Objective.To establish the occurrence, distribution, and secular time trend of nosocomial infections (NIs) in a surgical intensive care unit (ICU).

Design and Setting.Follow‐up study in a teaching hospital in Spain.

Methods.In May 1995 we established an nosocomial infection surveillance system in our surgical ICU. We collected information daily for all patients who were in the ICU for at least 48 hours (546 patients from 1996 through 2000). We used the Centers for Disease Control and Prevention definitions and criteria for infections. Monthly, we determined the site‐specific incidence densities of NIs, the rates of medical device use, and the Poisson probability distribution, which determined whether the case count equalled the number of expected cases (the mean number of cases during the previous year, with extreme values excluded). We compared yearly and monthly infection rates by Poisson regression, using site‐specific NIs as a dependent variable and year and month as dummy variables. We tested annual trends with an alternative Poisson regression model fitting a single linear trend.

Results.The average rate of catheter‐associated urinary tract infections was 8.4 per 1000 catheter‐days; that of ventilator‐associated pneumonia, 21 per 1000 ventilator‐days; and that of central line–associated bloodstream infections, 30 per 1000 central line–days. The rate of urinary tract infections did not change over the study period, but there was a trend toward decreases in the rates of central line–associated bloodstream infections and ventilator‐associated pneumonia.

Conclusion.An NI surveillance and control program contributed to a progressive decrease in NI rates.