机器学习 社会心理学风险因子 非扁平输入 预测青少年抑郁

Non-suicidal self-injury (NSSI) in adolescent girls is a critical predictor of subsequent depression and suicide risk, yet current tools lack both accuracy and clinical interpretability. We developed the first explainable machine learning model integrating multicenter psychosocial data to predict depression among Chinese adolescent girls with NSSI, addressing the critical need for culturally tailored risk stratification tools. In this cross - sectional observational study, our model was developed using data from 14 hospitals. We used five categories of data as predictors, including individual, family, school, psychosocial, behavioral and lifestyle factors. We compared seven machine learning models and selected the best one to develop final model and the Shapley Additive exPlanations (SHAP) method were used to explain model prediction. The Random Forest (RF) model was compared against six other machine learning algorithms. We assessed the discrimination using the area under receiver operating characteristic (AUROC) with 95 % CIs. Using the development dataset (n = 1163) and predictive model building process, a simplified model containing only the top 20 features had similar predictive performance to the full model, the RF model outperformed six algorithms (AUROC = 0.964 [0.945-0.975]), demonstrating superior discriminative power and robustness. The top ten risk predictors were Borderline personality, Rumination, Perceived stress, Hopelessness, Self-esteem, Sleep quality, Loneliness, Resilience, Parental care, and Problem-focused coping. We developed a three-tiered, color-coded web-based clinical tool to operationalize predictions, enabling real-time risk stratification and personalized interventions. Our study bridges machine learning and clinical interpretability to advance precision mental health interventions for vulnerable adolescent populations.

Adolescent depression is a major public health concern. Physical health indicators are rarely included in risk tools. We examined whether adding dry eye disease (DED) to psychosocial and behavioral factors improves prediction of depressive symptoms. We analyzed 2,076 adolescent questionnaires (94.5% response) covering ocular health, sleep, electronic device use, social support, and demographics. Five machine-learning classifiers were trained with cross-validation and evaluated for discrimination and calibration. Models that included DED achieved strong discrimination (AUC ≈ 0.84) and good calibration, with highest accuracy for no and severe depression and lower performance for mild/moderate categories. Integrating ocular health with psychosocial factors enhances machine-learning prediction of adolescent depression and may support earlier, school-based identification and referral. Given the low-cost, questionnaire-based inputs and favorable calibration, this approach shows promise for population screening and targeted prevention, pending external validation and prospective testing.

Depression is a major cause of disability and mortality for young people worldwide and is typically first diagnosed during adolescence. In this work, we present a machine learning framework to predict adolescent depression occurring between ages 12 and 18 years using environmental, biological, and lifestyle features of the child, mother, and partner from the child’s prenatal period to age 10 years using data from 8467 participants enrolled in the Avon Longitudinal Study of Parents and Children (ALSPAC). We trained and compared several cross-sectional and longitudinal machine learning techniques and found the resulting models predicted adolescent depression with recall (0.59 ± 0.20), specificity (0.61 ± 0.17), and accuracy (0.64 ± 0.13), using on average 39 out of the 885 total features (4.4%) included in the models. The leading informative features in our predictive models of adolescent depression were female sex, parental depression and anxiety, and exposure to stressful events or environments. This work demonstrates how using a broad array of evidence-driven predictors from early in life can inform the development of preventative decision support tools to assist in the early detection of risk for mental illness.

Abstract Background The prevalence of adolescent mental health conditions such as depression and anxiety has significantly increased. Despite the potential of machine learning (ML), there is a shortage of models that use real-world data (RWD) to enhance early detection and intervention for these conditions. Objective This study aimed to identify depression and anxiety in adolescents using ML techniques on RWD and social determinants of health (SDoH). Methods We analyzed RWD of adolescents aged 10‐17 years, considering various factors such as demographics, prior diagnoses, prescribed medications, medical procedures, and laboratory measurements recorded before the onset of anxiety or depression. Clinical data were linked with SDoH at the block-level. Three separate models were developed to predict anxiety, depression, and both conditions. Our ML model of choice was Extreme Gradient Boosting (XGBoost) and we evaluated its performance using the nested cross-validation technique. To interpret the model predictions, we used the Shapley additive explanation method. Results Our cohort included 52,054 adolescents, identifying 12,572 with anxiety, 7812 with depression, and 14,019 with either condition. The models achieved area under the curve values of 0.80 for anxiety, 0.81 for depression, and 0.78 for both combined. Excluding SDoH data had a minimal impact on model performance. Shapley additive explanation analysis identified gender, race, educational attainment, and various medical factors as key predictors of anxiety and depression. Conclusions This study highlights the potential of ML in early identification of depression and anxiety in adolescents using RWD. By leveraging RWD, health care providers may more precisely identify at-risk adolescents and intervene earlier, potentially leading to improved mental health outcomes.

