Comprehensive Summary
This study by Fu et al. attempts to better understand factors associated with adolescent non-suicidal self-injury (NSSI) by applying interpretable machine learning methods. Survey data were collected from approximately 3,000 students in eastern China. The survey data included demographic information and seven validated psychological scales. Using this dataset, the researchers trained six algorithms to predict NSSI. The CatBoost algorithm showed the best performance. The analysis identified 23 influential factors, which were divided into groups such as situational anxiety, depressive symptoms, negative self-esteem, bullying and aggression, and interpersonal difficulties. Simultaneously, protective factors such as positive daily functioning and self-acceptance were identified. These results expand on the integrated theoretical model of the NSSI by showing how both resilience and risk factors shape behavior.
Outcomes and Implications
This study shows that depression, anxiety, and negative self-image are key risks for adolescent NSSI. It also showed that strong routines, resilience, and social support serve as important protective factors. The research shows that for medical professionals, screening should focus on both vulnerabilities and strengths rather than just on symptoms. The explainable AI tools helped the researchers uncover patterns that would have been easy to miss, like how anxiety plays out in certain situations or how being separated from parents can shape risk. These findings have the potential to help providers spot risks earlier and give support that fits each adolescent’s needs. While further longitudinal research is needed, this work shows the potential of machine learning to strengthen prevention and early support for adolescent mental health.