BACKGROUND: Intravascular devices are a leading cause of nosocomial infection. Specific prevention strategies and improved guidelines for the use of intravascular devices can decrease the rate of infection; however, the impact of a combination of these strategies on rates of vascular-access infection in intensive-care units (ICUs) is not known. We implemented a multiple-approach prevention programme to decrease the occurrence of vascular-access infection in an 18-bed medical ICU at a tertiary centre. METHODS: 3154 critically ill patients, admitted between October, 1995, and November, 1997, were included in a cohort study with longitudinal assessment of an overall catheter-care policy targeted at the reduction of vascular-access infections and based on an educational campaign for vascular-access insertion and on device use and care. Incidence of ICU-acquired infections was measured by means of on-site surveillance. FINDINGS: 613 infections occurred in 353 patients (19.4 infections per 100 admissions). The incidence density of exit-site catheter infection was 9.2 episodes per 1000 patient-days before the intervention, and 3.3 episodes per 1000 patient-days afterwards (relative risk 0.36 [95% CI 0.20-0.63]). Corresponding rates for bloodstream infection were 11.3 and 3.8 episodes per 1000 patient-days, respectively (0.33 [0.20-0.56]) due to decreased rates of both microbiologically documented infections and clinical sepsis. Rates of respiratory and urinary-tract infections remained unchanged, whereas those of skin or mucous-membrane infections decreased from 11.4 to 7.0 episodes per 1000 patient-days (0.62 [0.41-0.93]). Overall, the incidence of nosocomial infections decreased from 52.4 to 34.0 episodes per 1000 patient-days (0.65 [0.54-0.78]). INTERPRETATION: A multiple-approach prevention strategy, targeted at the insertion and maintenance of vascular access, can decrease rates of vascular-access infections and can have a substantial impact on the overall incidence of ICU-acquired infections.