To date, research on risk for self-harming behavior has been oflimited success. This has led to difficulty in predicting self-harm, anddeveloping effective intervention programs. The existing predictionliterature has focused on static risk factors, failed to adequately define or measure many important constructs, and has not yet developed predictionschemes of adequate accuracy. The current intervention literature is bothsmall and methodologically poor. This paper presents results from a

more »

... mining risk for self-harm in individuals with a history of psychiatrichospitalization. The data is taken from the MacArthur Risk Assessmentstudy, and corrects for the methodological problems that have marked theprediction literature. Logistic regressions found several risk and protective factors for both suicidal and non-suicidal acts of self harm respectively. Classification tree analyses correctly classified a high percentage of subjects, and provided descriptions of those at low, moderate and high risk for both suicidal and non-suicidal acts of self-harm. Implications for conceptualization, study and professional reaction to acts of self-harm are discussed.