Nosocomial infections caused by methicillin-resistant Staphylococcus aureus (MRSA) are a significant epidemiological problem. Detecting the sources of epidemic strains and preventing their access to patients, however, depend upon the availability of techniques to reliably distinguish among MRSA strains. We evaluated restriction enzyme analysis of plasmid DNA for use as an epidemiological marker of MRSA strains. The diversity of plasmid types was assessed by examining 120 clinical and environmental MRSA isolates from five southern California hospitals and from the American Type Culture Collection. Thirty-seven distinctive EcoRI digestion patterns were observed. We characterized each strain by the number of plasmids it contained and the sizes of the fragments that were generated by EcoRI. Very few of the isolates (4.2%) lacked plasmids, and some (6.7%) contained DNA that was not digested by EcoRI. Several isolates (12.5%) contained two or more plasmids. We were able to assess the stability of MRSA plasmid types by tracking epidemic strains over a 2-year period. We also examined successive isolates from 10 individual patients during their hospitalization. In all but one case, the patient's plasmid profiles remained unchanged. We conclude that the diversity and stability of MRSA plasmid types make them excellent epidemiological markers. In support of this conclusion, we found that our data provided significant epidemiological insights. Two epidemic strains, accounting for more than half of the infections, were identified in the five hospitals. The remaining cases were sporadic, caused by MRSA strains that appeared very infrequently and that may have originated from sources outside the hospitals.