during a period of low INR. Women and patients anticoagulated for venous thromboembolism were particularly likely to experience low INR. Improving adherence, minimizing interruptions of therapy, and addressing low INR more promptly could reduce the risk of low INR.