The aim of the project was to evaluate the short‐term treatment outcomes of patients treated in a medium secure service for people with intellectual disability. A total of 138 patients, 77 discharged and 61 current inpatients, treated over a six‐year period were included in the audit. Information on demographic and clinical variables was collected on a pre‐designed data collection tool and analysed using appropriate statistical methods. The median length of stay for the discharged group was 2.8 years. About 90% of this group were discharged to lower levels of security and about a third went directly to community placements. None of the clinical and forensic factors examined was significantly associated with length of stay for this group. There was a ‘difficult to discharge long‐stay’ group which had more patients with criminal sections, restriction orders, history of abuse, fire setting, personality disorders and substance misuse. However, when regression analysis was done, most of these factors were not predictive of the length of stay. Clinical diagnosis or offending behaviour categories are poor predictors of length of hospital stay, and there is a need to identify empirically derived patient clusters using a variety of clinical and forensic variables. Common datasets and multi‐centre audits are needed to drive this.

No abstract