Revealing suicide risk of young adults based on comprehensive measurements using decision tree classification

Predicting suicide risk based on risk and protective factors is a critical and complex endeavor. In this study, we combined insights from comprehensive aetiological theories on suicide with the methodological strengths of machine learning techniques. Our primary objectives were twofold: a) to identify hazardous feature combinations that characterize a high risk of suicide, and b) to enhance our understanding of the potential interactions between risk and protective factors related to suicide. We established an interpretable decision tree model to classify young adults at high risk of suicide, utilizing fifty-five variables covering distal, developmental, proximal, and social context factors from a large-scale cross-sectional survey (N = 88,214). The results highlighted the significance of variables such as self-compassion and non-suicidal self-injury (NSSI), and the accumulation of depressive symptoms, medium-to-low self-compassion, and a history of NSSI as substantial indicators of heightened suicide risk. This study serves as a valuable reference for the clinical identification of individuals at risk of suicide.