Predicting depressive symptoms from sleep quality in adolescents: a longitudinal study using the XGBoost algorithm

Objective To investigate the impacts of different dimensions of sleep quality on depressive symptoms among adolescents, thus providing evidence for the prevention of depressive symptoms among adolescents.

Methods A baseline survey was conducted from September to November 2023 for 2 373 adolescents selected by stratified cluster sampling from two junior high schools, two senior high schools, and two vocational high schools in Wuhan city. A follow-up survey was conducted from September to November 2024. The Pittsburgh Sleep Quality Index (PSQI) and the Center for Epidemiologic Studies Depression Scale (CES-D) were used for the investigation. The Least Absolute Shrinkage and Selection Operator (LASSO) regression was employed to screen the influencing factors, and the eXtreme Gradient Boosting (XGBoost) algorithm was adopted to construct a prediction model. The prediction performance was evaluated by the area under the curve (AUC).

Results The detection rates of depressive symptoms among adolescents in Wuhan were 31.73% at baseline and 32.28% at the one-year follow-up. The median (interquartile range) PSQI total score was 4 (2–6). Adolescents with depressive symptoms at follow-up had higher PSQI total score (5 vs. 3), sleep latency subscale score (1 vs. 0), and daytime dysfunction subscale score (2 vs. 1) than those without depressive symptoms (all P < 0.05). LASSO regression results indicated that baseline depressive symptoms (β = 0.15), daytime dysfunction (β = 0.04), sleep latency (β = 0.02), and sleep disturbances (β = 0.02) were influencing factors for follow-up depressive symptoms. The prediction model achieved an AUC of 0.775 (95%CI: 0.739–0.812). Baseline depressive symptoms and daytime dysfunction had the greatest impacts on follow-up depressive symptoms (with contributions of 0.36 and 0.20, respectively), and their interaction effect value was 0.022 (95%CI: 0.020–0.023).

Conclusions Daytime dysfunction is a key sleep-related risk factor for future depressive symptoms among adolescents. It interacts with baseline depressive symptoms, significantly amplifying the risk of subsequent depression.