Machine Learning-Based Prediction Model and Key Determinants of Adolescent Depression in China: A Multicenter Cross-Sectional Study

Machine Learning-Based Prediction Model and Key Determinants of Adolescent Depression in China: A Multicenter Cross-Sectional Study | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Machine Learning-Based Prediction Model and Key Determinants of Adolescent Depression in China: A Multicenter Cross-Sectional Study Liuyuan Li, Shuhua Zhang, wenhui Fan This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6887876/v1 This work is licensed under a CC BY 4.0 License Status: Under Revision Version 1 posted 17 You are reading this latest preprint version Abstract This study developed a multidimensional machine learning model grounded in ecological systems theory to enable early depression detection and targeted interventions among Chinese adolescents. Utilizing data from 7,169 middle school students (aged 14.98 ± 1.58 years, depression prevalence 36.01%) in Liaoning Province, China, 21 key predictors were selected via LASSO regression and Boruta algorithm. Five models (XGBoost, random forest, logistic regression, multilayer perceptron, and support vector machine) were evaluated. XGBoost demonstrated optimal performance (AUC = 0.912). SHAP analysis identified five core predictors: sleep quality (primary factor), authenticity, meaning in life, sense of control, and perceived peer-related stress. Protective factors included sleep quality, authenticity, and sense of control, while perceived peer-related stress was a risk factor. Nonlinear associations emerged between meaning in life and depression, with age-stratified thresholds (12–13 years: moderate meaning linked to highest risk; 14–16 years: moderate meaning reduced risk; 17–18 years: high meaning mitigated risk). Findings suggest a tri-level intervention framework: sleep-exercise programs (biological), age-specific resilience training (cognitive), and AI-driven school stress monitoring (environmental). Adolescent depression Machine learning Predictive modeling SHAP Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 1 Introduction Adolescent depression, which is evolving into a global public health problem, has a profound impact on the physical and mental health and social functioning of individual adolescents (Thapar et al., 2022 ). According to recent research, the pooled prevalence of depression in children and adolescents has reached 21.3% (Lu et al., 2024 ). A notable finding is that a particular study revealed a pooled prevalence of depression in Chinese children and adolescents as high as 26.17%, underscoring the severity of the problem in this demographic (Zhou et al., 2024 ). The consequences of depression for adolescents are extensive and far-reaching, manifesting in symptoms such as low mood, academic decline, social withdrawal, and an increased risk of self-injury and suicide. These effects can persist into adulthood, impacting career development and social adjustment (Shorey et al., 2022 ; Thapar et al., 2022 ; Zhou et al., 2024 ). Prior research suggests that adolescence is a time of high prevalence of depression (Shorey et al., 2022 ) and a critical period for identification and intervention (Hankin & Griffith, 2023 ; Thapar et al., 2022 ). Consequently, there is a pressing need to develop a predictive model that incorporates a range of factors and accurately identifies early underlying factors that precipitate depression in adolescents. The development of adolescent depression constitutes a multifaceted process shaped by the interplay of macro-level sociocultural contexts, proximal environmental factors (e.g., family and school dynamics), and individual-level intrinsic traits and behavioral patterns (Courtney et al., 2022 ). Existing studies have predominantly conceptualised these factors as independent variables and have investigated their associations with adolescent depression in isolation. However, this univariate analysis paradigm is challenging to employ comprehensively to elucidate the intricate mechanisms that underpin the occurrence and development of depression. Ecosystem theory emphasises the two-way interaction mechanism between the environment and the individual (Bronfenbrenner & Morris, 2006 ), and adolescent depression is essentially the result of the dynamic interaction between microsystems such as family, teachers, and peers, and the individual's contextual, psychological, and behavioural characteristics. Firstly, stressors from family, teachers, and peers can all have profound effects on adolescent psychology (Herres et al., 2016 ; Yang et al., 2024 ). Excessive parental discipline (Manuele et al., 2023 ), discordant family environments (Mastrotheodoros et al., 2020 ), teacher criticism (Guo et al., 2024 ), and peer rejection or isolation behaviours (Potter & Yoon, 2023 ) have been shown to increase the risk of adolescent depression. Secondly, sociodemographic variables, understood as individual background characteristics, have been demonstrated to play a significant role in the development of adolescent depression. Research has indicated that gender (Hua et al., 2024 ; Shorey et al., 2022 ; Thapar et al., 2022 ), age (Hua et al., 2024 ; Thapar et al., 2022 ; Zhou et al., 2024 ), and family economic status (Hua et al., 2024 ; Lu et al., 2024 ; Thapar et al., 2022 ), family structure (Wen et al., 2024 ), and parental education level (Xiang et al., 2024 ) have been identified as significant contributors to the risk of adolescent depression. Thirdly, disparities in individual psychological traits modulate adolescents' vulnerability to depression in response to stress (Monroe & Simons, 1991 ). Meaning in life (Baquero-Tomás et al., 2023 ; Ward et al., 2023 ), authenticity (Alchin et al., 2024 ; Xia & Xu, 2023 ), and sense of control (O'Neill et al., 2023), and self-identity (Wong & Hamza, 2023 ) as protective factors for depression have been shown to reduce psychopathological susceptibility to negative life events triggering depressive symptoms by shaping adaptive cognitive frameworks, moderating thresholds of emotional responses, optimizing the choice of coping strategies, and integrating social support resources, which together form the basis of adolescent psychological resilience. Fourthly, the role of lifestyle as an important behavioural characteristic in adolescent depression cannot be ignored. A substantial body of research has identified a nexus between unhealthy lifestyles and adolescent mental health, citing sleep disorders, inadequate exercise, poor dietary habits, sedentary behaviour, and excessive screen time as significant contributing factors (Garcia-Hermoso et al., 2022 ; Kleppang et al., 2023 ; Sampasa-Kanyinga et al., 2020 ). Utilising the aforementioned theoretical framework, this study has developed a multilevel predictive model incorporating four-dimensional variables, including interpersonal stressors (family-teacher-peers), demographic background, psychological quality, and lifestyle habits. The objective of this model is to elucidate the cross-system mechanism of action of adolescent depression and to furnish an integrative perspective for the early identification and intervention of this condition. Traditional statistical methods (e.g., multiple linear regression, structural equation modeling) rely on linear assumptions in multivariate modeling, which oversimplify complex nonlinear relationships into linear additive effects. This approach fails to identify curvilinear relationships, threshold effects, and higher-order interactions among variables. Furthermore, model complexity-sample size imbalance may induce parameter estimation bias and overfitting risks, while poor out-of-sample generalizability exacerbates the reproducibility crisis (Bzdok & Ioannidis, 2019 ; Dwyer & Koutsouleris, 2022 ). In contrast, machine learning (ML) methods employ nonparametric modeling (e.g., XGBoost’s gradient-boosted decision trees) to autonomously capture nonlinear interaction patterns. Regularization techniques effectively control model complexity, while Shapley value decomposition, partial dependence plots (PDPs), and accumulated local effects (ALE) plots enable global interpretation of feature contributions, overcoming the limitations of traditional linear frameworks without compromising predictive performance (Hastie et al., 2009 ). Leveraging ML’s analytical and predictive power (Shatte et al., 2019 ), this study integrates interpersonal stressors, sociodemographic backgrounds, psychological traits, and lifestyle factors into a unified framework to construct an ML prediction model capturing multidimensional interactions. This approach addresses three key limitations of traditional theory-driven research: 1. Reductionist tendencies in single-factor analysis; 2. Obscuration of complex interactions by linear assumptions; 3. Insufficient generalizability of predictive efficacy (Dwyer & Koutsouleris, 2022 ). SHAP interpretability methods—grounded in cooperative game theory’s Shapley values—precisely quantify each feature’s contribution to predictions, resolving the complexity of feature impact evaluation in ML models and limitations of traditional methods (Lundberg & Lee, 2017 ). PDPs visualize nonlinear relationships between individual predictors and depression risk (Hastie et al., 2009 ), while ALE plots illustrate cumulative contributions of variables to model predictions (Apley & Zhu, 2020 ). By employing multiple ML algorithms—multilayer perceptron (MLP), XGBoost, logistic regression, random forest (RF), and support vector machine (SVM)—this study aims to enhance the accuracy and efficacy of early adolescent depression detection, identify mechanisms of action of key factors, and establish a scientific foundation for targeted interventions. 2 Methods 2.1 Study Participants This study employed cluster sampling to recruit adolescents from 29 general middle schools in Liaoning Province, China. A total of 8,137 self-report questionnaires were distributed. After rigorous quality control (excluding invalid questionnaires with duplicate responses, missing values, or extreme values), 7,169 valid samples were retained, yielding an effective response rate of 88.10%. The final sample comprised 3,402 males (47.45%) and 3,767 females (52.55%), with geographic distribution as follows: urban (49.48%), county-level towns (21.59%), and rural areas (28.93%). Participants’ ages ranged from 11 to 20 years (14.98 ± 1.58). 2.2 Measurement Tools 2.2.1 Depression Assessment Depressive symptoms were assessed using the Patient Health Questionnaire-9 (PHQ-9; Kroenke et al., 2001 ). This 9-item scale adopts a Likert-4 scoring system (0 = "not at all" to 3 = "nearly every day"), with total scores ranging from 0 to 27. A cutoff score of ≥ 5 indicated depression. The Chinese version demonstrated good reliability and validity, with Cronbach’s α = 0.945 in this study. 2.2.2 Sleep The Pittsburgh Sleep Quality Index (PSQI; Buysse et al., 1989 ) evaluated sleep quality over the preceding month. This 18-item instrument assesses seven components: sleep quality, sleep latency, sleep duration, sleep efficiency, sleep disturbances, hypnotic medication use, and daytime dysfunction. Higher composite scores reflect poorer sleep quality. The Chinese version exhibited strong psychometric properties, achieving Cronbach’s α = 0.887 in this study. 2.2.3 Diet Dietary behaviors were assessed using the Dietary Behavior Questionnaire, which includes 10 items evaluating regularity of meals, binge eating, eating while studying or watching TV, and weekly frequencies of breakfast, fruit, vegetable, and milk consumption. Total scores range from 0 to 8, with scores ≥ 60% (≥ 4.8 points) classified as healthy. The Chinese version of this scale demonstrated good reliability (Cronbach’s α = 0.752). 2.2.4 Physical Activity Moderate-to-vigorous physical activity (MVPA) was measured using the World Health Organization Health Behavior in School-aged Children (WHO-HBSC) questionnaire. Participants reported the number of days per week (Monday–Friday and Saturday–Sunday) with ≥ 60 minutes of MVPA. Total weekly MVPA days ranged from 0 to 7. 2.2.5 Sedentary Time Sedentary behavior was assessed using the International Physical Activity Questionnaire (IPAQ) long-form. Based on the Chinese Physical Activity Guidelines for Children and Adolescents , sedentary time > 2 hours/day was classified as unhealthy. 2.2.6 Screen Time Screen time was defined as the total daily duration spent on electronic devices (TV, video games, smartphones, computers). Per national guidelines, > 2 hours/day was categorized as unhealthy. 2.2.7 Perceived Stress Assessment Interpersonal stress was evaluated using the Middle School Student Stressor Scale (Zheng Quanquan & Chen Shulin, 1999), comprising 23 items across four dimensions: perceived teacher-related stress, perceived family-environment stress, perceived parental discipline stress, and perceived peer-related stress. Items were rated on a 5-point Likert scale. The Chinese version exhibited excellent reliability: total scale Cronbach’s α = 0.918, with subscale α coefficients of 0.891 (perceived teacher-related stress), 0.725 (perceived family-environment stress), 0.713 (perceived parental discipline stress), and 0.824 (perceived peer-related stress). 2.2.8 Authenticity The 12-item Authenticity Scale (Wood et al., 2008 ) was administered using a 7-point Likert scale (1 = "strongly disagree" to 7 = "strongly agree"), with higher scores indicating greater authenticity. The Chinese version demonstrated good reliability (Cronbach’s α = 0.877). 2.2.9 Meaning in Life Meaning in life was assessed via the PIL-SF (Schulenberg et al., 2011 ), employing a 7-point scale (1 = "strongly disagree" to 7 = "strongly agree"). Higher scores reflect stronger life purpose and meaning. The Chinese adaptation showed excellent reliability (Cronbach’s α = 0.946). 2.2.10 Sense of Control A 12-item Sense of Control Scale (Lachman & Weaver, 1998 ) was used with a 7-point response format (1 = "strongly disagree" to 7 = "strongly agree"). Higher scores denote a greater sense of control. The Chinese version achieved good reliability (Cronbach’s α = 0.840). 2.2.11 Self-Identity The 12-item Self-Identity Scale (Kato, 1983 ) utilized a 6-point Likert scale (1 = "strongly disagree" to 6 = "strongly agree"), with higher scores indicating stronger self-identity. The Chinese version exhibited acceptable reliability (Cronbach’s α = 0.727). 2.2.12 Sociodemographic Variables Eleven demographic variables were collected: grade, gender, age, annual family income, boarding status, single or multi-child household, home location, family structure, mother’s education levels, and father’s education levels. 2.3 Statistical Analysis This study integrated 24 initial predictors from multidimensional domains to construct adolescent depression prediction models. To enhance model efficiency and mitigate overfitting, LASSO regression (Tibshirani, 1996 ) was applied to compress high-dimensional variable space, followed by the Boruta algorithm (Kursa & Rudnicki, 2010 ) to identify statistically significant predictors. After rigorous screening, 21 critical variables were retained for model development. Five machine learning algorithms (MLP, XGBoost, logistic regression, RF, and SVM)—were systematically evaluated. Data splitting adhered to stratified random sampling via the createDataPartition method (Kuhn, 2008 ), partitioning the dataset into training (80%) and independent test sets (20%). For XGBoost, SVM, and MLP, 25% of the training set was further allocated as a validation set (60-20-20 split) for hyperparameter tuning. To address class imbalance (depressed: non-depressed ≈ 1:3), differential correction strategies were implemented: Inverse class frequency weighting (Hastie et al., 2009 ) was applied to loss functions for logistic regression, RF, XGBoost, and SVM. Given MLP’s inability to directly incorporate sample weights, a synthetic minority oversampling technique (Random Over-Sampling Examples, ROSE) with boundary-smoothed resampling based on probability distributions was applied to the training set (Menardi & Torelli, 2014 ). Hyperparameter optimization employed grid search with 10-fold cross-validation. L1/L2 regularization mitigated overfitting in models, while tree-based algorithms (e.g., XGBoost, random forest) controlled complexity through depth limitation (max_depth), minimum leaf samples (min_samples_leaf), and feature subsampling (colsample_bytree), instance downsampling (subsample). Model performance was evaluated using AUC-ROC curves, with DeLong’s test (DeLong et al., 1988 ) comparing ROC curve differences. Additional metrics included accuracy, balanced accuracy, sensitivity, specificity, positive/negative predictive values, and F1-score derived from confusion matrices. Interpretability analyses leveraged SHAP values (Lundberg & Lee, 2017 ) to visualize predictor contributions, supplemented by partial dependence plots (Hastie et al., 2009 ) and accumulated local effects (ALE) plots (Apley & Zhu, 2020 ) to decode nonlinear predictor effects. The specific flow chart is shown in Fig. 1 . All analyses were conducted in R 4.3.3 and SPSS 27.0. 3 Results 3.1 Descriptive statistics The Supplementary table 1 shows the results of the descriptive statistics and the basic characteristics of the study variables. The screening results of this study indicated that 36.01% of the adolescents exhibited depressive symptoms. A subsequent analysis revealed significant disparities between participants with and without depressive symptoms with respect to sociodemographic indicators and scale scores. 