Year: 2026
Type: preprint
Study design: Observational Study
Abstract: Background: Depressive symptoms and learning weariness represent overlapping dimensions of adolescent well-being. Characterizing their joint distribution may inform more differentiated assessment and support. This study examined joint screening phenotypes, identified socioecological correlates of co-occurrence, and compared three models for classifying the co-occurring phenotype. Methods: This school-based cross-sectional study included 57,753 students aged 10–18 years in Jiangsu Province, China. Screening results defined four mutually exclusive phenotypes: double negative, learning weariness only, depressive symptoms only, and co-occurring depressive symptoms and learning weariness. Multinomial logistic regression examined adjusted associations with co-occurrence relative to the double-negative group. Logistic regression, random forest (RF), and gradient boosting machine (GBM) models classified co-occurrence versus the remaining phenotypes using stratified training, validation, and test sets (70%, 15%, and 15%). Performance was assessed using areas under the receiver operating characteristic curve (AUROC) and precision–recall curve (AUPRC), Brier scores, sensitivity, and specificity. SHapley Additive exPlanations (SHAP) quantified predictor contributions to GBM outputs. Results: The double-negative, learning-weariness-only, depressive-symptoms-only, and co-occurring phenotypes accounted for 61.3%, 21.8%, 8.7%, and 8.2% of students, respectively. Bullying victimization, sleep problems, and internet addiction were associated with higher odds of co-occurrence, with adjusted odds ratios of 3.20, 2.68, and 2.51, respectively. Residential setting and family characteristics were also associated with co-occurrence. GBM achieved the highest test-set AUROC (0.886) and AUPRC (0.628), followed by RF (0.852 and 0.556) and logistic regression (0.811 and 0.441). Corresponding Brier scores were 0.096, 0.108, and 0.123. Sleep problems, internet addiction, and bullying victimization contributed most to GBM classification. Conclusions: Joint screening differentiated domain-specific and co-occurring emotional and academic difficulties, identifying a substantial learning-weariness-only group and an 8.2% co-occurring group. GBM demonstrated the strongest discrimination among the evaluated models. These findings highlight the potential value of integrated emotional and academic assessment and provide an empirical basis for further research on differentiated school-based support.
Source: Research Square
Institutions: Southeast University, Nanjing Audit University
Language: English
Subfield: Clinical Psychology
Field: Psychology
Domain: Social Sciences
Open Access status: green
Award: G6918834406
OpenAlex ID: W7225382621