Effective physician involvement in quality assurance in hospitals is necessary both to identify problems in patient care and to bring about problem resolution. This report describes a controlled, longitudinal study designed to measure effects on physician performance of involving physicians in setting medical audit criteria, reviewing audit results, and receiving concurrent reminders on a case-by- case basis. Unexpected low hemoglobin levels was the topic for audit. During phase 1 of the study, three randomly composed physician groups were involved in criteria setting plus review of audit results, review of audit results alone, or no treatment. During phase 2, 6 months later, the physicians in all three groups were exposed to 4 months of concurrent reminders. During phase 1, the group involved in review of audit results only achieved 51% compliance with the preset criteria compared with 26% for the group involved in criteria setting and review of audit results (P = 0.002). During phase 2, when concurrent reminders were provided, the group that was formerly the control group in phase 1, achieved 77% compliance as compared with 56% for the group previously involved in criteria setting and review of audit results (P = 0.004). These findings challenge the common assumption that physicians should be involved in criteria setting. They also suggest that providing the individual physician with information concurrently on a case-by-case basis may be superior to reviewing audit results for changing physician behavior.