Unresolved debates about the appropriate number of beds in acute psychiatry nave mostly relied on data from population level relationships between deprivation and admission rates. We suggest a need for new analytical approaches that utilise data on patient characteristics and resource utilisation that are sensitive to the changing dynamics of admission processes. We apply statistical methods appropriate for 'time to event' utilisation outcomes to describe and predict time to discharge, i.e. duration of admission in a 6-month cohort admitted by community mental health teams in Nottingham. Kaplan-Meier estimates of median lengths of stay across sector teams ranged from 19 to 35 days reflecting different discharge rates. Significant sector differences remained after adjusting for patient level sociodemographic characteristics, routine clinical diagnoses, and other prognostic factors for length of stay in Cox regression models. These data provide evidence that sector teams differ in their patient management strategies having implications for resource management. The survival/event history analysis approach is a potentially useful method of feeding back current service activity. Future studies investigating community mental health teams should incorporate 'true' (non-proxy) outcomes such as health status, needs or social functioning, as well as service outcomes.