3.2 Feature Selection A comparative analysis of LASSO regression and Boruta algorithm results (see Supplementary Fig. 1) was conducted to identify a shared subset of 21 key features that are strongly associated with adolescent depression for model construction. These include gender, age, home location, boarding status, family structure, single or multi-child household, mother's education level, father's education level, sleep quality, diet, weekend sedentary time, physical activity, weekly screen time, weekend screen time, perceived teacher-related stress, perceived family environment stress, perceived parental discipline stress, perceived peer-related stress, sense of control, authenticity, and meaning in life. 3.3 Model Performance As shown in Table 1 , this study comprehensively compared the predictive performance of five machine learning models: MLP, XGBoost, logistic regression, RF, and SVM. Results demonstrated that the XGBoost model achieved a significantly superior area under the receiver operating characteristic curve (AUC) compared to other models (see Fig. 2). On the test set, the XGBoost model exhibited a classification accuracy of 0.8238 (95% CI: 0.812–0.835), significantly exceeding the no-information rate (0.656, p < 0.001). Cohen’s Kappa coefficient of 0.629 indicated substantial agreement between predicted and actual classifications (Landis & Koch, 1977). For depression identification, the model achieved a sensitivity of 0.855, specificity of 0.807, positive predictive value (PPV) of 0.699, negative predictive value (NPV) of 0.914, balanced accuracy of 0.831, and F1-score of 0.769. DeLong’s test confirmed the statistical significance of XGBoost’s AUC superiority over other models ( p < 0.05), establishing its optimal discriminative power and clinical utility in adolescent depression prediction. 3.4 Variable Importance This study employed SHAP analysis to interpret the XGBoost model, generating SHAP summary plots (Fig. 3 A) and beeswarm plots (Fig. 3 B). The top five predictors of adolescent depression were identified as sleep quality, authenticity, meaning in life, sense of control, and perceived peer-related stress. Specifically, sleep disturbances and perceived peer-related stress exhibited positive correlations with depression risk, whereas authenticity and sense of control showed protective effects. Notably, meaning in life demonstrated a nonlinear association: both low and high levels served as protective factors, while moderate levels increased depression risk. Additional risk factors included female gender, rural/county residence, older age, non-only-child status, higher paternal education, weekly screen time > 2 hours, weekend sedentary time > 2 hours, perceived teacher-related stress, and perceived teacher-related stress. Protective factors encompassed daily moderate-intensity physical activity ≥ 1 hour. To further explore the mechanisms of the effects of the key variables, the present study plotted a partial dependence plot (PDP, see Supplementary Fig. 2) and a cumulative local effect plot (ALE, see Fig. 4 ). Initially, the results of both the PDP and ALE plots for sleep demonstrated that the predicted probability of depression in adolescents increased significantly as sleep disturbance scores increased, suggesting that good sleep quality may play a protective role in the development of depression. Secondly, the PDP and ALE plots for Authenticity exhibited a tendency towards a negative correlation between authenticity and the predicted probability of depression. Additionally, the PDP and ALE plots for sense of control demonstrated that after scores surpassed 40 (scale scores of 45%-50%), the level of sense of control exhibited an essentially negative correlation with the probability of depression. This finding suggests that higher levels of authenticity and sense of control may contribute to a reduced risk of depression. The PDP for meaning in life exhibited a positive association with depression below 8 points (scale scores of 25%-50%) and a negative association above 8 points, a threshold that may reflect dominant trends at the group level but is susceptible to affecting confounding variables such as age, gender, etc. The ALE controlled for feature covariance by adjusting for the conditional distributions, and the results showed a 12-point (scale scores of 25%-50%) threshold. The results of the ALE demonstrated that a score of 12 points was identified as the critical value. Scores below 12 points were found to be positively correlated, while scores above 12 points were negatively correlated. This finding was more closely aligned with the local causal effect, suggesting that the actual threshold may be higher and necessitate dynamic evaluation based on individual characteristics. The PDP and ALE plots of perceived peer-related stress demonstrated that perceived peer-related stress exhibited a positive association with the probability of depression when scores were below approximately 13 (scale scores of 25%-50%). The association leveled off after exceeding the threshold. In order to further explore the nonlinear relationship of meaning in life in different age groups of adolescents, the present study conducted a SHAP interaction analysis between age and meaning in life (see Fig. 5 ). The findings revealed a robust positive interaction between low values of meaning in life and age among the high age group (17–18 years old), along with a significant negative interaction between high values of meaning in life and age. These results suggest that low levels of meaning in life are associated with an elevated risk of depression, while high levels of meaning in life offer a protective effect against depression among adolescents in the high age group. In the middle age group (14–16 years), the interaction between meaning in life and age was relatively weak, with both higher and lower levels of meaning in life potentially exacerbating the risk of depression, and moderate levels of meaning in life mitigating the risk of depression. In the lower age group (12–13 years), a robust positive interaction emerged between moderate values of meaning in life (15–18 points) and age, suggesting that moderate levels of meaning in life amplify the risk of depression in lower-age adolescents. Conversely, a significant negative interaction was observed between low (10–11 points) and higher (19–20 points) values of meaning in life and age, indicating that lower levels of meaning in life may mitigate the risk of depression in these younger adolescents. 4 Discussion The following discussion will elaborate on the findings of this study, which compared five machine learning models commonly used to predict depression in Chinese adolescents. The results showed that the XGBoost model exhibited good performance with an accuracy of 0.824 (balanced precision = 0.831), an AUC of 0.912, a sensitivity of 0.855, and a specificity of 0.807. These findings are consistent with those reported in Mardini et al.'s (Mardini et al., 2025 ) cross-national study (N = 15,632) and Kuang et al.'s (Kuang et al., 2025 ) Chinese mega-cohort study (N = 583,405) to form a methodological validation, which collectively confirms the superiority of XGBoost in predicting depression in adolescents. As one of the most commonly used machine learning models for classification, the XGBoost model, through regularization and the DART tree integration strategies, can effectively prevent overfitting to noisy data and tends to have better performance when dealing with tabular datasets (Shwartz-Ziv & Armon, 2022 ). This study also examined the importance of each predictor. The findings of this study indicated that sleep quality, authenticity, sense of control, meaning in life, and perceived peer-related stressors were significant in predicting adolescent depression. Characteristic importance results indicated that sleep quality was the most significant factor contributing to depression, and as sleep quality improved, the risk of adolescent depression decreased. This finding aligns with the results of a prior machine learning study by Kuang et al. ( 2025 ), which identified sleep as the most significant segmented characteristic in predicting depression. Additionally, the machine learning study by Olfati et al. ( 2024 ) corroborated the importance of sleep quality in predicting the severity of depressive symptoms. Liang et al. ( 2021 ) found through a meta-analysis that Chinese adolescents' sleep disorders had a comorbid prevalence of 26%, and a study by Wang et al. ( 2021 ) noted that 70% of high school students did not get enough sleep during the school year. Sleep disorders have been demonstrated to be a risk factor for adolescent mental health (Palagini et al., 2022 ; Palmer et al., 2024 ; Wang et al., 2021 ), exacerbating adolescent mental health problems and suicide risk (Lam & Lam, 2021 ; Woodfield et al., 2024 ). Consequently, the present study endorses the incorporation of sleep interventions as a strategy for the prevention of adolescent depression (Freeman et al., 2020 ; Scott et al., 2021 ). It is recommended that educational institutions incorporate sleep education into their health education curricula, systematically assess the quality of sleep and psychological well-being of students, and devise customized psycho-sleep intervention programs based on the assessment outcomes. Research has identified authenticity as a significant protective factor against depression in adolescents. Adolescence is a period of exploration and the formation of authenticity, which serves as a foundation for self-perception (Alchin et al., 2024 ). Extensive research has indicated a negative correlation between authenticity and depression (Assor et al., 2021 ; Ionescu et al., 2022 ). According to the stress trait theory of depression (Monroe & Simons, 1991 ), the formation of authenticity fosters coherence in the integration of internalized self-narratives, thereby enhancing adolescents' ability to regulate their emotions and improve their psychological well-being (Xia & Xu, 2023 ). Conversely, instability in the core self-concept has been associated with feelings of division in social situations and significant fluctuations in emotional regulation, which can increase the risk of depression (Alchin et al., 2024 ). The present study found that the effect of meaning in life as a significant predictor of depression in adolescents showed significant age-specific effects. In the 17–18 age group, a low meaning in life was identified as a risk factor for depression, while a high meaning in life exhibited a protective effect, demonstrating psychological resilience. In the 14–16 age group, a moderate meaning in life served as a buffering factor against the risk of depression, with higher or lower levels likely to exacerbate the risk. In the 12–13 age group, an inverse effect was observed, with a moderate meaning in life exacerbating depression, and higher or lower level decreasing. This nonlinear relationship is rooted in the dynamic developmental processes of the cognitive-affective system. From a neurodevelopmental perspective, the prefrontal lobe functions as a higher-order cognitive center, and the degree of its functional integration with the limbic system determines the psychological effects of meaning in life (Andrews et al., 2021 ). Prefrontal-limbic integration is incomplete in 12- to 13-year-old adolescents, resulting in an inability to relate abstract meanings to their personal experiences, despite being able to recapitulate social norms (e.g., "it is important to study hard") (Wehmeyer et al., 2017 ). Individuals who possess a moderate meaning in life may exhibit superficial compliance with adult expectations, while concurrently experiencing internal conflict regarding their values. This proclivity for cognitive dissonance, exacerbated by rumination, has been observed to engender a heightened mental load (Chu & Fung, 2021 ). The localization of meaning, whether elevated or diminished, has been demonstrated to facilitate adaptation by attenuating cognitive demands, thereby mitigating the risk of depression. With age, prefrontal myelination fosters cognitive-emotional integration, enabling adolescents aged 17 to 18 to translate abstract meaning into personal narratives. At this stage, high sense makers reconstruct stressful events as growth narratives through dialectical thinking (Dulaney et al., 2018 ), while low sense makers are more prone to existential anxiety under multiple stressors due to a lack of goal orientation. 14–16 years of age serve as a transition period where the development of metacognitive skills supports the construction of meaning through reflection and adjustment through trial and error, and medium sense makers are able to maintain both a certain set of core values, but also have the flexibility to adjust meaning frameworks, whereas higher or lower meaning orientations can hinder role exploration in Erikson's theory and exacerbate depression through social comparison. This developmental trajectory reveals that adolescents are experiencing the “growing pains” of the cognitive revolution. Early risk for a moderate meaning in life stems from neurodevelopmental lags, in which any mental processes requiring higher-order integration may temporarily become cognitively burdensome when the emotional intensity of the limbic system outstrips the ability of the prefrontal lobes to regulate it. With the development of prefrontal-limbic functioning, the dichotomies of surface cognition and deep beliefs, social attachment and self-identity are gradually integrated, and the meaning in life is eventually transformed from a developmental risk factor to a kernel of psychological resilience. Therefore, intervention strategies need to match the developmental stage of adolescents: avoiding abstract questioning of meaning and guiding meaning exploration through figurative activities in 12–16-year-olds, and strengthening the reconstruction of stressful events and dialectical thinking training in 17–18-year-olds. The fourth significant predictor was sense of control, the level of which demonstrated a significant negative correlation with adolescent depression risk. This finding is consistent with the theory of learned helplessness (Seligman, 1975 ), which posits that individuals who repeatedly encounter uncontrollable negative events may develop generalized attributional patterns of hopelessness, potentially resulting in the onset of depressive symptoms. Empirical studies have further confirmed that sense of control, an important cornerstone of mental health, exerts a moderating effect on depressive risk primarily through the establishment of cognitive-emotional pathways of behavioral regulation (Ge et al., 2024 ; Stolz et al., 2020 ). Ge et al.'s study (2024) found that sense of control, by enhancing emotionally controllable beliefs and adaptive regulatory strategies, was able to reduce the risk of depression in adolescents. Research on neural mechanisms has identified a protective effect of perceptual control through the cognitive-emotional integration pathway in the prefrontal-limbic system. Individuals with strong control beliefs have been observed to employ adaptive strategies, such as cognitive reappraisal, in response to stress, and experience heightened positive emotions (Stolz et al., 2020 ). Chronic interpersonal stress has emerged as a robust predictor of adolescent depression (Losiewicz et al., 2023 ). Among various stressors, perceived peer stress and perceived teacher stress exerted the strongest influences in this study. Drawing from the interpersonal theory of depression, negative social interactions—such as peer rejection and critical teacher feedback—elevate depressive risk by altering self-perception and emotional regulation processes (Coyne, 1976 ). Adolescents undergoing psychological individuation prioritize peer and teacher relationships as central to their social lives during this developmental stage. The intensifying importance of peer validation in adolescence makes negative peer experiences (e.g., exclusion, bullying) particularly traumatic (Kim, 2021 ; Potter & Yoon, 2023 ). Longitudinal evidence indicates that adolescents reporting high perceived peer-related stress demonstrate increased odds of developing persistent depressive symptoms compared to peers with low peer stress (Agoston & Rudolph, 2016 ; Liao et al., 2022 ). Perceived teacher-related stressors also play a critical role. Negative teacher-student interactions (e.g., excessive criticism, biased evaluations) trigger stronger emotional reactivity in adolescents than parent-child conflicts or peer rejection, with significant longitudinal associations to depressive symptom persistence (Herres et al., 2016 ). Additionally, academic pressures—including unrealistic performance expectations, frequent assessments, and overloaded assignments—exacerbate perceived teacher-related stress, leading to an increase in depression risk among Chinese adolescents (Jiang et al., 2021 ). These findings underscore the need for school-based interventions targeting: (1) peer relationship skills training to mitigate rejection sensitivity, (2) teacher-student communication workshops to reduce negative feedback, and (3) dynamic stress monitoring systems to identify at-risk individuals. The study also found that engaging in at least one hour of moderate-intensity physical activity daily, limiting screen time to ≤ 2 hours on weekdays, and restricting weekend sedentary time to ≤ 2 hours were significantly associated with a reduced risk of adolescent depression. In line with prior research findings (Rodriguez-Ayllon et al., 2019 ; Li et al., 2025 ), the present study corroborated that adolescents who adhered strictly to physical activity recommendations exhibited a reduced risk of depression. A substantial body of research has demonstrated the efficacy of physical activity interventions in enhancing adolescents' mental well-being (Recchia et al., 2023 ; Tomkinson et al., 2018 ). Consequently, this study proposes the optimization of the design of physical education programs on campuses. Furthermore, the study identified demographic variables such as female gender, residing in rural or county areas, advanced age, not being an only child, and elevated paternal education levels as risk factors for depression among adolescents. This finding underscores the necessity to prioritize this population. 5 Limitations This study breaks through the traditional unidimensional analysis framework and systematically integrates environmental factors such as family, teachers, and peers with individual background characteristics, psychological characteristics, and behavioral characteristics to construct a multidimensional depression prediction system for Chinese adolescents, but some limitations still exist. First, the samples were all from adolescents in Northeast China, which may limit the generalizability of the findings. Second, the cross-sectional design made it difficult to distinguish the temporal relationship between predictors and depressive symptoms, e.g., sleep disturbance may be both a precursor symptom of depression and an outcome variable due to environmental stress (Palagini et al., 2022 ), which needs to be further validated through follow-up studies. Third, in the current depression assessment system, the self-rating scale (PHQ − 9) is mainly relied upon to determine the depression status of adolescents and is not combined with clinical diagnosis and neurobiological indicators (e.g., fMRI, cortisol levels) for a comprehensive assessment. 