A preliminary statistical model for identifying repeaters of parasuicide

This paper presents a statistical model constructed using logisticregression to identify those at high-risk of repeating parasuicide. Thesubjects in the study are Cork city residents who exhibited parasuicidalbehaviour between 1 January and 30 June 1995. Repetition of the behaviourwithin six months of the index episode distinguishes repeaters fromnon-repeaters. The model was designed so that it could be used bynon-clinicians and hence does not require information relating topsychiatric diagnosis or use of psychiatric services. The proportion ofsubjects correctly classified remained stable across a range of cut-pointprobabilities (mean = 86%, range: 83.9–87.5%). Using acut-point of 0.2, 96% of repeaters and 81% of non-repeaters were correctly classified. Using 0.45 led to the correct identification of81% of repeaters and 90% of non-repeaters. If these highlevels of sensitivity and specificity are maintained in validation tests onfuture cohorts in Cork city then the model could form the basis of anintervention programme designed to prevent the repetition of parasuicide.