LAUSR

This study aimed to explore the main influencing factors of suicide risk among Chinese students and establish an early warning model to provide interventions for high-risk students. We conducted surveys of students in their first and third years from a cohort study at Jining Medical College. Logistic regression models were used to screen the early warning factors, and four machine learning models were used to establish early warning models. There were 8 factors related to suicide risk that were eventually obtained through screening, including age, having a rough father, and CES-D, OHQ, ASLEC-4, BFI-Neuroticism, BFI-Openness, and MMC-AF-C scores. A random forest model with SMOTE was adopted, and it verified that these 8 early warning signs, for suicide risk can effectively predict suicide risk within 2 years with an AUC score of 0.947. Among the factors, we constructed a model that indicated that different personality traits affected suicide risk by different paths. Moreover, the factors obtained by screening can be used to identify college students in the same year with a high risk of suicide, with an AUC score that reached 0.953. Based on this study, we suggested some interventions to prevent students going high suicide risk.