BACKGROUND: Patient-specific risk factors for acquisition of vancomycin-resistant enterococci (VRE) among hospitalized patients are becoming well defined. However, few studies have reported data on the institutional risk factors, including rates of antimicrobial use, that predict rates of VRE. Identifying modifiable institutional factors can advance quality-improvement efforts to minimize hospital-acquired infections with VRE. OBJECTIVE: To determine the independent importance of any association between antimicrobial use and risk factors for nosocomial infection on rates of VRE in intensive care units (ICUs). DESIGN: Prospective ecologic study. SETTING: 126 adult ICUs from 60 U.S. hospitals from January 1996 through July 1999. PATIENTS: All patients admitted to participating ICUs. MEASUREMENTS: Monthly use of antimicrobial agents (defined daily doses per 1000 patient-days), nosocomial infection rates, and susceptibilities of all tested enterococci isolated from clinical cultures. RESULTS: Prevalence of VRE (median, 10%; range, 0% to 59%) varied by type of ICU and by teaching status and size of the hospital. Prevalence of VRE was strongly associated with VRE prevalence among inpatient non-ICU areas and outpatient areas in the hospital, ventilator-days per 1000 patient-days, and rate of parenteral vancomycin use. In a weighted linear regression model controlling for type of ICU and rates of VRE among non-ICU inpatient areas, rates of vancomycin use (P < 0.001) and third-generation cephalosporin use (P = 0.02) were independently associated with VRE prevalence. CONCLUSIONS: Higher rates of vancomycin or third-generation cephalosporin use were associated with increased prevalence of VRE, independent of other ICU characteristics and the endemic VRE prevalence elsewhere in the hospital. Decreasing the use rates of these antimicrobial agents could reduce rates of VRE in ICUs.