We analyzed data from a prospective observational cohort study that included 108 adult intensive care units (ICUs) in 41 United States hospitals. Use of vancomycin (defined daily doses per 1,000 patient-days), nosocomial infection rates, and proportion of all Staphylococcus aureus isolates resistant to methicillin (MRSA rate) were recorded from January 1996 through November 1997. The median rate of vancomycin use was lowest in coronary care ICUs and highest in general Surgical ICUs. Prior approval before use of vancomycin was required in only 26 (24%)of the 108 ICUs. In a multivariate linear regression model, rates of MRSA, central line-associated bloodstream infection, and the type of ICU were independent predictors of vancomycin use. None of the vancomycin control practices was associated with lower rates of vancomycin use; however, it is important to recognize that this database was not designed to measure rates of inappropriate use. Vancomycin use is heavily determined by rates of endemic MRSA and central line-associated bloodstream infection. Efforts to reduce these rates through infection control activities should be included in hospitals' efforts to reduce vancomycin use.