This study examined the prognostic significance of atrial fibrillation (AF) in a community-based cohort. AF, the most common cardiac dysrhythmia, frequently occurs in the presence of concomitant medical illness. Population-based studies have associated AF with excess mortality, and this risk of death is independent of concomitant cardiovascular disease. The effect of noncardiovascular medical illnesses on mortality in patients who have AF has not been determined. We examined a community-based cohort of 390 residents of Olmsted County, Minnesota, who had newly diagnosed AF; patients were identified retrospectively but followed prospectively. The cohort included all patients who had electrocardiographically proved first-onset AF; patients were not excluded on the basis of medical illness. Mean age of the cohort was 73 +/- 14 years (56% were men). Mean follow-up was 2.7 +/- 1.7 years. Onset of AF frequently occurred during hospitalization (78%). One hundred sixty-six deaths occurred, a death rate significantly higher than expected for the cohort. Most deaths had a noncardiovascular cause. This trend was maintained for patients who had no previous cardiovascular disease and for those who had a cardiovascular diagnosis at the time AF was diagnosed. AF is observed frequently among hospitalized patients who are medically ill. The survival rate of these patients is low, but AF may be only a minor component of the excess mortality.