The purpose of this study was to evaluate performance feedback delivered by on-site presentations compared to mailed feedback on improving acute myocardial infarction (AMI) care. We used a randomized trial including 18 hospitals nested within the Cooperative Cardiovascular Project. Patients comprised AMI Medicare patients admitted before (n = 929, 1994 and 1995) and after intervention (n = 438, 1996). Control hospitals received written feedback by mail. The experimental intervention group received a presentation led by a cardiologist and a quality improvement specialist. We assessed the proportion of patients receiving appropriate AMI care before and after the intervention. Both univariate and multivariate analyses demonstrated no effect of the intervention in increasing the proportion of patients who received reperfusion, aspirin, beta-blockers, or angiotensin-converting enzyme inhibitors. On-site feedback presentations were not associated with a larger improvement in AMI care compared to the mailed feedback. Other interventions, such as opinion leaders and patient-directed interventions, may be necessary in order to improve the care of AMI patients.