6 Conclusion This study compared the effectiveness of five machine learning models in predicting depression in Chinese adolescents. The results show that XGBoost has optimal efficacy in predicting depression in Chinese adolescents, and the key predictors identified include sleep quality, authenticity, sense of control, meaning in life, perceived peer-source pressure, and perceived teacher-source pressure. Among the key predictors, sleep quality served as the most important protective factor, suggesting that sleep quality monitoring should be incorporated into the campus routine mental health assessment system. Authenticity and sense of control reduce the risk of depression by strengthening the internal stability and external adaptability of psychological resilience, suggesting that schools should increase the content of related psychological education; the age-specificity of the meaning in life suggests a differentiated intervention strategy: for the group of 12-16-year-olds, it is recommended to avoid early abstract questioning and integrate meaning construction into practical experience; for the group of 17-18-year-olds, it is recommended that they should be guided to reconstruct the stressful events into self-growth narratives. In addition, perceived peer-related stress and perceived teacher-related pressure were found to be risk factors for depression, suggesting the need to build a dynamic early warning and support network for campus stress. Future research could further integrate the ecological transient assessment data to reveal the cross-level interaction mechanisms of the predictors. Declarations a. Ethics approval and consent to participate All participants were fully informed about the purpose, procedures, and potential risks of the study. Written informed consent was obtained from each participant prior to their involvement. For participants under the age of 18, consent was obtained from their parents or legal guardians. The study was conducted in accordance with the ethical standards of the relevant institutional review board.. All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The study was approved by the Bioethics Committee of China Medical University, as indicated by the ethics approval number 2024/169. b. Consent for publication Not Applicable. c. Availability of data and materials The current study’s findings are available from the corresponding author upon reasonable request. d. Competing interests All participants provided written informed consent to participate in this study. e. Funding The authors did not receive support from any organization for the submitted work. f. Authors' contribution Liuyuan Li played a lead role in conceptualization, data collection, formal analysis, and writing–original draft, and played a supporting role in writing–review and editing. Shuhua Zhang played a supporting role in conceptualization, and a lead role in writing–review and editing. Wenhui Fan played a supporting role in literature review. All authors read and approved the final manuscript. g. Acknowledgments We are grateful to the participating schools that helped collect the data. We also thank parents for granting permission and students for their participation. h. Authors' information (optional) FIRST AUTHOR: Liuyuan Li, Affiliation: College of Medical Humanities, China Medical University, Shenyang, China; Address: China Medical University, 78 Puhe Road, Shenbei New District, Shenyang City, Liaoning Province, China; ORCID：0009-0009-0190-9367; Email: [email protected] ; THIRD AUTHOR: Wenhui Fan, Affiliation: College of Educational Science, Shenyang Normal University, Shenyang, China; Address: Shenyang Normal University, 253 Huanghe North Street, Huanggu District, Shenyang City, Liaoning Province, China; Email: [email protected] ； SECOND AUTHOR: Shuhua Zhang, Affiliation: College of Medical Humanities, China Medical University, Shenyang, China; Address: China Medical University, 78 Puhe Road, Shenbei New District, Shenyang City, Liaoning Province, China. Email: [email protected] ; Correspondence concerning this article should be addressed to Shuhua Zhang, Email: [email protected] ; Tel: 18900916785. References Agoston, A. M., & Rudolph, K. D. (2016). Interactive contributions of cumulative peer stress and executive function deficits to depression in early adolescence. The Journal of Early Adolescence , 36 (8), 1070-1094. https://doi.org/10.1177/0272431615593176 Alchin, C. E., Machin, T. M., Martin, N., & Burton, L. J. (2024). Authenticity and Inauthenticity in Adolescents: A Scoping Review. Adolescent Research Review , 9 (2), 279-315. https://doi.org/10.1007/s40894-023-00218-8 Andrews, J. L., Ahmed, S. P., & Blakemore, S.-J. (2021). Navigating the Social Environment in Adolescence: The Role of Social Brain Development. Biological Psychiatry , 89 (2), 109–118. https://doi.org/10.1016/j.biopsych.2020.09.012 Apley, D. W., & Zhu, J. (2020). Visualizing the effects of predictor variables in black box supervised learning models. Journal of the Royal Statistical Society Series B: Statistical Methodology , 82 (4), 1059-1086. https://doi.org/10.1111/rssb.12377 Assor, A., Benita, M., Shi, Y., Goren, R., Yitshaki, N., & Wang, Q. (2021). The Authentic Inner Compass as a Well-Being Resource: Predictive Effects on Vitality, and Relations with Self-Esteem, Depression and Behavioral Self-realization. JOURNAL OF HAPPINESS STUDIES , 22 (8), 3435-3455. https://doi.org/10.1007/s10902-021-00373-6 Baquero-Tomás, M., Grau, M. D., Moliner, A.-R., & Sanchis-Sanchis, A. (2023). Meaning in life as a protective factor against depression. FRONTIERS IN PSYCHOLOGY , 14 , 1180082. https://doi.org/10.3389/fpsyg.2023.1180082 Bronfenbrenner, U., & Morris, P. A. (2006). The bioecological model of human development. In Handbook of child psychology: Theoretical models of human development (6th ed., Vol. 1, pp. 793–828). John Wiley & Sons. Buysse, D. J., Reynolds, C. F., Monk, T. H., Berman, S. R., & Kupfer, D. J. (1989). The pittsburgh sleep quality index: A new instrument for psychiatric practice and research. Psychiatry Research , 28 (2), 193-213. https://doi.org/10.1016/0165-1781(89)90047-4 Bzdok, D., & Ioannidis, J. P. A. (2019). Exploration, inference, and prediction in neuroscience and biomedicine. Trends in Neurosciences , 42 (4), 251–262. https://doi.org/10.1016/j.tins.2019.02.001 Chu, S. T.-W., & Fung, H. H.-L. (2021). Is the search for meaning related to the presence of meaning? Moderators of the longitudinal relationship. JOURNAL OF HAPPINESS STUDIES , 22 (1), 127-145. https://doi.org/10.1007/s10902-020-00222-y Courtney, D. B., Watson, P., Krause, K. R., Chan, B. W. C., Bennett, K., Gunlicks-Stoessel, M., Rodak, T., Neprily, K., Zentner, T., & Szatmari, P. (2022). Predictors, moderators, and mediators associated with treatment outcome in randomized clinical trials among adolescents with depression: A scoping review. JAMA Network Open , 5 (2), e2146331. https://doi.org/10.1001/jamanetworkopen.2021.46331 Coyne, J. C. (1976). Toward an interactional description of depression. Psychiatry . DeLong, E. R., DeLong, D. M., & Clarke-Pearson, D. L. (1988). Comparing the areas under two or more correlated receiver operating characteristic curves: A nonparametric approach. Biometrics , 44 (3), 837-845. Dulaney, E. S., Graupmann, V., Grant, K. E., Adam, E. K., & Chen, E. (2018). Taking on the stress-depression link: Meaning as a resource in adolescence. JOURNAL OF ADOLESCENCE , 65 , 39-49. https://doi.org/10.1016/j.adolescence.2018.02.011 Dwyer, D., & Koutsouleris, N. (2022). Annual research review: Translational machine learning for child and adolescent psychiatry. JOURNAL OF CHILD PSYCHOLOGY AND PSYCHIATRY , 63 (4), 421-443. https://doi.org/10.1111/jcpp.13545 Freeman, D., Sheaves, B., Waite, F., Harvey, A. G., & Harrison, P. J. (2020). Sleep disturbance and psychiatric disorders. The Lancet Psychiatry , 7 (7), 628-637. https://doi.org/10.1016/S2215-0366(20)30136-X Garcia-Hermoso, A., Ezzatvar, Y., Ramirez-Velez, R., Lopez-Gil, J. F., & Izquierdo, M. (2022). Trajectories of 24-h movement guidelines from middle adolescence to adulthood on depression and suicidal ideation: a 22-year follow-up study. INTERNATIONAL JOURNAL OF BEHAVIORAL NUTRITION AND PHYSICAL ACTIVITY , 19 (1), 135. https://doi.org/10.1186/s12966-022-01367-0 Ge, W., Wang, H., Chen, J., Yang, X., & Wang, Y. (2024). How general control beliefs relate to depressive symptoms in adolescents: The role of emotion controllability beliefs and emotion regulation. CURRENT PSYCHOLOGY , 43 (35), 28531-28543. https://doi.org/10.1007/s12144-024-06489-1 Guo, N., Huebner, E. S., Gong, X., & Tian, L. (2024). Psychological maltreatment by teachers and peer victimization in Chinese youth: Depression and aggression as mediators. Child Abuse and Neglect , 153 , 106809. https://doi.org/10.1016/j.chiabu.2024.106809 Hankin, B. L., & Griffith, J. M. (2023). What do we know about depression among youth and how can we make progress toward improved understanding and reducing distress? A new hope. Clinical Child and Family Psychology Review , 26 (4), 919-942. https://doi.org/10.1007/s10567-023-00437-4 Hastie, T., Tibshirani, R., & Friedman, J. (2009). The elements of statistical learning . Springer. Herres, J., Ewing, E. S. K., & Kobak, R. (2016). Emotional reactivity to negative adult and peer events and the maintenance of adolescent depressive symptoms: A daily diary design. JOURNAL OF ABNORMAL CHILD PSYCHOLOGY , 44 (3), 471-481. https://doi.org/10.1007/s10802-015-0043-6 Hua, Z., Wang, S., & Yuan, X. (2024). Trends in age-standardized incidence rates of depression in adolescents aged 10–24 in 204 countries and regions from 1990 to 2019. JOURNAL OF AFFECTIVE DISORDERS , 350 , 831-837. https://doi.org/10.1016/j.jad.2024.01.009 Ionescu, D., Burtăverde, V., Avram, E., & Ene, C. (2022). Authenticity, Volition, and Motivational Persistence Predicting Well-being: a Self-determination Theoretical Perspective. Trends in Psychology , 30 (3), 468-479. https://doi.org/10.1007/s43076-021-00138-3 Jiang, S., Ren, Q., Jiang, C., & Wang, L. (2021). Academic stress and depression of Chinese adolescents in junior high schools: Moderated mediation model of school burnout and self-esteem. JOURNAL OF AFFECTIVE DISORDERS , 295 , 384-389. https://doi.org/10.1016/j.jad.2021.08.085 Kim, K. M. (2021). What makes adolescents psychologically distressed? Life events as risk factors for depression and suicide. European Child and Adolescent Psychiatry , 30 (3), 359-367. https://doi.org/10.1007/s00787-020-01520-9 Kleppang, A. L., Vettore, M. V., Hartz, I., Haugland, S. H., & Stea, T. H. (2023). Patterns of unhealthy behaviours during adolescence and subsequent anxiety and depression in adulthood: a prospective register linkage study of the HUNT survey and health registries. INTERNATIONAL JOURNAL OF BEHAVIORAL NUTRITION AND PHYSICAL ACTIVITY , 20 (1), 6. https://doi.org/10.1186/s12966-023-01408-2 Kroenke, K., Spitzer, R. L., & Williams, J. B. W. (2001). The PHQ-9. Journal of General Internal Medicine , 16 (9), 606-613. https://doi.org/10.1046/j.1525-1497.2001.016009606.x Kuang, Y., Liao, X., Jiang, Z., Gu, Y., Liu, B., Tan, C., Zhang, W., & Li, K. (2025). Federated learning-based prediction of depression among adolescents across multiple districts in China. JOURNAL OF AFFECTIVE DISORDERS , 369 , 625-632. https://doi.org/10.1016/j.jad.2024.10.027 Kuhn, M. (2008). Building predictive models in R using the caret package. JOURNAL OF STATISTICAL SOFTWARE , 28 , 1-26. https://doi.org/10.18637/jss.v028.i05 Kursa, M. B., & Rudnicki, W. R. (2010). Feature selection with the boruta package. JOURNAL OF STATISTICAL SOFTWARE , 36 , 1-13. https://doi.org/10.18637/jss.v036.i11 Lachman, M. E., & Weaver, S. L. (1998). The Sense of Control as a Moderator of Social Class Differences in Health and Well-Being: Journal of Personality & Social Psychology. Journal of Personality & Social Psychology , 74 (3), 763-773. https://doi.org/10.1037/0022-3514.74.3.763 Lam, L. T., & Lam, M. K. (2021). Sleep Disorders in Early Childhood and the Development of Mental Health Problems in Adolescents: A Systematic Review of Longitudinal and Prospective Studies. INTERNATIONAL JOURNAL OF ENVIRONMENTAL RESEARCH AND PUBLIC HEALTH , 18 (22), 11782. https://doi.org/10.3390/ijerph182211782 Li, L., Guo, D., Shi, C., & Zheng, Y. (2025). The predictive role of sedentary behavior and physical activity on adolescent depressive symptoms: A machine learning approach. Journal of Affective Disorders , 378 , 81–89. https://doi.org/10.1016/j.jad.2025.02.085 Liang, M., Guo, L., Huo, J., & Zhou, G. (2021). Prevalence of sleep disturbances in Chinese adolescents: A systematic review and meta-analysis. PLOS ONE , 16 (3), e0247333. https://doi.org/10.1371/journal.pone.0247333 Liao, Y., Chen, J., Zhang, Y., & Peng, C. (2022). The reciprocal relationship between peer victimization and internalizing problems in children and adolescents: A meta-analysis of longitudinal studies. Acta Psychologica Sinica , 54 (7), 828-849. https://doi.org/10.3724/SP.J.1041.2022.00828 Losiewicz, O. M., Metts, A. V., Zinbarg, R. E., Hammen, C., & Craske, M. G. (2023). Examining the indirect contributions of irritability and chronic interpersonal stress on symptoms of anxiety and depression in adolescents. JOURNAL OF AFFECTIVE DISORDERS , 329 , 350-358. https://doi.org/10.1016/j.jad.2023.02.125 Lu, B., Lin, L., & Su, X. (2024). Global burden of depression or depressive symptoms in children and adolescents: A systematic review and meta-analysis. JOURNAL OF AFFECTIVE DISORDERS , 354 , 553-562. https://doi.org/10.1016/j.jad.2024.03.074 Lundberg, S. M., & Lee, S.-I. (2017, 2017-12-04). A unified approach to interpreting model predictions. NIPS'17 Manuele, S. J., Yap, M. B. H., Lin, S. C., Pozzi, E., & Whittle, S. (2023). Associations between paternal versus maternal parenting behaviors and child and adolescent internalizing problems: A systematic review and meta-analysis. Clinical Psychology Review , 105 , 102339. https://doi.org/10.1016/j.cpr.2023.102339 Mardini, M. T., Khalil, G. E., Bai, C., DivaKaran, A. M., & Ray, J. M. (2025). Identifying adolescent depression and anxiety through real-world data and social determinants of health: Machine learning model development and validation. JMIR MENTAL HEALTH , 12 , e66665. https://doi.org/10.2196/66665 Mastrotheodoros, S., Canário, C., Cristina Gugliandolo, M., Merkas, M., & Keijsers, L. (2020). Family Functioning and Adolescent Internalizing and Externalizing Problems: Disentangling between-, and Within-Family Associations. JOURNAL OF YOUTH AND ADOLESCENCE , 49 (4), 804-817. https://doi.org/10.1007/s10964-019-01094-z Menardi, G., & Torelli, N. (2014). Training and assessing classification rules with imbalanced data. Data Mining and Knowledge Discovery , 28 (1), 92-122. https://doi.org/10.1007/s10618-012-0295-5 Monroe, S. M., & Simons, A. D. (1991). Diathesis-stress theories in the context of life stress research: Implications for the depressive disorders. PSYCHOLOGICAL BULLETIN , 110 (3), 406-425. https://doi.org/10.1037/0033-2909.110.3.406 O‘Neill, A., Humphrey, N., & Stapley, E. (2023). A systematic review of qualitative research focusing on emotional distress among adolescents: Perceived cause and help-seeking. Adolescent Research Review , 8 (3), 387-402. https://doi.org/10.1007/s40894-022-00203-7 Olfati, M., Samea, F., Faghihroohi, S., Balajoo, S. M., Küppers, V., Genon, S., Patil, K., Eickhoff, S. B., & Tahmasian, M. (2024). Prediction of depressive symptoms severity based on sleep quality, anxiety, and gray matter volume: A generalizable machine learning approach across three datasets. eBioMedicine , 108 . https://doi.org/10.1016/j.ebiom.2024.105313 Palagini, L., Hertenstein, E., Riemann, D., & Nissen, C. (2022). Sleep, insomnia and mental health. Journal of Sleep Research , 31 (4), e13628. https://doi.org/10.1111/jsr.13628 Palmer, C. A., Bower, J. L., Cho, K. W., Clementi, M. A., Lau, S., Oosterhoff, B., & Alfano, C. A. (2024). Sleep loss and emotion: A systematic review and meta-analysis of over 50 years of experimental research. PSYCHOLOGICAL BULLETIN , 150 (4), 440-463. https://doi.org/10.1037/bul0000410 Potter, J. R., & Yoon, K. L. (2023). Interpersonal factors, peer relationship stressors, and gender differences in adolescent depression. Current Psychiatry Reports , 25 (11), 759-767. https://doi.org/10.1007/s11920-023-01465-1 Recchia, F., Bernal, J. D. K., Fong, D. Y., Wong, S. H. S., Chung, P.-K., Chan, D. K. C., Capio, C. M., Yu, C. C. W., Wong, S. W. S., Sit, C. H. P., Chen, Y.-J., Thompson, W. R., & Siu, P. M. (2023). Physical activity interventions to alleviate depressive symptoms in children and adolescents: A systematic review and meta-analysis. JAMA Pediatrics , 177 (2), 132-140. https://doi.org/10.1001/jamapediatrics.2022.5090 Rodriguez-Ayllon, M., Cadenas-Sánchez, C., Estévez-López, F., Muñoz, N. E., Mora-Gonzalez, J., Migueles, J. H., Molina-García, P., Henriksson, H., Mena-Molina, A., Martínez-Vizcaíno, V., Catena, A., Löf, M., Erickson, K. I., Lubans, D. R., Ortega, F. B., & Esteban-Cornejo, I. (2019). Role of Physical Activity and Sedentary Behavior in the Mental Health of Preschoolers, Children and Adolescents: A Systematic Review and Meta-Analysis. Sports Medicine (Auckland, N.Z.) , 49 (9), 1383-1410. https://doi.org/10.1007/s40279-019-01099-5 Sampasa-Kanyinga, H., Colman, I., Goldfield, G. S., Janssen, I., Wang, J., Podinic, I., Tremblay, M. S., Saunders, T. J., Sampson, M., & Chaput, J.