To describe the outcome of patients after noninvasive ventilation in a high dependency unit (HDU) of an emergency department (ED). Secondary aims were to define the role of intensive care consultation and to identify variables associated with mortality. Observational, prospective 6-month study. Two hundred and nine cases were analysed. Thirty-four patients were initially rejected by the intensive care unit (ICU). Physicians in the ED did not request ICU consultation in the remaining 175 (83%) because of 'belief of improvable medical condition in the ED in patients without therapeutic limits' in 93 (group 1) and to 'preset therapeutic limits' or 'comfort measures only' in 82 (groups 2 and 3). Ten out of these 175 were subsequently admitted to the ICU. The global in-hospital mortality rate was 22% (3.3% in the high dependency unit), but only 10% in group 1. Place of referral for ventilation (P<0.001), absence of subsequent ventilation on the general ward (P<0.001), group of assignation (P=0.004), intensive care initial rejection (P=0.022), no previous home ventilation (P=0.028), older age (P=0.03) and longer duration on ventilation (P=0.047) were significantly associated with mortality. In the multivariate regression model, ventilating patients from general wards (odds ratio=7.1; 2.3-25, 95% confidence interval) and ventilation under preset limits (odds ratio=3.57; 1.42-8.98, 95% confidence interval) remained significantly associated with mortality. Noninvasive ventilation is a relatively safe and effective treatment in the ED when performed in carefully controlled settings. ICU consultation may be securely deferred in this setting.