Despite seclusion being described as one of the most ethically- and legally-controversial management options available, it remains a widely-used clinical strategy for managing disruptive, aggressive, and violent behaviour. This study sought to determine how frequently seclusion was used, the common characteristics of those secluded and not secluded, and the degree to which the Level of Service Inventory – Revised: Screening Version (LSI-R: SV) could predict seclusion. The study was retrospective, covering the first 2 years of operation of a statewide forensic psychiatry hospital in Victoria, Australia. Data were collected from individual case files, electronic databases, and paper copies of records pertaining to violent incidents and episodes of seclusion. Eighty five (44%) of the 193 patients admitted during this period were secluded. Those secluded were significantly younger and had a more established psychiatric history. LSI-R: SV scores were significantly and positively associated with being secluded. A statistical model containing three LSI-R: SV items, along with age on admission and psychiatric history, achieved an area under the curve of 0.74. Seclusion is used on a regular basis in response to a range of different forms of aggressive behaviour of different severity. The LSI-R: SV demonstrated moderate-to-good accuracy in predicting seclusion and warrants further research using detailed prospective methodologies.