-P. (2020). Combinations of physical activity, sedentary time, and sleep duration and their associations with depressive symptoms and other mental health problems in children and adolescents: a systematic review. The International Journal of Behavioral Nutrition and Physical Activity , 17 (1), 72. https://doi.org/10.1186/s12966-020-00976-x Schulenberg, S. E., Schnetzer, L. W., & Buchanan, E. M. (2011). The Purpose in Life Test-Short Form: Development and Psychometric Support. JOURNAL OF HAPPINESS STUDIES , 12 (5), 861-876. https://doi.org/10.1007/s10902-010-9231-9 Scott, J., Kallestad, H., Vedaa, O., Sivertsen, B., & Etain, B. (2021). Sleep disturbances and first onset of major mental disorders in adolescence and early adulthood: A systematic review and meta-analysis. SLEEP MEDICINE REVIEWS , 57 , 101429. https://doi.org/10.1016/j.smrv.2021.101429 Seligman, M. E. P. (1975). Helplessness: on depression, development, and death . W H Freeman/Times Books/ Henry Holt & Co. Shatte, A. B. R., Hutchinson, D. M., & Teague, S. J. (2019). Machine learning in mental health: a scoping review of methods and applications. Psychological Medicine , 49 (9), 1426-1448. https://doi.org/10.1017/S0033291719000151 Shorey, S., Ng, E. D., & Wong, C. H. J. (2022). Global prevalence of depression and elevated depressive symptoms among adolescents: A systematic review and meta-analysis. British Journal of Clinical Psychology , 61 (2), 287-305. https://doi.org/10.1111/bjc.12333 Shwartz-Ziv, R., & Armon, A. (2022). Tabular data: Deep learning is not all you need. Information Fusion , 81 , 84-90. https://doi.org/10.1016/j.inffus.2021.11.011 Stolz, D. S., Müller-Pinzler, L., Krach, S., & Paulus, F. M. (2020). Internal control beliefs shape positive affect and associated neural dynamics during outcome valuation. Nature Communications , 11 (1), 1230. https://doi.org/10.1038/s41467-020-14800-4 Thapar, A., Eyre, O., Patel, V., & Brent, D. (2022). Depression in young people. Lancet , 400 (10352), 617-631. https://doi.org/10.1016/S0140-6736(22)01012-1 Tibshirani, R. (1996). Regression shrinkage and selection via the lasso. Journal of the Royal Statistical Society: Series B (methodological) , 58 (1), 267-288. https://doi.org/10.1111/j.2517-6161.1996.tb02080.x Tomkinson, G. R., Carver, K. D., Atkinson, F., Daniell, N. D., Lewis, L. K., Fitzgerald, J. S., Lang, J. J., & Ortega, F. B. (2018). European normative values for physical fitness in children and adolescents aged 9-17 years: Results from 2 779 165 eurofit performances representing 30 countries. British Journal of Sports Medicine , 52 (22), 1445-1456. https://doi.org/10.1136/bjsports-2017-098253 Wang, W., Du, X., Guo, Y., Li, W., Teopiz, K. M., Shi, J., Guo, L., Lu, C., & McIntyre, R. S. (2021). The associations between sleep situations and mental health among Chinese adolescents: A longitudinal study. Sleep Medicine , 82 , 71-77. https://doi.org/10.1016/j.sleep.2021.03.009 Ward, S., Womick, J., Titova, L., & King, L. (2023). Meaning in Life and Coping With Everyday Stressors. Personality & Social Psychology Bulletin , 49 (3), 460-476. https://doi.org/10.1177/01461672211068910 Wehmeyer, M. L., Shogren, K. A., Little, T. D., & Lopez, S. J. (Eds.). (2017). Development of Self-Determination Through the Life-Course . Springer Netherlands. https://doi.org/10.1007/978-94-024-1042-6 Wen, S., Yu, X., Jin, L., Gong, J., Zhang, X., Sun, J., Zhang, S., & Lyu, H. (2024). A three-level meta-analysis of the relationship between family dysfunction and mental health of children and adolescents. Advances in Psychological Science , 32 (5), 771. https://doi.org/10.3724/SP.J.1042.2024.00771 Wong, T. K. Y., & Hamza, C. A. (2023). Longitudinal Associations among Identity Processes and Mental Health in Young Adulthood: The Mediating Role of Social Support. JOURNAL OF YOUTH AND ADOLESCENCE . https://doi.org/10.1007/s10964-023-01883-7 Wood, A. M., Linley, P. A., Maltby, J., Baliousis, M., & Joseph, S. (2008). The authentic personality: A theoretical and empirical conceptualization and the development of the Authenticity Scale. Journal of Counseling Psychology , 55 (3), 385-399. https://doi.org/10.1037/0022-0167.55.3.385 Woodfield, M., Butler, N. G., & Tsappis, M. (2024). Impact of sleep and mental health in adolescence: an overview. CURRENT OPINION IN PEDIATRICS , 36 (4), 375-381. https://doi.org/10.1097/MOP.0000000000001358 Xia, M., & Xu, X. (2023). Does authenticity always breed mental health? A cross-cultural comparison between the United States and China. ASIAN JOURNAL OF SOCIAL PSYCHOLOGY , 26 (1), 132-145. https://doi.org/10.1111/ajsp.12549 Xiang, Y., Cao, R., & Li, X. (2024). Parental education level and adolescent depression: A multi-country meta-analysis. JOURNAL OF AFFECTIVE DISORDERS , 347 , 645-655. https://doi.org/10.1016/j.jad.2023.11.081 Yang, X., Fang, Y., Wang, Y., Liu, S., Cai, J., Li, H., & Chen, Z. (2024). Multidimensional stressors and depressive and anxiety symptoms in adolescents: A network analysis through simulations. JOURNAL OF AFFECTIVE DISORDERS , 347 , 364-374. https://doi.org/10.1016/j.jad.2023.11.057 Zhou, J., Liu, Y., Ma, J., Feng, Z., Hu, J., Hu, J., & Dong, B. (2024). Prevalence of depressive symptoms among children and adolescents in China: A systematic review and meta-analysis. Child And Adolescent Psychiatry And Mental HEALTH , 18 (1), 150. https://doi.org/10.1186/s13034-024-00841-w Kato, A. (1983). Aspects and structure of identity among college students [in Japanese]. Japanese Journal of Educational Psychology , 31(4), 292–302. https://doi.org/10.5926/jjep1953.31.4_292 Zheng, Q. Q., & Chen, S. L. (1999). Preliminary development of a stressor scale for middle school students [in Chinese]. Psychological Development and Education , (4), 45–49. Supplementary Material Supplementary Table 1 and Supplementary Figures 1-2 are not available with this version. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Revision Version 1 posted Editorial decision: Revision requested 21 May, 2026 Reviews received at journal 18 May, 2026 Reviewers agreed at journal 14 May, 2026 Reviews received at journal 16 Aug, 2025 Reviews received at journal 30 Jul, 2025 Reviewers agreed at journal 30 Jul, 2025 Reviews received at journal 30 Jul, 2025 Reviewers agreed at journal 30 Jul, 2025 Reviews received at journal 29 Jul, 2025 Reviewers agreed at journal 28 Jul, 2025 Reviewers agreed at journal 28 Jul, 2025 Reviewers agreed at journal 24 Jul, 2025 Reviewers invited by journal 21 Jul, 2025 Editor assigned by journal 27 Jun, 2025 Editor invited by journal 19 Jun, 2025 Submission checks completed at journal 18 Jun, 2025 First submitted to journal 18 Jun, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6887876","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":488564419,"identity":"38b14e52-f5ae-468c-bb7d-ec920fb7bb5c","order_by":0,"name":"Liuyuan Li","email":"","orcid":"","institution":"China Medical University","correspondingAuthor":false,"prefix":"","firstName":"Liuyuan","middleName":"","lastName":"Li","suffix":""},{"id":488564420,"identity":"a4142386-1e35-49e8-9842-a547edb95fb5","order_by":1,"name":"Shuhua Zhang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA40lEQVRIie3PsYrCMBzH8ZSAXSJd/+WwzxAJFKGCr/IPB7qocweHTOno2sdQCs6Rwt1wce94j+DmcIdKcXIw1s0h3+0H/8/wJ8Tne8MiSve/Rz5Ootumz0lc6M9hmU9FrLoSbm36wWwtN6YrIQ1y6GsqxPfPFEieSRUejFMEJSKPdS9J7WIHxM6kYkt0EgpocKiZSJv+DgJdSwWMO0kPpDJSg6zKlpw7EMbqQBnL5QZaojoQCDUNVI4C7KIa4ddMaDZ3k0kdnf7++SWJisO2Oa6ywTq0bnIftt+9cO/z+Xy+B10BOG1GEe+mzKUAAAAASUVORK5CYII=","orcid":"","institution":"China Medical University","correspondingAuthor":true,"prefix":"","firstName":"Shuhua","middleName":"","lastName":"Zhang","suffix":""},{"id":488564421,"identity":"09c7aa84-964c-44bb-859e-00938ce2599a","order_by":2,"name":"wenhui Fan","email":"","orcid":"","institution":"Shenyang Normal University","correspondingAuthor":false,"prefix":"","firstName":"wenhui","middleName":"","lastName":"Fan","suffix":""}],"badges":[],"createdAt":"2025-06-13 11:38:20","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6887876/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6887876/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":87677984,"identity":"d90381c6-3906-41c5-b967-8db9056a5555","added_by":"auto","created_at":"2025-07-27 18:44:20","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":249949,"visible":true,"origin":"","legend":"\u003cp\u003eMachine Learning Model Processing of the Study\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6887876/v1/e1ff03fafe3897f3826495fe.png"},{"id":87678039,"identity":"325d5661-4caa-47dd-af39-511ed695ca38","added_by":"auto","created_at":"2025-07-27 18:52:20","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":297357,"visible":true,"origin":"","legend":"\u003cp\u003eComparison of ROC Curves Across Machine Learning Models for Adolescent Depression Identification in Training and Test Sets (Left: ROC curves on the training set; Right: ROC curves on the test set)\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6887876/v1/f47c371a4281e67b736c420d.png"},{"id":87677987,"identity":"2dc9759f-9ef8-4720-8ca3-180d281f86e9","added_by":"auto","created_at":"2025-07-27 18:44:20","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":252124,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP Value Analysis of the XGBoost Model in Adolescent Depression Prediction. Figure (A) shows the average importance of each feature based on the magnitude of the SHAP value through a bar chart, which helps to identify the variables that contribute the most to the model's prediction. Figure (B) is a honeycomb chart that shows the distribution of the importance of features in each prediction instance. Each data point represents a prediction instance, and its position on the x-axis indicates the relative impact of the feature on the model's prediction, while the color of the point indicates the specific level of the predictor variable according to the color reference bar on the right. The darker the color, the lower the level, and the brighter the color, the higher the level.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-6887876/v1/72a43b795cb9ac0ce9fb4afa.png"},{"id":87678238,"identity":"7c98f796-43a0-4e40-b932-aad735e1e6ea","added_by":"auto","created_at":"2025-07-27 19:00:20","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":155591,"visible":true,"origin":"","legend":"\u003cp\u003eAccumulated Local Effects (ALEs) of Key Variables\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-6887876/v1/415e1d137885ddaae8763145.png"},{"id":87677997,"identity":"13b11422-fb8e-4c25-ae81-bdd5801a3365","added_by":"auto","created_at":"2025-07-27 18:44:20","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":101578,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP Interaction Plot of Meaning in life and Age\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-6887876/v1/f4109c126985107e205b2175.png"},{"id":87678378,"identity":"727dbea9-0bdc-47f0-afea-022153fba409","added_by":"auto","created_at":"2025-07-27 19:08:25","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1804285,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6887876/v1/ba099fb2-25b9-498e-8ea3-d795f613e8cd.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Machine Learning-Based Prediction Model and Key Determinants of Adolescent Depression in China: A Multicenter Cross-Sectional Study","fulltext":[{"header":"1 Introduction","content":"\u003cp\u003eAdolescent depression, which is evolving into a global public health problem, has a profound impact on the physical and mental health and social functioning of individual adolescents (Thapar et al., \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). According to recent research, the pooled prevalence of depression in children and adolescents has reached 21.3% (Lu et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). A notable finding is that a particular study revealed a pooled prevalence of depression in Chinese children and adolescents as high as 26.17%, underscoring the severity of the problem in this demographic (Zhou et al., \u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). The consequences of depression for adolescents are extensive and far-reaching, manifesting in symptoms such as low mood, academic decline, social withdrawal, and an increased risk of self-injury and suicide. These effects can persist into adulthood, impacting career development and social adjustment (Shorey et al., \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Thapar et al., \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Zhou et al., \u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Prior research suggests that adolescence is a time of high prevalence of depression (Shorey et al., \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) and a critical period for identification and intervention (Hankin \u0026amp; Griffith, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Thapar et al., \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Consequently, there is a pressing need to develop a predictive model that incorporates a range of factors and accurately identifies early underlying factors that precipitate depression in adolescents.\u003c/p\u003e\u003cp\u003eThe development of adolescent depression constitutes a multifaceted process shaped by the interplay of macro-level sociocultural contexts, proximal environmental factors (e.g., family and school dynamics), and individual-level intrinsic traits and behavioral patterns (Courtney et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Existing studies have predominantly conceptualised these factors as independent variables and have investigated their associations with adolescent depression in isolation. However, this univariate analysis paradigm is challenging to employ comprehensively to elucidate the intricate mechanisms that underpin the occurrence and development of depression. Ecosystem theory emphasises the two-way interaction mechanism between the environment and the individual (Bronfenbrenner \u0026amp; Morris, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2006\u003c/span\u003e), and adolescent depression is essentially the result of the dynamic interaction between microsystems such as family, teachers, and peers, and the individual's contextual, psychological, and behavioural characteristics. Firstly, stressors from family, teachers, and peers can all have profound effects on adolescent psychology (Herres et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Yang et al., \u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Excessive parental discipline (Manuele et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), discordant family environments (Mastrotheodoros et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), teacher criticism (Guo et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), and peer rejection or isolation behaviours (Potter \u0026amp; Yoon, \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) have been shown to increase the risk of adolescent depression. Secondly, sociodemographic variables, understood as individual background characteristics, have been demonstrated to play a significant role in the development of adolescent depression. Research has indicated that gender (Hua et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Shorey et al., \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Thapar et al., \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), age (Hua et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Thapar et al., \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Zhou et al., \u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), and family economic status (Hua et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Lu et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Thapar et al., \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), family structure (Wen et al., \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), and parental education level (Xiang et al., \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) have been identified as significant contributors to the risk of adolescent depression. Thirdly, disparities in individual psychological traits modulate adolescents' vulnerability to depression in response to stress (Monroe \u0026amp; Simons, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e1991\u003c/span\u003e). Meaning in life (Baquero-Tom\u0026aacute;s et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Ward et al., \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), authenticity (Alchin et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Xia \u0026amp; Xu, \u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), and sense of control (O'Neill et al., 2023), and self-identity (Wong \u0026amp; Hamza, \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) as protective factors for depression have been shown to reduce psychopathological susceptibility to negative life events triggering depressive symptoms by shaping adaptive cognitive frameworks, moderating thresholds of emotional responses, optimizing the choice of coping strategies, and integrating social support resources, which together form the basis of adolescent psychological resilience. Fourthly, the role of lifestyle as an important behavioural characteristic in adolescent depression cannot be ignored. A substantial body of research has identified a nexus between unhealthy lifestyles and adolescent mental health, citing sleep disorders, inadequate exercise, poor dietary habits, sedentary behaviour, and excessive screen time as significant contributing factors (Garcia-Hermoso et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Kleppang et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Sampasa-Kanyinga et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Utilising the aforementioned theoretical framework, this study has developed a multilevel predictive model incorporating four-dimensional variables, including interpersonal stressors (family-teacher-peers), demographic background, psychological quality, and lifestyle habits. The objective of this model is to elucidate the cross-system mechanism of action of adolescent depression and to furnish an integrative perspective for the early identification and intervention of this condition.