Predicting suicide attempts in a high-risk clinical cohort of adolescents using machine-learning

While suicide prevention is a major global health priority in youth, it remains challenging to identify those at risk for future suicidal behavior. This study aimed to investigate the potential of machine learning (ML) based approaches to predict suicide attempts (SA) among high-risk adolescents. We applied three ML-algorithms – elastic net, random forest and extreme gradient boosting – to longitudinal clinical data, and compared their performance to a traditional statistical approach – logistic regression (LR). Models were trained using patient data from a clinical cohort (N = 255) of adolescents with self-harming and risk-taking behaviors. Forty-four predictor variables including sociodemographic information, features of suicidal thoughts and behaviors, psychiatric disorders, global functioning scores and adverse childhood events were obtained at baseline and selected to predict future SAs within a two-year follow-up period. Performance metrics included area under the curve (AUC), Brier score, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). Moreover, individual predictor importance was explored. ML-algorithms outperformed LR in terms of overall predictive accuracy (AUC = 0.75–0.79 vs. 0.72) and model calibration (Brier scores = 0.18–0.20 vs. 0.24). A prior SA stood out as the most important predictor variable across all algorithms. Our findings demonstrate that SA can be predicted with good overall accuracy, even among patients with high prevalence of suicidal behaviors. The clinical utility of ML-based SA-predictions is subject to several limitations (e.g., ethical, legal, clinical). Alongside improvements of predictive performance, future research needs to address clinical and ethical implications of ML-based risk detection.