Compared with usual practice, clinical trials often exclude patients with relative contraindications. A study of real-world warfarin use could help inform trials of new medications that could potentially replace warfarin. The objective of this study was to describe potential barriers to warfarin use among patients with atrial fibrillation. This was a retrospective study of electronic medical records (1998-2007) from an inner-city public hospital and affiliated primary care clinics and included adults aged 18 years or more with atrial fibrillation. Exclusions included mitral or aortic valve replacement, hyperthyroidism, or no clinical encounter within 1 year after first diagnosis. Warfarin exposure was defined by electronic pharmacy or physician order data or, in a second definition, international normalized ratio > 1.3. A history of potential barriers to warfarin was defined by International Classification of Diagnoses, 9th revision codes or electronic medical record "dictionary" terms. Among 3329 patients, CHADS2 scores were 0 (17%), 1 (26%), 2-6 (57%). Among 1276 patients with CHADS2 scores >0 who were prescribed warfarin, rates of potential barriers to warfarin were gastrointestinal or genitourinary hemorrhage (20%), alcohol abuse (16%), renal insufficiency (15%), predisposition to falls (8%), cirrhosis/hepatitis (5%), intracranial hemorrhage (1%), other hemorrhage (6%), and age 75 years or more (23%). Among 1475 patients with CHADS2 scores >0 who were not prescribed warfarin, these rates differed by not >3% except for predisposition to falls (16%) and age 75 years or more (43%). In real-world practice, many patients given warfarin have contraindications that would exclude them from clinical trials, and many patients apparently eligible for warfarin do not receive it.