Objective: To observe the nosocomial infection (NI) distribution in ventilated patients of a single intensive care unit (ICU) according to the kind of control of the upper airways: noninvasive positive pressure ventilation (NPPV) versus endotracheal intubation (ETI).
Setting: ICU of a general hospital.
Design: Prospective clinical and epidemiologic survey.
Patients: In the period December 1994-March 1997, 761 patients were included who needed mechanical ventilation for more than 48 h: 129 were ventilated by NPPV (NPPV group), 607 were intubated (ETI group) and 25 required intubation after a period of NPPV (NPPV-ETI group).
Measurements and results: The data used were prospectively collected according to the NI epidemiologic surveillance protocol of "C. CLIN Sud Est, Réa Sud Est", France. NI included a ventilator-associated pneumonia (VAP), catheter-related infection, urinary tract infection and bacteremia. Occurrence of NI was estimated by the density of incidence. Covariate-adjusted NI and VAP risk factors were assessed by the Cox model. The incidence density of total NI was lower for NPPV than for ETI (14.2 versus 30.3 per 1000 patient-days, p < 0.01). The Cox model showed that the use of noninvasive ventilation, adjusted to the severity of illness (SAPS II), reduced not only the VAP risk (hazard ratio (HR) = 4.07) but also the NI risk (HR = 1.95).
Conclusion: The use of NPPV reduces the risk of VAP and NI, compared to ETI, irrespective of the severity of the patient's illness.