BACKGROUND: In addition to controversies as to the definition of nosocomial pneumonia (NP) because of the lack of a widely accepted diagnostic standard, there has been no agreement concerning the time span from hospital admission to disease onset. This study aims at both estimating the time span, in hours, from admission to the occurrence of suspected NP and investigating risk factors that might influence this time span. METHODS: This is a cohort study, and subjects were patients with nosocomial infection acquired in the intensive care unit of Edgard Santos University Hospital (HUPES/ICU) in Salvador, Brazil, from January 1995 to December 1997. Patients were observed from admission to 48 hours after discharge from the intensive care unit. The time span from admission to occurrence of suspected NP, the reason for admission, patient's origin, history of surgery, general anesthesia, mechanical ventilation, and use of antibiotic were analyzed and given a multivariate analysis using Cox regression model. RESULTS: Among 246 patients with nosocomial infection, 198 (80.5%) were suspected cases of NP, whereas 48 patients (19.5%) were not classified as such. The mean time, in hours, for the NP-free time span was 85.1 +/- 3.5 hours, and the median time was 72 hours when estimated by Kaplan-Meier method. Patients admitted from surgical heart procedures who had been given general anesthesia, mechanical ventilation, and antibiotics showed statistically significant shorter mean time spans from admission to NP occurrence when compared with the other patients. Age >or=50 years, use of mechanical ventilation, and use of antibiotics were associated with NP. CONCLUSIONS: Our finding for the estimated mean NP-free time span at the HUPES/ICU is somewhat in accordance with the literature (48 to 72 hours). Patients at HUPES/ICU might be considered as showing early NP, because they were diagnosed before the 5th day after admission. Preventive measures to NP should be reviewed and intensified at the HUPES/ICU, especially as related to mechanical ventilation.