\u003c/p\u003e\u003cp\u003eTraditional statistical methods (e.g., multiple linear regression, structural equation modeling) rely on linear assumptions in multivariate modeling, which oversimplify complex nonlinear relationships into linear additive effects. This approach fails to identify curvilinear relationships, threshold effects, and higher-order interactions among variables. Furthermore, model complexity-sample size imbalance may induce parameter estimation bias and overfitting risks, while poor out-of-sample generalizability exacerbates the reproducibility crisis (Bzdok \u0026amp; Ioannidis, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Dwyer \u0026amp; Koutsouleris, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). In contrast, machine learning (ML) methods employ nonparametric modeling (e.g., XGBoost\u0026rsquo;s gradient-boosted decision trees) to autonomously capture nonlinear interaction patterns. Regularization techniques effectively control model complexity, while Shapley value decomposition, partial dependence plots (PDPs), and accumulated local effects (ALE) plots enable global interpretation of feature contributions, overcoming the limitations of traditional linear frameworks without compromising predictive performance (Hastie et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2009\u003c/span\u003e). Leveraging ML\u0026rsquo;s analytical and predictive power (Shatte et al., \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), this study integrates interpersonal stressors, sociodemographic backgrounds, psychological traits, and lifestyle factors into a unified framework to construct an ML prediction model capturing multidimensional interactions. This approach addresses three key limitations of traditional theory-driven research: 1. Reductionist tendencies in single-factor analysis; 2. Obscuration of complex interactions by linear assumptions; 3. Insufficient generalizability of predictive efficacy (Dwyer \u0026amp; Koutsouleris, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). SHAP interpretability methods\u0026mdash;grounded in cooperative game theory\u0026rsquo;s Shapley values\u0026mdash;precisely quantify each feature\u0026rsquo;s contribution to predictions, resolving the complexity of feature impact evaluation in ML models and limitations of traditional methods (Lundberg \u0026amp; Lee, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). PDPs visualize nonlinear relationships between individual predictors and depression risk (Hastie et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2009\u003c/span\u003e), while ALE plots illustrate cumulative contributions of variables to model predictions (Apley \u0026amp; Zhu, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). By employing multiple ML algorithms\u0026mdash;multilayer perceptron (MLP), XGBoost, logistic regression, random forest (RF), and support vector machine (SVM)\u0026mdash;this study aims to enhance the accuracy and efficacy of early adolescent depression detection, identify mechanisms of action of key factors, and establish a scientific foundation for targeted interventions.\u003c/p\u003e"},{"header":"2 Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Study Participants\u003c/h2\u003e\u003cp\u003eThis study employed cluster sampling to recruit adolescents from 29 general middle schools in Liaoning Province, China. A total of 8,137 self-report questionnaires were distributed. After rigorous quality control (excluding invalid questionnaires with duplicate responses, missing values, or extreme values), 7,169 valid samples were retained, yielding an effective response rate of 88.10%. The final sample comprised 3,402 males (47.45%) and 3,767 females (52.55%), with geographic distribution as follows: urban (49.48%), county-level towns (21.59%), and rural areas (28.93%). Participants\u0026rsquo; ages ranged from 11 to 20 years (14.98\u0026thinsp;\u0026plusmn;\u0026thinsp;1.58).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2 Measurement Tools\u003c/h2\u003e\u003cdiv id=\"Sec5\" class=\"Section3\"\u003e\u003ch2\u003e2.2.1 Depression Assessment\u003c/h2\u003e\u003cp\u003eDepressive symptoms were assessed using the Patient Health Questionnaire-9 (PHQ-9; Kroenke et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2001\u003c/span\u003e). This 9-item scale adopts a Likert-4 scoring system (0 = \"not at all\" to 3 = \"nearly every day\"), with total scores ranging from 0 to 27. A cutoff score of \u0026ge;\u0026thinsp;5 indicated depression. The Chinese version demonstrated good reliability and validity, with Cronbach\u0026rsquo;s α\u0026thinsp;=\u0026thinsp;0.945 in this study.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec6\" class=\"Section3\"\u003e\u003ch2\u003e2.2.2 Sleep\u003c/h2\u003e\u003cp\u003eThe Pittsburgh Sleep Quality Index (PSQI; Buysse et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e1989\u003c/span\u003e) evaluated sleep quality over the preceding month. This 18-item instrument assesses seven components: sleep quality, sleep latency, sleep duration, sleep efficiency, sleep disturbances, hypnotic medication use, and daytime dysfunction. Higher composite scores reflect poorer sleep quality. The Chinese version exhibited strong psychometric properties, achieving Cronbach\u0026rsquo;s α\u0026thinsp;=\u0026thinsp;0.887 in this study.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section3\"\u003e\u003ch2\u003e2.2.3 Diet\u003c/h2\u003e\u003cp\u003eDietary behaviors were assessed using the Dietary Behavior Questionnaire, which includes 10 items evaluating regularity of meals, binge eating, eating while studying or watching TV, and weekly frequencies of breakfast, fruit, vegetable, and milk consumption. Total scores range from 0 to 8, with scores\u0026thinsp;\u0026ge;\u0026thinsp;60% (\u0026ge;\u0026thinsp;4.8 points) classified as healthy. The Chinese version of this scale demonstrated good reliability (Cronbach\u0026rsquo;s α\u0026thinsp;=\u0026thinsp;0.752).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec8\" class=\"Section3\"\u003e\u003ch2\u003e2.2.4 Physical Activity\u003c/h2\u003e\u003cp\u003eModerate-to-vigorous physical activity (MVPA) was measured using the World Health Organization Health Behavior in School-aged Children (WHO-HBSC) questionnaire. Participants reported the number of days per week (Monday\u0026ndash;Friday and Saturday\u0026ndash;Sunday) with \u0026ge;\u0026thinsp;60 minutes of MVPA. Total weekly MVPA days ranged from 0 to 7.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section3\"\u003e\u003ch2\u003e2.2.5 Sedentary Time\u003c/h2\u003e\u003cp\u003eSedentary behavior was assessed using the International Physical Activity Questionnaire (IPAQ) long-form. Based on \u003cem\u003ethe Chinese Physical Activity Guidelines for Children and Adolescents\u003c/em\u003e, sedentary time\u0026thinsp;\u0026gt;\u0026thinsp;2 hours/day was classified as unhealthy.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec10\" class=\"Section3\"\u003e\u003ch2\u003e2.2.6 Screen Time\u003c/h2\u003e\u003cp\u003eScreen time was defined as the total daily duration spent on electronic devices (TV, video games, smartphones, computers). Per national guidelines, \u0026gt;\u0026thinsp;2 hours/day was categorized as unhealthy.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section3\"\u003e\u003ch2\u003e2.2.7 Perceived Stress Assessment\u003c/h2\u003e\u003cp\u003eInterpersonal stress was evaluated using the Middle School Student Stressor Scale (Zheng Quanquan \u0026amp; Chen Shulin, 1999), comprising 23 items across four dimensions: perceived teacher-related stress, perceived family-environment stress, perceived parental discipline stress, and perceived peer-related stress. Items were rated on a 5-point Likert scale. The Chinese version exhibited excellent reliability: total scale Cronbach\u0026rsquo;s α\u0026thinsp;=\u0026thinsp;0.918, with subscale α coefficients of 0.891 (perceived teacher-related stress), 0.725 (perceived family-environment stress), 0.713 (perceived parental discipline stress), and 0.824 (perceived peer-related stress).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section3\"\u003e\u003ch2\u003e2.2.8 Authenticity\u003c/h2\u003e\u003cp\u003eThe 12-item Authenticity Scale (Wood et al., \u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e2008\u003c/span\u003e) was administered using a 7-point Likert scale (1 = \"strongly disagree\" to 7 = \"strongly agree\"), with higher scores indicating greater authenticity. The Chinese version demonstrated good reliability (Cronbach\u0026rsquo;s α\u0026thinsp;=\u0026thinsp;0.877).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section3\"\u003e\u003ch2\u003e2.2.9 Meaning in Life\u003c/h2\u003e\u003cp\u003eMeaning in life was assessed via the PIL-SF (Schulenberg et al., \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2011\u003c/span\u003e), employing a 7-point scale (1 = \"strongly disagree\" to 7 = \"strongly agree\"). Higher scores reflect stronger life purpose and meaning. The Chinese adaptation showed excellent reliability (Cronbach\u0026rsquo;s α\u0026thinsp;=\u0026thinsp;0.946).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section3\"\u003e\u003ch2\u003e2.2.10 Sense of Control\u003c/h2\u003e\u003cp\u003eA 12-item Sense of Control Scale (Lachman \u0026amp; Weaver, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e1998\u003c/span\u003e) was used with a 7-point response format (1 = \"strongly disagree\" to 7 = \"strongly agree\"). Higher scores denote a greater sense of control. The Chinese version achieved good reliability (Cronbach\u0026rsquo;s α\u0026thinsp;=\u0026thinsp;0.840).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section3\"\u003e\u003ch2\u003e2.2.11 Self-Identity\u003c/h2\u003e\u003cp\u003eThe 12-item Self-Identity Scale (Kato, \u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e1983\u003c/span\u003e) utilized a 6-point Likert scale (1 = \"strongly disagree\" to 6 = \"strongly agree\"), with higher scores indicating stronger self-identity. The Chinese version exhibited acceptable reliability (Cronbach\u0026rsquo;s α\u0026thinsp;=\u0026thinsp;0.727).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section3\"\u003e\u003ch2\u003e2.2.12 Sociodemographic Variables\u003c/h2\u003e\u003cp\u003eEleven demographic variables were collected: grade, gender, age, annual family income, boarding status, single or multi-child household, home location, family structure, mother\u0026rsquo;s education levels, and father\u0026rsquo;s education levels.\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003e2.3 Statistical Analysis\u003c/h2\u003e\u003cp\u003eThis study integrated 24 initial predictors from multidimensional domains to construct adolescent depression prediction models. To enhance model efficiency and mitigate overfitting, LASSO regression (Tibshirani, \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e1996\u003c/span\u003e) was applied to compress high-dimensional variable space, followed by the Boruta algorithm (Kursa \u0026amp; Rudnicki, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2010\u003c/span\u003e) to identify statistically significant predictors. After rigorous screening, 21 critical variables were retained for model development. Five machine learning algorithms (MLP, XGBoost, logistic regression, RF, and SVM)\u0026mdash;were systematically evaluated. Data splitting adhered to stratified random sampling via the createDataPartition method (Kuhn, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2008\u003c/span\u003e), partitioning the dataset into training (80%) and independent test sets (20%). For XGBoost, SVM, and MLP, 25% of the training set was further allocated as a validation set (60-20-20 split) for hyperparameter tuning. To address class imbalance (depressed: non-depressed\u0026thinsp;\u0026asymp;\u0026thinsp;1:3), differential correction strategies were implemented: Inverse class frequency weighting (Hastie et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2009\u003c/span\u003e) was applied to loss functions for logistic regression, RF, XGBoost, and SVM. Given MLP\u0026rsquo;s inability to directly incorporate sample weights, a synthetic minority oversampling technique (Random Over-Sampling Examples, ROSE) with boundary-smoothed resampling based on probability distributions was applied to the training set (Menardi \u0026amp; Torelli, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Hyperparameter optimization employed grid search with 10-fold cross-validation. L1/L2 regularization mitigated overfitting in models, while tree-based algorithms (e.g., XGBoost, random forest) controlled complexity through depth limitation (max_depth), minimum leaf samples (min_samples_leaf), and feature subsampling (colsample_bytree), instance downsampling (subsample). Model performance was evaluated using AUC-ROC curves, with DeLong\u0026rsquo;s test (DeLong et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e1988\u003c/span\u003e) comparing ROC curve differences. Additional metrics included accuracy, balanced accuracy, sensitivity, specificity, positive/negative predictive values, and F1-score derived from confusion matrices. Interpretability analyses leveraged SHAP values (Lundberg \u0026amp; Lee, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2017\u003c/span\u003e) to visualize predictor contributions, supplemented by partial dependence plots (Hastie et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2009\u003c/span\u003e) and accumulated local effects (ALE) plots (Apley \u0026amp; Zhu, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) to decode nonlinear predictor effects. The specific flow chart is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. All analyses were conducted in R 4.3.3 and SPSS 27.0.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"3 Results","content":"\u003cdiv id=\"Sec19\" class=\"Section2\"\u003e\u003ch2\u003e3.1 Descriptive statistics\u003c/h2\u003e\u003cp\u003eThe Supplementary table 1 shows the results of the descriptive statistics and the basic characteristics of the study variables. The screening results of this study indicated that 36.01% of the adolescents exhibited depressive symptoms. A subsequent analysis revealed significant disparities between participants with and without depressive symptoms with respect to sociodemographic indicators and scale scores.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec20\" class=\"Section2\"\u003e\u003ch2\u003e3.2 Feature Selection\u003c/h2\u003e\u003cp\u003eA comparative analysis of LASSO regression and Boruta algorithm results (see Supplementary Fig.\u0026nbsp;1) was conducted to identify a shared subset of 21 key features that are strongly associated with adolescent depression for model construction. These include gender, age, home location, boarding status, family structure, single or multi-child household, mother's education level, father's education level, sleep quality, diet, weekend sedentary time, physical activity, weekly screen time, weekend screen time, perceived teacher-related stress, perceived family environment stress, perceived parental discipline stress, perceived peer-related stress, sense of control, authenticity, and meaning in life.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec21\" class=\"Section2\"\u003e\u003ch2\u003e3.3 Model Performance\u003c/h2\u003e\u003cp\u003eAs shown in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, this study comprehensively compared the predictive performance of five machine learning models: MLP, XGBoost, logistic regression, RF, and SVM. Results demonstrated that the XGBoost model achieved a significantly superior area under the receiver operating characteristic curve (AUC) compared to other models (see Fig.\u0026nbsp;2). On the test set, the XGBoost model exhibited a classification accuracy of 0.8238 (95% CI: 0.812\u0026ndash;0.835), significantly exceeding the no-information rate (0.656, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Cohen\u0026rsquo;s Kappa coefficient of 0.629 indicated substantial agreement between predicted and actual classifications (Landis \u0026amp; Koch, 1977). For depression identification, the model achieved a sensitivity of 0.855, specificity of 0.807, positive predictive value (PPV) of 0.699, negative predictive value (NPV) of 0.914, balanced accuracy of 0.831, and F1-score of 0.769. DeLong\u0026rsquo;s test confirmed the statistical significance of XGBoost\u0026rsquo;s AUC superiority over other models (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05), establishing its optimal discriminative power and clinical utility in adolescent depression prediction.\u003c/p\u003e\u003cp\u003e\u003cimg src=\"data:image/png;base64,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\"\u003e\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n \u003cdiv align=\"left\" class=\"colspec\"\u003e3.4 Variable Importance\u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec22\" class=\"Section2\"\u003e\n \u003cp\u003eThis study employed SHAP analysis to interpret the XGBoost model, generating SHAP summary plots (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003eA) and beeswarm plots (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003eB). The top five predictors of adolescent depression were identified as sleep quality, authenticity, meaning in life, sense of control, and perceived peer-related stress. Specifically, sleep disturbances and perceived peer-related stress exhibited positive correlations with depression risk, whereas authenticity and sense of control showed protective effects. Notably, meaning in life demonstrated a nonlinear association: both low and high levels served as protective factors, while moderate levels increased depression risk. Additional risk factors included female gender, rural/county residence, older age, non-only-child status, higher paternal education, weekly screen time\u0026thinsp;\u0026gt;\u0026thinsp;2 hours, weekend sedentary time\u0026thinsp;\u0026gt;\u0026thinsp;2 hours, perceived teacher-related stress, and perceived teacher-related stress. Protective factors encompassed daily moderate-intensity physical activity\u0026thinsp;\u0026ge;\u0026thinsp;1 hour.