Mental health issues, especially depressive symptoms, among young adults represent a public health challenge. Conventional psychological assessment tools have limited sensitivity and specificity for identifying individuals at risk. This study aims to develop an explainable machine learning-based model to stratify concurrent depression risk in young adults. This study included 100,257 college students and collected mental health variables including depression, anxiety, resilience, parent-child relationship, and duration of mobile phone usage. The screening capabilities of 13 machine learning algorithms were systematically evaluated and compared. The SHapley Additive exPlanations (SHAP) framework was employed for the interpretability of the final model. The median scores for parent-child relationship, resilience, anxiety, and mobile phone usage time was 42.0, 28.0, 1.0 and 28.0, respectively. Among the 13 machine learning algorithms, the XGBoost model demonstrated superior performance. The final multivariate screening model achieved an area under the curve (AUC) of 0.887, a sensitivity of 0.787, a specificity of 0.830, and an accuracy of 0.816 in classifying young adults' concurrent depression risk. The SHAP analysis showed the importance of each variable: anxiety (2.303) > resilience (0.774) > parent-child relationship (0.708) > mobile phone usage time (0.411). The final multivariate model exhibited stable performance during cross-validation (AUC = 0.885 ± 0.032), significantly better than the single-variable model (P < 0.001) and better screening reliability (Brier score 0.153). The final multivariate XGBoost model provides a highly accurate and interpretable approach for young adults' depression risk stratification. As the model was developed using cross-sectional data collected during the COVID-19 campus lockdown, prospective validation is required before clinical deployment. Notably, anxiety level emerged as the most influential risk factor, and resilience demonstrated a significant protective effect.
University enrollment coincides with a developmental period of heightened vulnerability to the onset of mental health difficulties, compounded by academic pressure, financial strain, social transition, and, for many students, distance from established support networks. This paper examines the prevalence and correlates of depression, anxiety, stress, and related outcomes among university students, and synthesizes an ecological framework spanning individual, academic, social, institutional, digital/lifestyle, and financial/environmental risk factors. Using a cross-sectional survey design modeled on validated screening instruments, we estimate symptom prevalence, identify statistically significant risk and protective factors via multivariable logistic regression, and evaluate the association between specific preventive strategies and reductions in self-reported symptom burden. Elevated stress (46.1%) and sleep disturbance (41.3%) were the most commonly reported concerns, and prior mental health diagnosis (adjusted OR = 3.12), low social support (OR = 2.48), and financial strain (OR = 2.31) emerged as the strongest independent risk factors, while regular physical activity and strong peer networks were protective. Preventive strategies — peer support programs, mindfulness-based stress reduction, adjusted academic workload policies, expanded counseling access, and physical activity programming — were each associated with meaningful reductions in elevated-symptom prevalence in program evaluation data. Stigma, lack of time, and low awareness of services were the most commonly cited barriers to help-seeking. We discuss implications for campus mental health policy, propose a multi-tiered prevention model, and outline limitations and directions for future longitudinal research. This paper synthesizes research findings for educational and policy purposes and is not a substitute for individualized clinical assessment.
Dr Emon Hasan, Akash Banerjee, MD Hasib Hossain· Journal of Information Techn...· 0 citations
Personal values may be important for university students’ mental health. However, most studies focus on single indicators such as anxiety, depression, or well-being, with limited attention to overall mental health profiles. How different value dimensions relate to mental health heterogeneity from a dual-factor perspective remains unclear. In this cross-sectional study, latent profile analysis and multinomial logistic regression were used to identify mental health profiles and examine their associations with personal values. A total of 14,929 Chinese university students completed self-report measures of personal values, anxiety, depression, life satisfaction, and psychological resilience. Three mental health profiles were identified: complete mental health, vulnerable, and troubled. The symptomatic but content profile was not found. Multinomial logistic regression showed that self-transcendence was linked to more favorable profiles, whereas self-enhancement was linked to less favorable profiles. Openness to change showed a conditional pattern: it was generally linked to lower likelihood of vulnerable or troubled membership, but among vulnerable and troubled students, higher openness to change was linked to higher likelihood of troubled membership. Conservation showed a weaker role and mainly distinguished the complete mental health and vulnerable profiles. This study provides a person-centered view of mental health heterogeneity among Chinese university students. The findings suggest that personal values are differentially associated with overall mental health profiles and support integrating symptom screening, value guidance, and positive psychological resource development in university mental health education.
Generative artificial intelligence (GenAI) has become a routine academic tool for university students, but dysregulated use may co-occur with anxiety, sleep problems, psychological distress, and impaired academic functioning. This study examined GenAI-related digital behavior as a public health issue and assessed whether health literacy and self-regulated learning were associated with lower-risk profiles.
