Background. Routine culturing of patients in intensive care units (ICUs) for methicillin-resistant Staphylococcus aureus (MRSA) identifies unrecognized carriers and facilitates timely isolation. However, the benefit of surveillance in detecting prevalent and incident carriers likely varies among ICUs. In addition, many assessments underestimate the incidence of acquisition by including prevalent carriers in the at-risk population. Methods. We performed a retrospective cohort study using accurate at-risk populations to evaluate the range of benefit of admission and weekly surveillance cultures in detecting otherwise unrecognized MRSA in 12 ICUs in 5 states. Results. We assessed 142 ICU-months. Among the 12 ICUs, the admission prevalence of imported MRSA was 5%-21%, with admission surveillance providing 30%-135% increases in rates of detection. The monthly hospital-associated incidence was 2%-6%, with weekly surveillance providing 7%-157% increases in detection. The common practice of reporting incidence using the total number of patients or total patient-days underestimated incidence by one-third. Surgical ICUs had lower MRSA importation but higher MRSA incidence. Overall, routine surveillance prevented the misclassification of 17% (unit range, 11%-29%) of "incident" carriers, compared with clinical cultures, and increased precaution days by 18% (unit range, 11%-91%). Conclusions. Routine surveillance significantly increases the detection of MRSA, but this benefit is not uniform across ICUs, even with high compliance and the use of correct denominators.