The long-term efficacy of a program to control methicillin-resistant Staphylococcus aureus (MRSA) was evaluated in a 350-bed university hospital. Three periods were monitored: pre-epidemic (January 1989-November 1989), outbreak (December 1989-June 1990) and control program (July 1990-December 1992) periods. Control measures included cohort isolation, patient care measures and therapy (oral cotrimoxazole plus fusidic acid ointment) of MRSA carriage in patients, roommates and personnel. A total of 117 MRSA-infected patients were detected. For each period respectively, MRSA incidence (number of cases per 1,000 patient-days) was 3.2, 8.2 and 2.0 in the intensive care unit (ICU) and 0.08, 0.23 and 0.26 in the general wards. During the outbreak there was a 2.7-fold overall increase of baseline MRSA incidence (p < 0.02). The crude mortality was 68% and the attributable mortality was estimated to be 50%. The program was estimated to have prevented 76% (CI95 28-91, p < 0.0001) of expected MRSA cases and 85% (CI95 62-94, p < 0.0001) of expected fatalities due to MRSA in the ICU, but it had no significant effect in the general wards. The program did not control vancomycin consumption.