A cluster of methicillin-resistant Staphylococcus aureus (MRSA) infections among patients on an intensive care unit (ICU) was detected by routine infection control surveillance. In the period from 5 January to 22 June 1995, 10 patients on the ICU and a further 6 patients (5 on one ward that had received colonized patients transferred from the ICU) were affected by MRSA strains with the same antibiotic susceptibility patterns. Seven (44%) of these 16 colonized patients developed MRSA bacteremia. MRSA isolates with the same characteristics were also found on the hands of one member of the ICU staff. The isolates were untypeable by phage typing, but 15 of 17 outbreak strains analyzed genetically had identical randomly amplified polymorphic DNA (RAPD) and pulsed-field gel electrophoresis (PFGE) profiles. A single strain of MRSA that was nontypeable by phage typing and that was isolated on the ICU on 1 January and six nontypeable and epidemiologically unrelated MRSA isolates all had RAPD profiles distinct from that of the outbreak strain. Implementation of strict infection control measures stopped the further spread of MRSA on the ICU, the affected general ward, and seven other wards that received MRSA carriers from the ICU. Although nontypeable by phage typing and not previously recognized as an epidemic strain, this strain of MRSA was readily transmissible and highly virulent. RAPD typing was found to be a simple, rapid, and effective method for the epidemiological investigation of this outbreak, and performance of typing by this method was simpler and less time-consuming than that of typing by PFGE. RAPD typing may have more general application for the study of S. aureus infections in hospitals.