We describe our initial experience with the admission of 129 patients for 132 episodes of ventilator-dependence to a self-contained ventilator-dependent unit (VDU) in a general hospital and present a survival comparison between VDU patients and a historic control population from the same institution. Forty-three patients were screened and denied admission to the VDU because long-term ventilator dependence was not felt to be a probable outcome (56%); they were medically unstable, often requiring electrocardiographic monitoring (19%), they had poor rehabilitation potential because of markedly depressed mental status (13%), or they preferred to be treated closer to their homes (12%). Thirteen (9.8%) of the VDU patients died in the hospital compared to 44 (42%) in the historic control group. After exclusion of patients with multiorgan failure (who made up 26% of the control group) and using a proportional hazard model to adjust for group differences in age and disease class, the difference in hospital mortality remained highly significant (p < or = 0.01). Ninety-one of the 119 VDU patients (77%) were ultimately able to return home; 16 (13%) continued to use a ventilator intermittently at night; 26 patients (22%) were permanently placed in nursing homes, all off of the ventilator. Overall, 88% of the 119 patients discharged had been liberated from mechanical ventilation. Ninety-seven (82%) and 86 (72%) remain alive 1 and 2 years after discharge, respectively. Some of the survival benefits may be directly attributed to the VDU. Others reflect a change in treatment philosophy, which was nevertheless reinforced by our VDU experience.