To investigate whether isolating patients with MRSA in private rooms in ICUs (or cohorting) is a protective factor for nosocomial MRSA infection. Association between nosocomial MRSA infection rates and ICU structure and process parameters in the German Nosocomial Infection Surveillance System (KISS). Two hundred twelve ICUs participating in KISS in 2001. In June 2001, a structured questionnaire was sent to the participating ICUs regarding their preventive measures, their type and size, their patient-to-personnel ratios, and routine cultures. Univariate and multivariate analyses were conducted to identify risk factors for nosocomial MRSA infection. The questionnaire was completed by 164 (77.4%) of the ICUs. These ICUs had 325 nosocomial MRSA infections in a 5-year period (1997 to 2001). The mean incidence density of nosocomial MRSA infections was 0.3/1,000 patient-days. Ninety-one ICUs (55.5%) did not register any nosocomial MRSA infections during the observation period. Forty-two ICUs had an incidence density of at least 0.3/1,000 patient-days (75th percentile). Surgical ICUs were found to be a risk factor for a nosocomial MRSA infection rate above this threshold. Multivariate analysis found surgical ICUs to be an independent predictor and isolation in private rooms (or cohorts) to be a protective factor (OR, 0.36; CI95, 0.17-0.79). Many (34.4%) of the German ICUs have not isolated MRSA patients in private rooms or cohorts, a procedure associated with lower MRSA infection rates in this study.