Our objectives were to determine the incidence of endocarditis in patients whose Staphylococcus aureus bacteremia was community-acquired, related to hemodialysis, or hospital-acquired; to assess clinical factors that would reliably distinguished between S. aureus bacteremia and S. aureus endocarditis; to assess the emergence of methicillin-resistant S. aureus (MRSA) as a cause of endocarditis; and to examine risk factors for mortality in patients with S. aureus endocarditis. We conducted a prospective observational study in 6 university teaching hospitals; we evaluated 505 consecutive patients with Staphylococcus aureus bacteremia. Thirteen percent of patients with S. aureus bacteremia were found to have endocarditis, including 21% with community-acquired S. aureus bacteremia, 5% with hospital-acquired bacteremia, and 12% on hemodialysis. Infection was due to MRSA in 31%. Factors predictive of endocarditis included underlying valvular heart disease, history of prior endocarditis, intravenous drug use, community acquisition of bacteremia, and an unrecognized source. Twelve patients with bacteremia had a prosthetic valve; 17% developed endocarditis. Unexpectedly, nonwhite race proved to be an independent risk factor for endocarditis by both univariate and multivariate analyses. Persistent bacteremia (positive blood cultures at day 3 of appropriate therapy) was identified as an independent risk factor for both endocarditis and mortality, a unique observation not reported in other prospective studies of S. aureus bacteremia. Patients with endocarditis due to MRSA were significantly more likely to have complicating renal insufficiency and to experience persistent bacteremia than those with endocarditis due to MSSA. The 30-day mortality was 31% among patients with endocarditis compared to 21% in patients who had bacteremia without endocarditis (p = 0.055). Risk factors for death due to endocarditis included severity of illness at onset of bacteremia (as measured by Apache III and Pitt bacteremia score), MRSA infection, and presence of atrioventricular block on electrocardiogram. Patients with S. aureus bacteremia who have community acquisition of infection, underlying valvular heart disease, intravenous drug use, unknown portal of entry, history of prior endocarditis, and possibly, nonwhite race should undergo echocardiography to screen for the presence of endocarditis. We recommend that blood cultures be repeated 3 days following initiation of antistaphylococcal antibiotic therapy in all patients with S. aureus bacteremia. Positive blood cultures at 3 days may prove to be a useful marker in promoting more aggressive management, including more potent antibiotic therapy and surgical resection of the valve in endocarditis cases. MRSA as the infecting organism should be added to the list of risk factors for consideration of valvular resection in cases of endocarditis.