Machine learning-based predictive modeling of depressive symptoms in Chinese adolescents

Comprehensive Summary

This study examined whether machine learning models could predict depressive symptoms among Chinese adolescents using lifestyle behaviors and socioeconomic factors. Data from over 32,000 students in grades 4–12 were analyzed, with depressive symptoms identified using the CES-D scale. The researchers applied multiple machine learning approaches and found that a random forest model performed best, achieving strong predictive accuracy. Key predictors included self-rated health, sleep duration, screen time, physical activity, dietary habits such as breakfast and egg intake, and parental support for exercise. The findings demonstrate that routinely collected, modifiable lifestyle factors can be effectively used to identify adolescents at higher risk for depressive symptoms.

Outcomes and Implications

This study highlights the potential for machine learning to support early mental health screening using non-invasive, low-cost data already available in school and healthcare settings. By relying on modifiable lifestyle behaviors rather than clinical diagnoses or neuroimaging biomarkers, the model offers a practical tool for large-scale prevention efforts, especially in populations where stigma limits mental health help-seeking. In the medical field, such approaches could help clinicians, public health professionals, and school-based health programs identify at-risk adolescents earlier and prioritize preventive interventions. More broadly, the study supports the integration of explainable AI into mental health care, shifting focus toward proactive, lifestyle-based strategies that align with precision medicine and preventive psychiatry.