Machine learning identifies prominent risk factors for depressive symptoms among Chinese children and adolescents

Identifying key risk factors for depressive symptoms in children and adolescents is crucial for prevention. However, few studies have explored this topic. This study aimed to examine the prevalence of depressive symptoms in Chinese children and adolescents and rank prominent risk factors.

A total of 198,062 children and adolescents were recruited from Chongqing, China. The Center for Epidemiological Studies Depression Scale for Children was used to assess depressive symptoms. Covariates were collected via single-item questions and scales. Logistic regression and three machine learning (ML) methods (Random Forest, Random Ferns, and Extreme Gradient Boosting) were used to identify and rank risk factors. Subgroup analysis was conducted to examine variations among different demographics.

The prevalence of depressive symptoms among children and adolescents was 29.6%. Risk factors included socio-demographics, family background, lifestyle, academic performance, trauma experiences, relationships, and personality traits (adjusted odds ratio = 1.03-1.88, all p < 0.05). ML analysis highlighted nine key risk factors: psychological resilience, age, sleep satisfaction, self-expectations, gender, relationship with father, academic performance, relationship with classmates, and breakfast frequency. A minimal model was established to identify depressive symptoms, achieving an AUC value of 0.844. Subgroup analysis showed similar factor rankings pattern in the overall sample.

Our study identified psychological resilience, age, sleep satisfaction, self-expectations, gender, relationship with father, academic performance, relationship with classmates, and breakfast frequency may be the prominent factors for Chinese children and adolescents. These factors should be prioritized in the prevention process of depressive symptoms.