Predicting suicide behaviours in incarcerated settings

Attempts to predict suicide behaviours have produced a number of useful clinical tools. Unfortunately, these have been largely designed with a specific psychiatric population or institutional setting in mind and are not easily transferred to an incarcerated setting. In 1983 the authors developed a suicide checklist which could be used to aid screening of new admissions to Remand Centres in the Province of Alberta. Studies completed to date have revealed this checklist to be a practical and reliable method of standardizing the suicide screening process. This paper presents findings from a study designed to evaluate the predictive power (discriminant validity) of the checklist. A stepwise multivariate framework is used to assess the overall ability of checklist items to discriminate a high risk group. As well, the relative importance of specific socio-demographic, clinical and historical variables is assessed. Data were collected on a cohort of consecutive admissions to the Edmonton Remand Centre during 1986. Findings reveal that marital status is the only direct statistical predictor of suicide risk. Inmates who had divorced or separated from a spouse were more likely to be active suicide risks at some time during their remand than those who were married or single. Symptom score was found to be generally predictive, however, closer assessment revealed this relationship to be heterogeneous across subgroups defined on the basis of certain criminological variables. The model which allowed for the interaction of clinical with criminological variables correctly discriminated 100% of the active suicide risks and 63% of inmates who had a prior history of suicide behaviours. The practical implications of these findings for suicide screening are discussed.