A cross-sectional survey was conducted among Chinese university students from six universities in eastern, central, and western China. A total of 2,140 questionnaires were submitted; after four-step data-quality screening, 1,872 valid responses were retained. Measures assessed GenAI-use dysregulation, academic anxiety, sleep problems, psychological distress, health literacy, self-regulated learning, academic engagement, and demographic and use-related covariates. Latent profile analysis identified GenAI-use and mental-health risk profiles. Profile differences were examined using omnibus and
post hoc
tests, multinomial logistic regression examined external factors associated with profile membership, and regularized network analysis identified central and bridge nodes.
A four-profile solution showed the best balance of fit, entropy, class size, and interpretability (entropy = 0.891). The profiles were low-risk adaptive users (39.9%), anxiety-prone dependent users (25.7%), sleep-disrupted overusers (19.8%), and high-risk dysregulated users (14.6%). Confirmatory factor analysis supported the measurement structure of the adapted and selected scales (CFI = 0.958, TLI = 0.951, RMSEA = 0.036, SRMR = 0.041), and Harman's single-factor test did not indicate serious common method bias. Higher health literacy and self-regulated learning were associated with greater odds of low-risk rather than high-risk membership. Network analysis identified loss of control, academic worry, delayed bedtime, and academic avoidance as central or bridge nodes, whereas credibility evaluation and time management showed negative bridge expected influence.
GenAI-related digital behavior among university students varied in ways that involved dysregulated use, academic anxiety, sleep problems, distress, and academic functioning. These findings may help universities identify students who report difficulty controlling AI use, academic worry, delayed bedtime, academic avoidance, or trouble setting limits on AI use. They also point to health literacy, self-regulation, and time management as candidate resources for future longitudinal and intervention studies.
Zeyu Zhang, Yinlan Zhang, Xiaomei Lu et al.· Frontiers in Public Health· 0 citations
Mental health issues among students require data-driven approaches for early identification. This study aims to classify students’ mental health levels using the Depression Anxiety Stress Scale (DASS-21) and perceived social support, measured by the Multidimensional Scale of Perceived Social Support (MSPSS), via machine learning algorithms. A supervised classification approach was employed using Random Forest, Support Vector Machine, and Logistic Regression on data collected from 450 respondents. The data were processed through scoring, labeling, encoding, balancing, and stratified 80:20 splitting. Model evaluation was conducted using hold-out testing and 5-fold cross-validation to ensure robust and reliable performance estimation. The results indicate that Random Forest achieved the best performance, with an accuracy of 0.97 on the test set, outperforming Support Vector Machine (0.81) and Logistic Regression (0.83). Improvements in recall and F1-score for minority classes demonstrate the effectiveness of the balancing process. These findings highlight the potential of machine learning for student mental health classification, although further validation on larger and more diverse datasets is required.
Mental disorders are a leading cause of morbidity among children and adolescents worldwide, with Chile showing particularly high levels of stress, anxiety, and depression in students. Multi-Tiered Systems of Support (MTSS) offer a structured framework for school-based mental health interventions, yet sustaining fidelity remains challenging. This cross-sectional study analyzed how the COM-B model dimensions (capability, opportunity, and motivation) relate to intervention fidelity in Chilean public schools. Data were collected via surveys from 359 education professionals across all 16 regions, using an adapted DIBQ-18 and a composite fidelity index. Partial least squares structural equation modeling revealed psychological capability and reflective motivation as direct predictors of fidelity, while physical and social opportunities and automatic motivation exerted indirect effects through psychological capability. The model explained 33.2% of fidelity variance. Findings highlight the central role of individual-level determinants, suggesting that capacity-building strategies should prioritize enhancing competencies to strengthen school-based mental health interventions.
R. Rojas-Andrade, Luiz Paulo Ribeiro· Doxa· 0 citations
Depression and anxiety disorders are among the most prevalent and debilitating mental health conditions worldwide, imposing substantial personal, social, and economic burdens. Although recent advances in Large Language Models (LLMs) have shown promise in supporting mental health assessment and intervention, existing approaches often lack contextual awareness, real-time adaptability, and privacy-preserving personalization. To address these limitations, we propose a novel, context-aware and privacy-preserving mental health evaluation architecture that synergistically integrates LLM-driven intelligence. The proposed system enables personalized, continuous, and stigma-free mental health support by combining structured multiple-choice questionnaires with advanced language models, including GPT-3.5-turbo and Groq, to analyze user inputs, identify behavioral patterns, and predict potential mental health conditions such as depression and anxiety. Furthermore, the platform provides individualized recommendations, including self-care strategies, lifestyle adjustments, mindfulness practices, and referrals to healthcare professionals when appropriate. Recognizing the critical importance of reliability in sensitive healthcare settings, we introduce an ensemble-based aggregation framework that explicitly incorporates classification confidence and uncertainty quantification across multiple LLMs. Experimental results demonstrate that the proposed approach outperforms existing LLM models. By prioritizing user anonymity and data privacy, the proposed system reduces psychological barriers to seeking mental health support and promotes early intervention.
Jashraj Jani, Sara Akif, Wassila Lalouani· International Conference on...· 0 citations