ObjectiveEffective suicide prevention for at-risk adolescents requires increased access to medical treatment. Investigating the unique characteristics of suicide in this vulnerable at-risk non-referred sample can contribute to establishing effective suicide prevention policies. This study aimed to (a) examine at-risk non-referred adolescents’ suicide attempt rate, (b) investigate influential multilevel factors in predicting these adolescents’ suicide attempts, and (c) compare the results of (a) and (b) by gender.MethodsA total of 401 samples (216 boys and 185 girls) were recruited through a school-based mental health project for at-risk adolescents. Multivariate hierarchical logistic regression analyses were performed at the individual-, contextual-, and protective levels to evaluate three multilevel models as well as to investigate predictabilities for the overall group and by gender.ResultsThe suicide attempt rate of the overall sample was 29.4% (boys: 18.1%, girls: 42.7%), which was significantly higher than that of community samples. For boys, individual-level predictors (depression and conduct problems) had the most significant contribution in predicting suicide attempts. In contrast, for girls, protective-level predictors (family satisfaction) contributed the most to the prediction of suicide attempts, followed by contextual-level predictors (academic grades).ConclusionThis study is an important step in understanding the unique characteristics of at-risk non-referred adolescents who have not yet been considered in mental health policies. Improving medical accessibility will be the first step in establishing effective suicide prevention policies for these vulnerable samples.