\u003c/p\u003e\n \u003cp\u003eTo further explore the mechanisms of the effects of the key variables, the present study plotted a partial dependence plot (PDP, see Supplementary Fig. 2) and a cumulative local effect plot (ALE, see Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e). Initially, the results of both the PDP and ALE plots for sleep demonstrated that the predicted probability of depression in adolescents increased significantly as sleep disturbance scores increased, suggesting that good sleep quality may play a protective role in the development of depression. Secondly, the PDP and ALE plots for Authenticity exhibited a tendency towards a negative correlation between authenticity and the predicted probability of depression. Additionally, the PDP and ALE plots for sense of control demonstrated that after scores surpassed 40 (scale scores of 45%-50%), the level of sense of control exhibited an essentially negative correlation with the probability of depression. This finding suggests that higher levels of authenticity and sense of control may contribute to a reduced risk of depression. The PDP for meaning in life exhibited a positive association with depression below 8 points (scale scores of 25%-50%) and a negative association above 8 points, a threshold that may reflect dominant trends at the group level but is susceptible to affecting confounding variables such as age, gender, etc. The ALE controlled for feature covariance by adjusting for the conditional distributions, and the results showed a 12-point (scale scores of 25%-50%) threshold. The results of the ALE demonstrated that a score of 12 points was identified as the critical value. Scores below 12 points were found to be positively correlated, while scores above 12 points were negatively correlated. This finding was more closely aligned with the local causal effect, suggesting that the actual threshold may be higher and necessitate dynamic evaluation based on individual characteristics. The PDP and ALE plots of perceived peer-related stress demonstrated that perceived peer-related stress exhibited a positive association with the probability of depression when scores were below approximately 13 (scale scores of 25%-50%). The association leveled off after exceeding the threshold.\u003c/p\u003e\n \u003cp\u003eIn order to further explore the nonlinear relationship of meaning in life in different age groups of adolescents, the present study conducted a SHAP interaction analysis between age and meaning in life (see Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e). The findings revealed a robust positive interaction between low values of meaning in life and age among the high age group (17\u0026ndash;18 years old), along with a significant negative interaction between high values of meaning in life and age. These results suggest that low levels of meaning in life are associated with an elevated risk of depression, while high levels of meaning in life offer a protective effect against depression among adolescents in the high age group. In the middle age group (14\u0026ndash;16 years), the interaction between meaning in life and age was relatively weak, with both higher and lower levels of meaning in life potentially exacerbating the risk of depression, and moderate levels of meaning in life mitigating the risk of depression. In the lower age group (12\u0026ndash;13 years), a robust positive interaction emerged between moderate values of meaning in life (15\u0026ndash;18 points) and age, suggesting that moderate levels of meaning in life amplify the risk of depression in lower-age adolescents. Conversely, a significant negative interaction was observed between low (10\u0026ndash;11 points) and higher (19\u0026ndash;20 points) values of meaning in life and age, indicating that lower levels of meaning in life may mitigate the risk of depression in these younger adolescents.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"4 Discussion","content":"\u003cp\u003eThe following discussion will elaborate on the findings of this study, which compared five machine learning models commonly used to predict depression in Chinese adolescents. The results showed that the XGBoost model exhibited good performance with an accuracy of 0.824 (balanced precision\u0026thinsp;=\u0026thinsp;0.831), an AUC of 0.912, a sensitivity of 0.855, and a specificity of 0.807. These findings are consistent with those reported in Mardini et al.'s (Mardini et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) cross-national study (N\u0026thinsp;=\u0026thinsp;15,632) and Kuang et al.'s (Kuang et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) Chinese mega-cohort study (N\u0026thinsp;=\u0026thinsp;583,405) to form a methodological validation, which collectively confirms the superiority of XGBoost in predicting depression in adolescents. As one of the most commonly used machine learning models for classification, the XGBoost model, through regularization and the DART tree integration strategies, can effectively prevent overfitting to noisy data and tends to have better performance when dealing with tabular datasets (Shwartz-Ziv \u0026amp; Armon, \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThis study also examined the importance of each predictor. The findings of this study indicated that sleep quality, authenticity, sense of control, meaning in life, and perceived peer-related stressors were significant in predicting adolescent depression. Characteristic importance results indicated that sleep quality was the most significant factor contributing to depression, and as sleep quality improved, the risk of adolescent depression decreased. This finding aligns with the results of a prior machine learning study by Kuang et al. (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), which identified sleep as the most significant segmented characteristic in predicting depression. Additionally, the machine learning study by Olfati et al. (\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) corroborated the importance of sleep quality in predicting the severity of depressive symptoms. Liang et al. (\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) found through a meta-analysis that Chinese adolescents' sleep disorders had a comorbid prevalence of 26%, and a study by Wang et al. (\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) noted that 70% of high school students did not get enough sleep during the school year. Sleep disorders have been demonstrated to be a risk factor for adolescent mental health (Palagini et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Palmer et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Wang et al., \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), exacerbating adolescent mental health problems and suicide risk (Lam \u0026amp; Lam, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Woodfield et al., \u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Consequently, the present study endorses the incorporation of sleep interventions as a strategy for the prevention of adolescent depression (Freeman et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Scott et al., \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). It is recommended that educational institutions incorporate sleep education into their health education curricula, systematically assess the quality of sleep and psychological well-being of students, and devise customized psycho-sleep intervention programs based on the assessment outcomes.\u003c/p\u003e\u003cp\u003eResearch has identified authenticity as a significant protective factor against depression in adolescents. Adolescence is a period of exploration and the formation of authenticity, which serves as a foundation for self-perception (Alchin et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Extensive research has indicated a negative correlation between authenticity and depression (Assor et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Ionescu et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). According to the stress trait theory of depression (Monroe \u0026amp; Simons, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e1991\u003c/span\u003e), the formation of authenticity fosters coherence in the integration of internalized self-narratives, thereby enhancing adolescents' ability to regulate their emotions and improve their psychological well-being (Xia \u0026amp; Xu, \u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Conversely, instability in the core self-concept has been associated with feelings of division in social situations and significant fluctuations in emotional regulation, which can increase the risk of depression (Alchin et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThe present study found that the effect of meaning in life as a significant predictor of depression in adolescents showed significant age-specific effects. In the 17\u0026ndash;18 age group, a low meaning in life was identified as a risk factor for depression, while a high meaning in life exhibited a protective effect, demonstrating psychological resilience. In the 14\u0026ndash;16 age group, a moderate meaning in life served as a buffering factor against the risk of depression, with higher or lower levels likely to exacerbate the risk. In the 12\u0026ndash;13 age group, an inverse effect was observed, with a moderate meaning in life exacerbating depression, and higher or lower level decreasing. This nonlinear relationship is rooted in the dynamic developmental processes of the cognitive-affective system. From a neurodevelopmental perspective, the prefrontal lobe functions as a higher-order cognitive center, and the degree of its functional integration with the limbic system determines the psychological effects of meaning in life (Andrews et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Prefrontal-limbic integration is incomplete in 12- to 13-year-old adolescents, resulting in an inability to relate abstract meanings to their personal experiences, despite being able to recapitulate social norms (e.g., \"it is important to study hard\") (Wehmeyer et al., \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Individuals who possess a moderate meaning in life may exhibit superficial compliance with adult expectations, while concurrently experiencing internal conflict regarding their values. This proclivity for cognitive dissonance, exacerbated by rumination, has been observed to engender a heightened mental load (Chu \u0026amp; Fung, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). The localization of meaning, whether elevated or diminished, has been demonstrated to facilitate adaptation by attenuating cognitive demands, thereby mitigating the risk of depression. With age, prefrontal myelination fosters cognitive-emotional integration, enabling adolescents aged 17 to 18 to translate abstract meaning into personal narratives. At this stage, high sense makers reconstruct stressful events as growth narratives through dialectical thinking (Dulaney et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2018\u003c/span\u003e), while low sense makers are more prone to existential anxiety under multiple stressors due to a lack of goal orientation. 14\u0026ndash;16 years of age serve as a transition period where the development of metacognitive skills supports the construction of meaning through reflection and adjustment through trial and error, and medium sense makers are able to maintain both a certain set of core values, but also have the flexibility to adjust meaning frameworks, whereas higher or lower meaning orientations can hinder role exploration in Erikson's theory and exacerbate depression through social comparison. This developmental trajectory reveals that adolescents are experiencing the \u0026ldquo;growing pains\u0026rdquo; of the cognitive revolution. Early risk for a moderate meaning in life stems from neurodevelopmental lags, in which any mental processes requiring higher-order integration may temporarily become cognitively burdensome when the emotional intensity of the limbic system outstrips the ability of the prefrontal lobes to regulate it. With the development of prefrontal-limbic functioning, the dichotomies of surface cognition and deep beliefs, social attachment and self-identity are gradually integrated, and the meaning in life is eventually transformed from a developmental risk factor to a kernel of psychological resilience. Therefore, intervention strategies need to match the developmental stage of adolescents: avoiding abstract questioning of meaning and guiding meaning exploration through figurative activities in 12\u0026ndash;16-year-olds, and strengthening the reconstruction of stressful events and dialectical thinking training in 17\u0026ndash;18-year-olds.\u003c/p\u003e\u003cp\u003eThe fourth significant predictor was sense of control, the level of which demonstrated a significant negative correlation with adolescent depression risk. This finding is consistent with the theory of learned helplessness (Seligman, \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e1975\u003c/span\u003e), which posits that individuals who repeatedly encounter uncontrollable negative events may develop generalized attributional patterns of hopelessness, potentially resulting in the onset of depressive symptoms. Empirical studies have further confirmed that sense of control, an important cornerstone of mental health, exerts a moderating effect on depressive risk primarily through the establishment of cognitive-emotional pathways of behavioral regulation (Ge et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Stolz et al., \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Ge et al.'s study (2024) found that sense of control, by enhancing emotionally controllable beliefs and adaptive regulatory strategies, was able to reduce the risk of depression in adolescents. Research on neural mechanisms has identified a protective effect of perceptual control through the cognitive-emotional integration pathway in the prefrontal-limbic system. Individuals with strong control beliefs have been observed to employ adaptive strategies, such as cognitive reappraisal, in response to stress, and experience heightened positive emotions (Stolz et al., \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eChronic interpersonal stress has emerged as a robust predictor of adolescent depression (Losiewicz et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Among various stressors, perceived peer stress and perceived teacher stress exerted the strongest influences in this study. Drawing from the interpersonal theory of depression, negative social interactions\u0026mdash;such as peer rejection and critical teacher feedback\u0026mdash;elevate depressive risk by altering self-perception and emotional regulation processes (Coyne, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e1976\u003c/span\u003e). Adolescents undergoing psychological individuation prioritize peer and teacher relationships as central to their social lives during this developmental stage. The intensifying importance of peer validation in adolescence makes negative peer experiences (e.g., exclusion, bullying) particularly traumatic (Kim, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Potter \u0026amp; Yoon, \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Longitudinal evidence indicates that adolescents reporting high perceived peer-related stress demonstrate increased odds of developing persistent depressive symptoms compared to peers with low peer stress (Agoston \u0026amp; Rudolph, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Liao et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Perceived teacher-related stressors also play a critical role. Negative teacher-student interactions (e.g., excessive criticism, biased evaluations) trigger stronger emotional reactivity in adolescents than parent-child conflicts or peer rejection, with significant longitudinal associations to depressive symptom persistence (Herres et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). Additionally, academic pressures\u0026mdash;including unrealistic performance expectations, frequent assessments, and overloaded assignments\u0026mdash;exacerbate perceived teacher-related stress, leading to an increase in depression risk among Chinese adolescents (Jiang et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). These findings underscore the need for school-based interventions targeting: (1) peer relationship skills training to mitigate rejection sensitivity, (2) teacher-student communication workshops to reduce negative feedback, and (3) dynamic stress monitoring systems to identify at-risk individuals.\u003c/p\u003e\u003cp\u003eThe study also found that engaging in at least one hour of moderate-intensity physical activity daily, limiting screen time to \u0026le;\u0026thinsp;2 hours on weekdays, and restricting weekend sedentary time to \u0026le;\u0026thinsp;2 hours were significantly associated with a reduced risk of adolescent depression. In line with prior research findings (Rodriguez-Ayllon et al., \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Li et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), the present study corroborated that adolescents who adhered strictly to physical activity recommendations exhibited a reduced risk of depression. A substantial body of research has demonstrated the efficacy of physical activity interventions in enhancing adolescents' mental well-being (Recchia et al., \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Tomkinson et al., \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Consequently, this study proposes the optimization of the design of physical education programs on campuses. Furthermore, the study identified demographic variables such as female gender, residing in rural or county areas, advanced age, not being an only child, and elevated paternal education levels as risk factors for depression among adolescents. This finding underscores the necessity to prioritize this population.\u003c/p\u003e"},{"header":"5 Limitations","content":"\u003cp\u003eThis study breaks through the traditional unidimensional analysis framework and systematically integrates environmental factors such as family, teachers, and peers with individual background characteristics, psychological characteristics, and behavioral characteristics to construct a multidimensional depression prediction system for Chinese adolescents, but some limitations still exist. First, the samples were all from adolescents in Northeast China, which may limit the generalizability of the findings. Second, the cross-sectional design made it difficult to distinguish the temporal relationship between predictors and depressive symptoms, e.g., sleep disturbance may be both a precursor symptom of depression and an outcome variable due to environmental stress (Palagini et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), which needs to be further validated through follow-up studies. Third, in the current depression assessment system, the self-rating scale (PHQ \u0026minus;\u0026thinsp;9) is mainly relied upon to determine the depression status of adolescents and is not combined with clinical diagnosis and neurobiological indicators (e.g., fMRI, cortisol levels) for a comprehensive assessment.