BACKGROUND The admission of a proportion of disabled people to hostels is inevitably followed by their transfer to nursing homes. Our hypothesis was that such admissions are justified in terms of quality of life and the cost to the community, notwithstanding the necessity of subsequent transfer. AIMS To test this hypothesis by measuring the retention and survival times of residents in hostel and in nursing home; to consider the relevance of these factors to the future policy of the two institutions. METHODS A retrospective study was made of 159 residents admitted over a period of 12 years to a hostel with 32 places. Times spent in the hostel and in the nursing home were recorded. Probabilities of survival in hostel and in nursing home were calculated according to the Kaplan-Meier method. Comparison with the expected survival of a matched cohort of the total population was determined. Estimation was made, using the SAS software package, of the likely number of places needed in nursing homes for residents following transfer. RESULTS Although the majority of hostel residents eventually needed nursing home care, a worthwhile proportion of their total institutional time (approximately two-thirds) was spent in the hostel. Ongoing support from the personnel in a geriatric service is likely to increase retention time in the hostel. Because of the ultimate outcome for the majority of residents, planning for hostel care should include consideration of places needed in nursing homes.