OBJECTIVE: To determine the influence of using different denominators on risk estimates of ventilator-associated pneumonia (VAP). DESIGN AND SETTING: Prospective cohort study in the medical ICU of a large teaching hospital. PATIENTS: All consecutive patients admitted for more than 48 h between October 1995 and November 1997. MEASUREMENTS AND RESULTS: We recorded all ICU-acquired infections using modified CDC criteria. VAP rates were reported per 1,000 patient-days, patient-days at risk, ventilator-days, and ventilator-days at risk. Of the 1,068 patients admitted, VAP developed in 106 (23.5%) of those mechanically ventilated. The incidence of the first episode of VAP was 22.8 per 1,000 patient-days (95% CI 18.7-27.6), 29.6 per 1,000 patient-days at risk (24.2-35.8), 35.7 per 1,000 ventilator-days (29.2-43.2), and 44.0 per 1,000 ventilator-days at risk (36.0-53.2). When considering all episodes of VAP (n=127), infection rates were 27.3 episodes per 1,000 ICU patient-days (95% CI 22.6-32.1) and 42.8 episodes per 1,000 ventilator-days (35.3-50.2). CONCLUSIONS: The method of reporting VAP rates has a significant impact on risk estimates. Accordingly, clinicians and hospital management in charge of patient-care policies should be aware of how to read and compare nosocomial infection rates.