\u003c/p\u003e"},{"header":"6 Conclusion","content":"\u003cp\u003eThis study compared the effectiveness of five machine learning models in predicting depression in Chinese adolescents. The results show that XGBoost has optimal efficacy in predicting depression in Chinese adolescents, and the key predictors identified include sleep quality, authenticity, sense of control, meaning in life, perceived peer-source pressure, and perceived teacher-source pressure. Among the key predictors, sleep quality served as the most important protective factor, suggesting that sleep quality monitoring should be incorporated into the campus routine mental health assessment system. Authenticity and sense of control reduce the risk of depression by strengthening the internal stability and external adaptability of psychological resilience, suggesting that schools should increase the content of related psychological education; the age-specificity of the meaning in life suggests a differentiated intervention strategy: for the group of 12-16-year-olds, it is recommended to avoid early abstract questioning and integrate meaning construction into practical experience; for the group of 17-18-year-olds, it is recommended that they should be guided to reconstruct the stressful events into self-growth narratives. In addition, perceived peer-related stress and perceived teacher-related pressure were found to be risk factors for depression, suggesting the need to build a dynamic early warning and support network for campus stress. Future research could further integrate the ecological transient assessment data to reveal the cross-level interaction mechanisms of the predictors.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003ea. Ethics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll participants were fully informed about the purpose, procedures, and potential risks of the study. Written informed consent was obtained from each participant prior to their involvement. For participants under the age of 18, consent was obtained from their parents or legal guardians. The study was conducted in accordance with the ethical standards of the relevant institutional review board.. All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The study was approved by the Bioethics Committee of China Medical University, as indicated by the ethics approval number 2024/169.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eb. Consent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot Applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ec. Availability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe current study\u0026rsquo;s findings are available from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ed. Competing interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll participants provided written informed consent to participate in this study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ee. Funding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors did not receive support from any organization for the submitted work.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ef. Authors\u0026apos; contribution\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eLiuyuan Li played a lead role in conceptualization, data collection, formal analysis, and writing\u0026ndash;original draft, and played a supporting role in writing\u0026ndash;review and editing. Shuhua Zhang played a supporting role in conceptualization, and a lead role in writing\u0026ndash;review and editing. Wenhui Fan played a supporting role in literature review. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eg. Acknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe are grateful to the participating schools that helped collect the data. We also thank parents for granting permission and students for their participation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eh. Authors\u0026apos; information (optional)\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFIRST AUTHOR: Liuyuan Li, Affiliation: College of Medical Humanities, China Medical University, Shenyang, China; Address: China Medical University, 78 Puhe Road, Shenbei New District, Shenyang City, Liaoning Province, China; ORCID：0009-0009-0190-9367; Email: [email protected];\u003c/p\u003e\n\u003cp\u003eTHIRD AUTHOR: Wenhui Fan, Affiliation: College of Educational Science, Shenyang Normal University, Shenyang, China; Address: Shenyang Normal University, 253 Huanghe North Street, Huanggu District, Shenyang City, Liaoning Province, China;\u0026nbsp;Email:\u0026nbsp;[email protected]；\u003c/p\u003e\n\u003cp\u003eSECOND AUTHOR: Shuhua Zhang, Affiliation: College of Medical Humanities, China Medical University, Shenyang, China; Address: China Medical University, 78 Puhe Road, Shenbei New District, Shenyang City, Liaoning Province, China. Email: [email protected];\u003c/p\u003e\n\u003cp\u003eCorrespondence concerning this article should be addressed to Shuhua Zhang, Email: [email protected]; Tel: 18900916785.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAgoston, A. M., \u0026amp; Rudolph, K. D. (2016). Interactive contributions of cumulative peer stress and executive function deficits to depression in early adolescence. \u003cem\u003eThe Journal of Early Adolescence\u003c/em\u003e,\u003cem\u003e 36\u003c/em\u003e(8), 1070-1094. https://doi.org/10.1177/0272431615593176\u003c/li\u003e\n\u003cli\u003eAlchin, C. E., Machin, T. M., Martin, N., \u0026amp; Burton, L. J. (2024). Authenticity and Inauthenticity in Adolescents: A Scoping Review. \u003cem\u003eAdolescent Research Review\u003c/em\u003e,\u003cem\u003e 9\u003c/em\u003e(2), 279-315. https://doi.org/10.1007/s40894-023-00218-8\u003c/li\u003e\n\u003cli\u003eAndrews, J. L., Ahmed, S. P., \u0026amp; Blakemore, S.-J. (2021). Navigating the Social Environment in Adolescence: The Role of Social Brain Development. \u003cem\u003eBiological Psychiatry\u003c/em\u003e, \u003cem\u003e89\u003c/em\u003e(2), 109\u0026ndash;118. https://doi.org/10.1016/j.biopsych.2020.09.012 \u003c/li\u003e\n\u003cli\u003eApley, D. W., \u0026amp; Zhu, J. (2020). Visualizing the effects of predictor variables in black box supervised learning models. \u003cem\u003eJournal of the Royal Statistical Society Series B: Statistical Methodology\u003c/em\u003e,\u003cem\u003e 82\u003c/em\u003e(4), 1059-1086. https://doi.org/10.1111/rssb.12377 \u003c/li\u003e\n\u003cli\u003eAssor, A., Benita, M., Shi, Y., Goren, R., Yitshaki, N., \u0026amp; Wang, Q. (2021). The Authentic Inner Compass as a Well-Being Resource: Predictive Effects on Vitality, and Relations with Self-Esteem, Depression and Behavioral Self-realization. \u003cem\u003eJOURNAL OF HAPPINESS STUDIES\u003c/em\u003e,\u003cem\u003e 22\u003c/em\u003e(8), 3435-3455. https://doi.org/10.1007/s10902-021-00373-6 \u003c/li\u003e\n\u003cli\u003eBaquero-Tom\u0026aacute;s, M., Grau, M. D., Moliner, A.-R., \u0026amp; Sanchis-Sanchis, A. (2023). Meaning in life as a protective factor against depression. \u003cem\u003eFRONTIERS IN PSYCHOLOGY\u003c/em\u003e,\u003cem\u003e 14\u003c/em\u003e, 1180082. https://doi.org/10.3389/fpsyg.2023.1180082 \u003c/li\u003e\n\u003cli\u003eBronfenbrenner, U., \u0026amp; Morris, P. A. (2006). The bioecological model of human development. In \u003cem\u003eHandbook of child psychology: Theoretical models of human development\u003c/em\u003e (6th ed., Vol. 1, pp. 793\u0026ndash;828). John Wiley \u0026amp; Sons.\u003c/li\u003e\n\u003cli\u003eBuysse, D. J., Reynolds, C. F., Monk, T. H., Berman, S. R., \u0026amp; Kupfer, D. J. (1989). The pittsburgh sleep quality index: A new instrument for psychiatric practice and research. \u003cem\u003ePsychiatry Research\u003c/em\u003e,\u003cem\u003e 28\u003c/em\u003e(2), 193-213. https://doi.org/10.1016/0165-1781(89)90047-4 \u003c/li\u003e\n\u003cli\u003eBzdok, D., \u0026amp; Ioannidis, J. P. A. (2019). Exploration, inference, and prediction in neuroscience and biomedicine. \u003cem\u003eTrends in Neurosciences\u003c/em\u003e, \u003cem\u003e42\u003c/em\u003e(4), 251\u0026ndash;262. https://doi.org/10.1016/j.tins.2019.02.001\u003c/li\u003e\n\u003cli\u003eChu, S. T.-W., \u0026amp; Fung, H. H.-L. (2021). Is the search for meaning related to the presence of meaning? Moderators of the longitudinal relationship. \u003cem\u003eJOURNAL OF HAPPINESS STUDIES\u003c/em\u003e,\u003cem\u003e 22\u003c/em\u003e(1), 127-145. https://doi.org/10.1007/s10902-020-00222-y \u003c/li\u003e\n\u003cli\u003eCourtney, D. B., Watson, P., Krause, K. R., Chan, B. W. C., Bennett, K., Gunlicks-Stoessel, M., Rodak, T., Neprily, K., Zentner, T., \u0026amp; Szatmari, P. (2022). Predictors, moderators, and mediators associated with treatment outcome in randomized clinical trials among adolescents with depression: A scoping review. \u003cem\u003eJAMA Network Open\u003c/em\u003e,\u003cem\u003e 5\u003c/em\u003e(2), e2146331. https://doi.org/10.1001/jamanetworkopen.2021.46331 \u003c/li\u003e\n\u003cli\u003eCoyne, J. C. (1976). Toward an interactional description of depression. \u003cem\u003ePsychiatry\u003c/em\u003e. \u003c/li\u003e\n\u003cli\u003eDeLong, E. R., DeLong, D. M., \u0026amp; Clarke-Pearson, D. L. (1988). Comparing the areas under two or more correlated receiver operating characteristic curves: A nonparametric approach. \u003cem\u003eBiometrics\u003c/em\u003e,\u003cem\u003e 44\u003c/em\u003e(3), 837-845. \u003c/li\u003e\n\u003cli\u003eDulaney, E. S., Graupmann, V., Grant, K. E., Adam, E. K., \u0026amp; Chen, E. (2018). Taking on the stress-depression link: Meaning as a resource in adolescence. \u003cem\u003eJOURNAL OF ADOLESCENCE\u003c/em\u003e,\u003cem\u003e 65\u003c/em\u003e, 39-49. https://doi.org/10.1016/j.adolescence.2018.02.011 \u003c/li\u003e\n\u003cli\u003eDwyer, D., \u0026amp; Koutsouleris, N. (2022). Annual research review: Translational machine learning for child and adolescent psychiatry. \u003cem\u003eJOURNAL OF CHILD PSYCHOLOGY AND PSYCHIATRY\u003c/em\u003e,\u003cem\u003e 63\u003c/em\u003e(4), 421-443. https://doi.org/10.1111/jcpp.13545 \u003c/li\u003e\n\u003cli\u003eFreeman, D., Sheaves, B., Waite, F., Harvey, A. G., \u0026amp; Harrison, P. J. (2020). Sleep disturbance and psychiatric disorders. \u003cem\u003eThe Lancet Psychiatry\u003c/em\u003e,\u003cem\u003e 7\u003c/em\u003e(7), 628-637. https://doi.org/10.1016/S2215-0366(20)30136-X \u003c/li\u003e\n\u003cli\u003eGarcia-Hermoso, A., Ezzatvar, Y., Ramirez-Velez, R., Lopez-Gil, J. F., \u0026amp; Izquierdo, M. (2022). Trajectories of 24-h movement guidelines from middle adolescence to adulthood on depression and suicidal ideation: a 22-year follow-up study. \u003cem\u003eINTERNATIONAL JOURNAL OF BEHAVIORAL NUTRITION AND PHYSICAL ACTIVITY\u003c/em\u003e,\u003cem\u003e 19\u003c/em\u003e(1), 135. https://doi.org/10.1186/s12966-022-01367-0 \u003c/li\u003e\n\u003cli\u003eGe, W., Wang, H., Chen, J., Yang, X., \u0026amp; Wang, Y. (2024). How general control beliefs relate to depressive symptoms in adolescents: The role of emotion controllability beliefs and emotion regulation. \u003cem\u003eCURRENT PSYCHOLOGY\u003c/em\u003e,\u003cem\u003e 43\u003c/em\u003e(35), 28531-28543. https://doi.org/10.1007/s12144-024-06489-1 \u003c/li\u003e\n\u003cli\u003eGuo, N., Huebner, E. S., Gong, X., \u0026amp; Tian, L. (2024). Psychological maltreatment by teachers and peer victimization in Chinese youth: Depression and aggression as mediators. \u003cem\u003eChild Abuse and Neglect\u003c/em\u003e,\u003cem\u003e 153\u003c/em\u003e, 106809. https://doi.org/10.1016/j.chiabu.2024.106809 \u003c/li\u003e\n\u003cli\u003eHankin, B. L., \u0026amp; Griffith, J. M. (2023). What do we know about depression among youth and how can we make progress toward improved understanding and reducing distress? A new hope. \u003cem\u003eClinical Child and Family Psychology Review\u003c/em\u003e,\u003cem\u003e 26\u003c/em\u003e(4), 919-942. https://doi.org/10.1007/s10567-023-00437-4 \u003c/li\u003e\n\u003cli\u003eHastie, T., Tibshirani, R., \u0026amp; Friedman, J. (2009). \u003cem\u003eThe elements of statistical learning\u003c/em\u003e. Springer. \u003c/li\u003e\n\u003cli\u003eHerres, J., Ewing, E. S. K., \u0026amp; Kobak, R. (2016). Emotional reactivity to negative adult and peer events and the maintenance of adolescent depressive symptoms: A daily diary design. \u003cem\u003eJOURNAL OF ABNORMAL CHILD PSYCHOLOGY\u003c/em\u003e,\u003cem\u003e 44\u003c/em\u003e(3), 471-481. https://doi.org/10.1007/s10802-015-0043-6 \u003c/li\u003e\n\u003cli\u003eHua, Z., Wang, S., \u0026amp; Yuan, X. (2024). Trends in age-standardized incidence rates of depression in adolescents aged 10\u0026ndash;24 in 204 countries and regions from 1990 to 2019. \u003cem\u003eJOURNAL OF AFFECTIVE DISORDERS\u003c/em\u003e,\u003cem\u003e 350\u003c/em\u003e, 831-837. https://doi.org/10.1016/j.jad.2024.01.009 \u003c/li\u003e\n\u003cli\u003eIonescu, D., Burtăverde, V., Avram, E., \u0026amp; Ene, C. (2022). Authenticity, Volition, and Motivational Persistence Predicting Well-being: a Self-determination Theoretical Perspective. \u003cem\u003eTrends in Psychology\u003c/em\u003e,\u003cem\u003e 30\u003c/em\u003e(3), 468-479. https://doi.org/10.1007/s43076-021-00138-3 \u003c/li\u003e\n\u003cli\u003eJiang, S., Ren, Q., Jiang, C., \u0026amp; Wang, L. (2021). Academic stress and depression of Chinese adolescents in junior high schools: Moderated mediation model of school burnout and self-esteem. \u003cem\u003eJOURNAL OF AFFECTIVE DISORDERS\u003c/em\u003e,\u003cem\u003e 295\u003c/em\u003e, 384-389. https://doi.org/10.1016/j.jad.2021.08.085 \u003c/li\u003e\n\u003cli\u003eKim, K. M. (2021). What makes adolescents psychologically distressed? Life events as risk factors for depression and suicide. \u003cem\u003eEuropean Child and Adolescent Psychiatry\u003c/em\u003e,\u003cem\u003e 30\u003c/em\u003e(3), 359-367. https://doi.org/10.1007/s00787-020-01520-9 \u003c/li\u003e\n\u003cli\u003eKleppang, A. L., Vettore, M. V., Hartz, I., Haugland, S. H., \u0026amp; Stea, T. H. (2023). Patterns of unhealthy behaviours during adolescence and subsequent anxiety and depression in adulthood: a prospective register linkage study of the HUNT survey and health registries. \u003cem\u003eINTERNATIONAL JOURNAL OF BEHAVIORAL NUTRITION AND PHYSICAL ACTIVITY\u003c/em\u003e,\u003cem\u003e 20\u003c/em\u003e(1), 6. https://doi.org/10.1186/s12966-023-01408-2 \u003c/li\u003e\n\u003cli\u003eKroenke, K., Spitzer, R. L., \u0026amp; Williams, J. B. W. (2001). The PHQ-9. \u003cem\u003eJournal of General Internal Medicine\u003c/em\u003e,\u003cem\u003e 16\u003c/em\u003e(9), 606-613. https://doi.org/10.1046/j.1525-1497.2001.016009606.x \u003c/li\u003e\n\u003cli\u003eKuang, Y., Liao, X., Jiang, Z., Gu, Y., Liu, B., Tan, C., Zhang, W., \u0026amp; Li, K. (2025). Federated learning-based prediction of depression among adolescents across multiple districts in China. \u003cem\u003eJOURNAL OF AFFECTIVE DISORDERS\u003c/em\u003e,\u003cem\u003e 369\u003c/em\u003e, 625-632. https://doi.org/10.1016/j.jad.2024.10.027 \u003c/li\u003e\n\u003cli\u003eKuhn, M. (2008). Building predictive models in R using the caret package. \u003cem\u003eJOURNAL OF STATISTICAL SOFTWARE\u003c/em\u003e,\u003cem\u003e 28\u003c/em\u003e, 1-26. https://doi.org/10.18637/jss.v028.i05 \u003c/li\u003e\n\u003cli\u003eKursa, M. B., \u0026amp; Rudnicki, W. R. (2010). Feature selection with the boruta package. \u003cem\u003eJOURNAL OF STATISTICAL SOFTWARE\u003c/em\u003e,\u003cem\u003e 36\u003c/em\u003e, 1-13. https://doi.org/10.18637/jss.v036.i11 \u003c/li\u003e\n\u003cli\u003eLachman, M. E., \u0026amp; Weaver, S. L. (1998). The Sense of Control as a Moderator of Social Class Differences in Health and Well-Being: Journal of Personality \u0026amp; Social Psychology. \u003cem\u003eJournal of Personality \u0026amp; Social Psychology\u003c/em\u003e,\u003cem\u003e 74\u003c/em\u003e(3), 763-773. https://doi.org/10.1037/0022-3514.74.3.763 \u003c/li\u003e\n\u003cli\u003eLam, L. T., \u0026amp; Lam, M. K. (2021). Sleep Disorders in Early Childhood and the Development of Mental Health Problems in Adolescents: A Systematic Review of Longitudinal and Prospective Studies. \u003cem\u003eINTERNATIONAL JOURNAL OF ENVIRONMENTAL RESEARCH AND PUBLIC HEALTH\u003c/em\u003e,\u003cem\u003e 18\u003c/em\u003e(22), 11782. https://doi.org/10.3390/ijerph182211782\u003c/li\u003e\n\u003cli\u003eLi, L., Guo, D., Shi, C., \u0026amp; Zheng, Y. (2025). The predictive role of sedentary behavior and physical activity on adolescent depressive symptoms: A machine learning approach. \u003cem\u003eJournal of Affective Disorders\u003c/em\u003e, \u003cem\u003e378\u003c/em\u003e, 81\u0026ndash;89. https://doi.org/10.1016/j.jad.2025.02.085\u003c/li\u003e\n\u003cli\u003eLiang, M., Guo, L., Huo, J., \u0026amp; Zhou, G. (2021). Prevalence of sleep disturbances in Chinese adolescents: A systematic review and meta-analysis. \u003cem\u003ePLOS ONE\u003c/em\u003e,\u003cem\u003e 16\u003c/em\u003e(3), e0247333. https://doi.org/10.1371/journal.pone.0247333 \u003c/li\u003e\n\u003cli\u003eLiao, Y., Chen, J., Zhang, Y., \u0026amp; Peng, C. (2022). The reciprocal relationship between peer victimization and internalizing problems in children and adolescents: A meta-analysis of longitudinal studies. \u003cem\u003eActa Psychologica Sinica\u003c/em\u003e,\u003cem\u003e 54\u003c/em\u003e(7), 828-849. https://doi.org/10.3724/SP.J.1041.2022.00828 \u003c/li\u003e\n\u003cli\u003eLosiewicz, O. M., Metts, A. V., Zinbarg, R. E., Hammen, C., \u0026amp; Craske, M. G. (2023). Examining the indirect contributions of irritability and chronic interpersonal stress on symptoms of anxiety and depression in adolescents. \u003cem\u003eJOURNAL OF AFFECTIVE DISORDERS\u003c/em\u003e,\u003cem\u003e 329\u003c/em\u003e, 350-358. https://doi.org/10.1016/j.jad.2023.02.125 \u003c/li\u003e\n\u003cli\u003eLu, B., Lin, L., \u0026amp; Su, X. (2024). Global burden of depression or depressive symptoms in children and adolescents: A systematic review and meta-analysis. \u003cem\u003eJOURNAL OF AFFECTIVE DISORDERS\u003c/em\u003e,\u003cem\u003e 354\u003c/em\u003e, 553-562. https://doi.org/10.1016/j.jad.2024.03.074 \u003c/li\u003e\n\u003cli\u003eLundberg, S. M., \u0026amp; Lee, S.