A prospective survey was performed over a period of 3 wk among 42 intensive care units to assess the incidence of use and effectiveness of noninvasive mechanical ventilation (NIV) in clinical practice. All patients requiring ventilatory support for acute respiratory failure (ARF), either with endotracheal intubation (ETI) or NIV, were included. Ventilatory support was required in 689 patients, 581 with ETI and 108 (16%) with NIV (35% of patients not intubated on admission). Reasons for mechanical ventilation were coma (30%), cardiogenic pulmonary edema (7%), and hypoxemic (48%) and hypercapnic ARF (15%). NIV was never used for patients in coma (who were excluded from further analysis), but was used in 14% of patients with hypoxemic ARF, in 27% of those with pulmonary edema, and in 50% of those with hypercapnic ARF. NIV was followed by ETI in 40% of cases. The incidence of both nosocomial pneumonia (10% versus 19%, p = 0.03), and mortality (22% versus 41%, p < 0.001) was lower in NIV patients than in those with ETI. After adjusting for differences at baseline, Simplified Acute Physiology Score (SAPS) II (odds ratio [OR] = 1.05 per point; confidence interval [CI]: 1.04 to 1.06), McCabe/Jackson score (OR = 2.18; CI: 1.57 to 3.03), and hypoxemic ARF (OR = 2.30; CI: 1.33 to 4.01) were identified as risk factors explaining mortality; success of NIV was associated with a lower risk of pneumonia (OR = 0.06; CI: 0.01 to 0.45) and of death (OR = 0.16; CI: 0.05 to 0.54). In NIV patients, SAPS II and a poor clinical tolerance predicted secondary ETI. Failure of NIV was associated with a longer length of stay. In conclusion, NIV can be successful in selected patients, and is associated with a lower risk of pneumonia and death than is ETI.