-I. (2017, 2017-12-04). A unified approach to interpreting model predictions.\u003cem\u003eNIPS\u0026apos;17\u003c/em\u003e \u003c/li\u003e\n\u003cli\u003eManuele, S. J., Yap, M. B. H., Lin, S. C., Pozzi, E., \u0026amp; Whittle, S. (2023). Associations between paternal versus maternal parenting behaviors and child and adolescent internalizing problems: A systematic review and meta-analysis. \u003cem\u003eClinical Psychology Review\u003c/em\u003e,\u003cem\u003e 105\u003c/em\u003e, 102339. https://doi.org/10.1016/j.cpr.2023.102339 \u003c/li\u003e\n\u003cli\u003eMardini, M. T., Khalil, G. E., Bai, C., DivaKaran, A. M., \u0026amp; Ray, J. M. (2025). Identifying adolescent depression and anxiety through real-world data and social determinants of health: Machine learning model development and validation. \u003cem\u003eJMIR MENTAL HEALTH\u003c/em\u003e,\u003cem\u003e 12\u003c/em\u003e, e66665. https://doi.org/10.2196/66665 \u003c/li\u003e\n\u003cli\u003eMastrotheodoros, S., Can\u0026aacute;rio, C., Cristina Gugliandolo, M., Merkas, M., \u0026amp; Keijsers, L. (2020). Family Functioning and Adolescent Internalizing and Externalizing Problems: Disentangling between-, and Within-Family Associations. \u003cem\u003eJOURNAL OF YOUTH AND ADOLESCENCE\u003c/em\u003e,\u003cem\u003e 49\u003c/em\u003e(4), 804-817. https://doi.org/10.1007/s10964-019-01094-z \u003c/li\u003e\n\u003cli\u003eMenardi, G., \u0026amp; Torelli, N. (2014). Training and assessing classification rules with imbalanced data. \u003cem\u003eData Mining and Knowledge Discovery\u003c/em\u003e,\u003cem\u003e 28\u003c/em\u003e(1), 92-122. https://doi.org/10.1007/s10618-012-0295-5 \u003c/li\u003e\n\u003cli\u003eMonroe, S. M., \u0026amp; Simons, A. D. (1991). Diathesis-stress theories in the context of life stress research: Implications for the depressive disorders. \u003cem\u003ePSYCHOLOGICAL BULLETIN\u003c/em\u003e,\u003cem\u003e 110\u003c/em\u003e(3), 406-425. https://doi.org/10.1037/0033-2909.110.3.406 \u003c/li\u003e\n\u003cli\u003eO\u0026lsquo;Neill, A., Humphrey, N., \u0026amp; Stapley, E. (2023). A systematic review of qualitative research focusing on emotional distress among adolescents: Perceived cause and help-seeking. \u003cem\u003eAdolescent Research Review\u003c/em\u003e,\u003cem\u003e 8\u003c/em\u003e(3), 387-402. https://doi.org/10.1007/s40894-022-00203-7 \u003c/li\u003e\n\u003cli\u003eOlfati, M., Samea, F., Faghihroohi, S., Balajoo, S. M., K\u0026uuml;ppers, V., Genon, S., Patil, K., Eickhoff, S. B., \u0026amp; Tahmasian, M. (2024). Prediction of depressive symptoms severity based on sleep quality, anxiety, and gray matter volume: A generalizable machine learning approach across three datasets. \u003cem\u003eeBioMedicine\u003c/em\u003e,\u003cem\u003e 108\u003c/em\u003e. https://doi.org/10.1016/j.ebiom.2024.105313 \u003c/li\u003e\n\u003cli\u003ePalagini, L., Hertenstein, E., Riemann, D., \u0026amp; Nissen, C. (2022). Sleep, insomnia and mental health. \u003cem\u003eJournal of Sleep Research\u003c/em\u003e,\u003cem\u003e 31\u003c/em\u003e(4), e13628. https://doi.org/10.1111/jsr.13628 \u003c/li\u003e\n\u003cli\u003ePalmer, C. A., Bower, J. L., Cho, K. W., Clementi, M. A., Lau, S., Oosterhoff, B., \u0026amp; Alfano, C. A. (2024). Sleep loss and emotion: A systematic review and meta-analysis of over 50 years of experimental research. \u003cem\u003ePSYCHOLOGICAL BULLETIN\u003c/em\u003e,\u003cem\u003e 150\u003c/em\u003e(4), 440-463. https://doi.org/10.1037/bul0000410 \u003c/li\u003e\n\u003cli\u003ePotter, J. R., \u0026amp; Yoon, K. L. (2023). Interpersonal factors, peer relationship stressors, and gender differences in adolescent depression. \u003cem\u003eCurrent Psychiatry Reports\u003c/em\u003e,\u003cem\u003e 25\u003c/em\u003e(11), 759-767. https://doi.org/10.1007/s11920-023-01465-1 \u003c/li\u003e\n\u003cli\u003eRecchia, F., Bernal, J. D. K., Fong, D. Y., Wong, S. H. S., Chung, P.-K., Chan, D. K. C., Capio, C. M., Yu, C. C. W., Wong, S. W. S., Sit, C. H. P., Chen, Y.-J., Thompson, W. R., \u0026amp; Siu, P. M. (2023). Physical activity interventions to alleviate depressive symptoms in children and adolescents: A systematic review and meta-analysis. \u003cem\u003eJAMA Pediatrics\u003c/em\u003e,\u003cem\u003e 177\u003c/em\u003e(2), 132-140. https://doi.org/10.1001/jamapediatrics.2022.5090 \u003c/li\u003e\n\u003cli\u003eRodriguez-Ayllon, M., Cadenas-S\u0026aacute;nchez, C., Est\u0026eacute;vez-L\u0026oacute;pez, F., Mu\u0026ntilde;oz, N. E., Mora-Gonzalez, J., Migueles, J. H., Molina-Garc\u0026iacute;a, P., Henriksson, H., Mena-Molina, A., Mart\u0026iacute;nez-Vizca\u0026iacute;no, V., Catena, A., L\u0026ouml;f, M., Erickson, K. I., Lubans, D. R., Ortega, F. B., \u0026amp; Esteban-Cornejo, I. (2019). Role of Physical Activity and Sedentary Behavior in the Mental Health of Preschoolers, Children and Adolescents: A Systematic Review and Meta-Analysis. \u003cem\u003eSports Medicine (Auckland, N.Z.)\u003c/em\u003e,\u003cem\u003e 49\u003c/em\u003e(9), 1383-1410. https://doi.org/10.1007/s40279-019-01099-5 \u003c/li\u003e\n\u003cli\u003eSampasa-Kanyinga, H., Colman, I., Goldfield, G. S., Janssen, I., Wang, J., Podinic, I., Tremblay, M. S., Saunders, T. J., Sampson, M., \u0026amp; Chaput, J.-P. (2020). Combinations of physical activity, sedentary time, and sleep duration and their associations with depressive symptoms and other mental health problems in children and adolescents: a systematic review. \u003cem\u003eThe International Journal of Behavioral Nutrition and Physical Activity\u003c/em\u003e,\u003cem\u003e 17\u003c/em\u003e(1), 72. https://doi.org/10.1186/s12966-020-00976-x \u003c/li\u003e\n\u003cli\u003eSchulenberg, S. E., Schnetzer, L. W., \u0026amp; Buchanan, E. M. (2011). The Purpose in Life Test-Short Form: Development and Psychometric Support. \u003cem\u003eJOURNAL OF HAPPINESS STUDIES\u003c/em\u003e,\u003cem\u003e 12\u003c/em\u003e(5), 861-876. https://doi.org/10.1007/s10902-010-9231-9 \u003c/li\u003e\n\u003cli\u003eScott, J., Kallestad, H., Vedaa, O., Sivertsen, B., \u0026amp; Etain, B. (2021). Sleep disturbances and first onset of major mental disorders in adolescence and early adulthood: A systematic review and meta-analysis. \u003cem\u003eSLEEP MEDICINE REVIEWS\u003c/em\u003e,\u003cem\u003e 57\u003c/em\u003e, 101429. https://doi.org/10.1016/j.smrv.2021.101429 \u003c/li\u003e\n\u003cli\u003eSeligman, M. E. P. (1975). \u003cem\u003eHelplessness: on depression, development, and death\u003c/em\u003e. W H Freeman/Times Books/ Henry Holt \u0026amp; Co. \u003c/li\u003e\n\u003cli\u003eShatte, A. B. R., Hutchinson, D. M., \u0026amp; Teague, S. J. (2019). Machine learning in mental health: a scoping review of methods and applications. \u003cem\u003ePsychological Medicine\u003c/em\u003e,\u003cem\u003e 49\u003c/em\u003e(9), 1426-1448. https://doi.org/10.1017/S0033291719000151 \u003c/li\u003e\n\u003cli\u003eShorey, S., Ng, E. D., \u0026amp; Wong, C. H. J. (2022). Global prevalence of depression and elevated depressive symptoms among adolescents: A systematic review and meta-analysis. \u003cem\u003eBritish Journal of Clinical Psychology\u003c/em\u003e,\u003cem\u003e 61\u003c/em\u003e(2), 287-305. https://doi.org/10.1111/bjc.12333 \u003c/li\u003e\n\u003cli\u003eShwartz-Ziv, R., \u0026amp; Armon, A. (2022). Tabular data: Deep learning is not all you need. \u003cem\u003eInformation Fusion\u003c/em\u003e,\u003cem\u003e 81\u003c/em\u003e, 84-90. https://doi.org/10.1016/j.inffus.2021.11.011 \u003c/li\u003e\n\u003cli\u003eStolz, D. S., M\u0026uuml;ller-Pinzler, L., Krach, S., \u0026amp; Paulus, F. M. (2020). Internal control beliefs shape positive affect and associated neural dynamics during outcome valuation. \u003cem\u003eNature Communications\u003c/em\u003e,\u003cem\u003e 11\u003c/em\u003e(1), 1230. https://doi.org/10.1038/s41467-020-14800-4 \u003c/li\u003e\n\u003cli\u003eThapar, A., Eyre, O., Patel, V., \u0026amp; Brent, D. (2022). Depression in young people. \u003cem\u003eLancet\u003c/em\u003e,\u003cem\u003e 400\u003c/em\u003e(10352), 617-631. https://doi.org/10.1016/S0140-6736(22)01012-1 \u003c/li\u003e\n\u003cli\u003eTibshirani, R. (1996). Regression shrinkage and selection via the lasso. \u003cem\u003eJournal of the Royal Statistical Society: Series B (methodological)\u003c/em\u003e,\u003cem\u003e 58\u003c/em\u003e(1), 267-288. https://doi.org/10.1111/j.2517-6161.1996.tb02080.x \u003c/li\u003e\n\u003cli\u003eTomkinson, G. R., Carver, K. D., Atkinson, F., Daniell, N. D., Lewis, L. K., Fitzgerald, J. S., Lang, J. J., \u0026amp; Ortega, F. B. (2018). European normative values for physical fitness in children and adolescents aged 9-17 years: Results from 2 779 165 eurofit performances representing 30 countries. \u003cem\u003eBritish Journal of Sports Medicine\u003c/em\u003e,\u003cem\u003e 52\u003c/em\u003e(22), 1445-1456. https://doi.org/10.1136/bjsports-2017-098253 \u003c/li\u003e\n\u003cli\u003eWang, W., Du, X., Guo, Y., Li, W., Teopiz, K. M., Shi, J., Guo, L., Lu, C., \u0026amp; McIntyre, R. S. (2021). The associations between sleep situations and mental health among Chinese adolescents: A longitudinal study. \u003cem\u003eSleep Medicine\u003c/em\u003e,\u003cem\u003e 82\u003c/em\u003e, 71-77. https://doi.org/10.1016/j.sleep.2021.03.009 \u003c/li\u003e\n\u003cli\u003eWard, S., Womick, J., Titova, L., \u0026amp; King, L. (2023). Meaning in Life and Coping With Everyday Stressors. \u003cem\u003ePersonality \u0026amp; Social Psychology Bulletin\u003c/em\u003e,\u003cem\u003e 49\u003c/em\u003e(3), 460-476. https://doi.org/10.1177/01461672211068910 \u003c/li\u003e\n\u003cli\u003eWehmeyer, M. L., Shogren, K. A., Little, T. D., \u0026amp; Lopez, S. J. (Eds.). (2017). \u003cem\u003eDevelopment of Self-Determination Through the Life-Course\u003c/em\u003e. Springer Netherlands. https://doi.org/10.1007/978-94-024-1042-6\u003c/li\u003e\n\u003cli\u003eWen, S., Yu, X., Jin, L., Gong, J., Zhang, X., Sun, J., Zhang, S., \u0026amp; Lyu, H. (2024). A three-level meta-analysis of the relationship between family dysfunction and mental health of children and adolescents. \u003cem\u003eAdvances in Psychological Science\u003c/em\u003e,\u003cem\u003e 32\u003c/em\u003e(5), 771. https://doi.org/10.3724/SP.J.1042.2024.00771 \u003c/li\u003e\n\u003cli\u003eWong, T. K. Y., \u0026amp; Hamza, C. A. (2023). Longitudinal Associations among Identity Processes and Mental Health in Young Adulthood: The Mediating Role of Social Support. \u003cem\u003eJOURNAL OF YOUTH AND ADOLESCENCE\u003c/em\u003e. https://doi.org/10.1007/s10964-023-01883-7 \u003c/li\u003e\n\u003cli\u003eWood, A. M., Linley, P. A., Maltby, J., Baliousis, M., \u0026amp; Joseph, S. (2008). The authentic personality: A theoretical and empirical conceptualization and the development of the Authenticity Scale. \u003cem\u003eJournal of Counseling Psychology\u003c/em\u003e,\u003cem\u003e 55\u003c/em\u003e(3), 385-399. https://doi.org/10.1037/0022-0167.55.3.385 \u003c/li\u003e\n\u003cli\u003eWoodfield, M., Butler, N. G., \u0026amp; Tsappis, M. (2024). Impact of sleep and mental health in adolescence: an overview. \u003cem\u003eCURRENT OPINION IN PEDIATRICS\u003c/em\u003e,\u003cem\u003e 36\u003c/em\u003e(4), 375-381. https://doi.org/10.1097/MOP.0000000000001358 \u003c/li\u003e\n\u003cli\u003eXia, M., \u0026amp; Xu, X. (2023). Does authenticity always breed mental health? A cross-cultural comparison between the United States and China. \u003cem\u003eASIAN JOURNAL OF SOCIAL PSYCHOLOGY\u003c/em\u003e,\u003cem\u003e 26\u003c/em\u003e(1), 132-145. https://doi.org/10.1111/ajsp.12549 \u003c/li\u003e\n\u003cli\u003eXiang, Y., Cao, R., \u0026amp; Li, X. (2024). Parental education level and adolescent depression: A multi-country meta-analysis. \u003cem\u003eJOURNAL OF AFFECTIVE DISORDERS\u003c/em\u003e,\u003cem\u003e 347\u003c/em\u003e, 645-655. https://doi.org/10.1016/j.jad.2023.11.081 \u003c/li\u003e\n\u003cli\u003eYang, X., Fang, Y., Wang, Y., Liu, S., Cai, J., Li, H., \u0026amp; Chen, Z. (2024). Multidimensional stressors and depressive and anxiety symptoms in adolescents: A network analysis through simulations. \u003cem\u003eJOURNAL OF AFFECTIVE DISORDERS\u003c/em\u003e,\u003cem\u003e 347\u003c/em\u003e, 364-374. https://doi.org/10.1016/j.jad.2023.11.057 \u003c/li\u003e\n\u003cli\u003eZhou, J., Liu, Y., Ma, J., Feng, Z., Hu, J., Hu, J., \u0026amp; Dong, B. (2024). Prevalence of depressive symptoms among children and adolescents in China: A systematic review and meta-analysis. \u003cem\u003eChild And Adolescent Psychiatry And Mental HEALTH\u003c/em\u003e,\u003cem\u003e 18\u003c/em\u003e(1), 150. https://doi.org/10.1186/s13034-024-00841-w \u003c/li\u003e\n\u003cli\u003eKato, A. (1983). Aspects and structure of identity among college students [in Japanese]. \u003cem\u003eJapanese Journal of Educational Psychology\u003c/em\u003e, 31(4), 292\u0026ndash;302.\u003cem\u003ehttps://doi.org/10.5926/jjep1953.31.4_292\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003eZheng, Q. Q., \u0026amp; Chen, S. L. (1999). Preliminary development of a stressor scale for middle school students [in Chinese]. \u003cem\u003ePsychological Development and Education\u003c/em\u003e, (4), 45\u0026ndash;49.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Supplementary Material","content":"\u003cp\u003eSupplementary Table 1 and Supplementary Figures 1-2 are not available with this version.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-public-health","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"pubh","sideBox":"Learn more about [BMC Public Health](http://bmcpublichealth.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/pubh/default.aspx","title":"BMC Public Health","twitterHandle":"@BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Adolescent depression, Machine learning, Predictive modeling, SHAP","lastPublishedDoi":"10.21203/rs.3.rs-6887876/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6887876/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis study developed a multidimensional machine learning model grounded in ecological systems theory to enable early depression detection and targeted interventions among Chinese adolescents. Utilizing data from 7,169 middle school students (aged 14.98\u0026thinsp;\u0026plusmn;\u0026thinsp;1.58 years, depression prevalence 36.01%) in Liaoning Province, China, 21 key predictors were selected via LASSO regression and Boruta algorithm. Five models (XGBoost, random forest, logistic regression, multilayer perceptron, and support vector machine) were evaluated. XGBoost demonstrated optimal performance (AUC\u0026thinsp;=\u0026thinsp;0.912). SHAP analysis identified five core predictors: sleep quality (primary factor), authenticity, meaning in life, sense of control, and perceived peer-related stress. Protective factors included sleep quality, authenticity, and sense of control, while perceived peer-related stress was a risk factor. Nonlinear associations emerged between meaning in life and depression, with age-stratified thresholds (12\u0026ndash;13 years: moderate meaning linked to highest risk; 14\u0026ndash;16 years: moderate meaning reduced risk; 17\u0026ndash;18 years: high meaning mitigated risk). Findings suggest a tri-level intervention framework: sleep-exercise programs (biological), age-specific resilience training (cognitive), and AI-driven school stress monitoring (environmental).\u003c/p\u003e","manuscriptTitle":"Machine Learning-Based Prediction Model and Key Determinants of Adolescent Depression in China: A Multicenter Cross-Sectional Study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-07-27 18:44:15","doi":"10.21203/rs.3.rs-6887876/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-05-21T15:47:17+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-18T06:42:22+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"180587170332839588641944006782516036976","date":"2026-05-14T15:14:09+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-08-16T14:52:03+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-07-31T03:13:39+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"198602723319964857932135074167387866925","date":"2025-07-31T02:27:51+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-07-30T11:02:37+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"292146521055597976673855191121557905946","date":"2025-07-30T09:28:32+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-07-29T09:00:26+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"106699348302406198350147202623313337877","date":"2025-07-28T20:58:27+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"185823906858557723144867451998948410967","date":"2025-07-28T13:56:16+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"314306999844845206709941722977016877396","date":"2025-07-24T23:39:28+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-07-21T12:25:07+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-06-27T12:34:56+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-06-19T07:51:00+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-06-18T13:20:50+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Public Health","date":"2025-06-18T12:17:13+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-public-health","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"pubh","sideBox":"Learn more about [BMC Public Health](http://bmcpublichealth.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/pubh/default.aspx","title":"BMC Public Health","twitterHandle":"@BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"9921717e-7075-4a53-9696-784b66727cfe","owner":[],"postedDate":"July 27th, 2025","published":true,"recentEditorialEvents":[{"type":"decision","content":"Revision requested","date":"2026-05-21T15:47:17+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-18T06:42:22+00:00","index":90,"fulltext":""},{"type":"reviewerAgreed","content":"180587170332839588641944006782516036976","date":"2026-05-14T15:14:09+00:00","index":89,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"in-revision","subjectAreas":[],"tags":[],"updatedAt":"2026-05-21T15:55:23+00:00","versionOfRecord":[],"versionCreatedAt":"2025-07-27 18:44:15","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6887876","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6887876","identity":"rs-6887876","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}