Abstract Background This study leveraged machine learning to evaluate the contribution of information from multiple developmental stages to prospective prediction of depression and anxiety in mid-adolescence. Methods A community sample (N = 374; 53.5% male) of children and their families completed tri-annual assessments across ages 3–15. The feature set included several important risk factors spanning psychopathology, temperament/personality, family environment, life stress, interpersonal relationships, neurocognitive, hormonal, and neural functioning, and parental psychopathology and personality. We used canonical correlation analysis (CCA) to reduce the large feature set to a lower dimensional space while preserving the longitudinal structure of the data. Ablation analysis was conducted to evaluate the relative contributions to prediction of information gathered at different developmental periods and relative to previous disorder status (i.e. age 12 depression or anxiety) and demographics (sex, race, ethnicity). Results CCA components from individual waves predicted age 15 disorder status better than chance across ages 3, 6, 9, and 12 for anxiety and 9 and 12 for depression. Only the components from age 12 for depression, and ages 9 and 12 for anxiety, improved prediction over prior disorder status and demographics. Conclusions These findings suggest that screening for risk of adolescent depression can be successful as early as age 9, while screening for risk of adolescent anxiety can be successful as early as age 3. Assessing additional risk factors at age 12 for depression, and going back to age 9 for anxiety, can improve screening for risk at age 15 beyond knowing standard demographics and disorder history.

No abstract available

No abstract available

Heterogeneity in the course of posttraumatic stress symptoms (PTSS) following a major life trauma such as childhood sexual abuse (CSA) can be attributed to numerous contextual factors, psychosocial risk, and family/peer support. The present study investigates a comprehensive set of baseline psychosocial risk and protective factors including online behaviors predicting empirically derived PTSS trajectories over time. Females aged 12-16 years (N = 440); 156 with substantiated CSA; 284 matched comparisons with various self-reported potentially traumatic events (PTEs) were assessed at baseline and then annually for 2 subsequent years. Latent growth mixture modeling (LGMM) was used to derive PTSS trajectories, and least absolute shrinkage and selection operator (LASSO) logistic regression was used to investigate psychosocial predictors including online behaviors of trajectories. LGMM revealed four PTSS trajectories: resilient (52.1%), emerging (9.3%), recovering (19.3%), and chronic (19.4%). Of the 23 predictors considered, nine were retained in the LASSO model discriminating resilient versus chronic trajectories including the absence of CSA and other PTEs, low incidences of exposure to sexual content online, minority ethnicity status, and the presence of additional psychosocial protective factors. Results provide insights into possible intervention targets to promote resilience in adolescence following PTEs.

No abstract available

Objective: The objective of this study was to develop and interpret an explainable machine learning model capable of accurately identifying psychological risk profiles associated with suicidal ideation among youth. Methods and Materials: This study employed a cross-sectional design involving adolescents and young adults recruited from educational settings. Participants completed standardized self-report measures assessing suicidal ideation, depressive and anxiety symptoms, emotional dysregulation, hopelessness, perceived family and peer support, academic stress, bullying exposure, and problematic digital use. Multiple supervised machine learning algorithms were trained to predict suicidal ideation, with model performance evaluated using area under the receiver operating characteristic curve, accuracy, precision, recall, and F1-score. The best-performing ensemble model was selected and interpreted using Shapley Additive Explanations to generate both global feature importance and individual-level explanatory profiles. Findings: Ensemble-based machine learning models significantly outperformed traditional classifiers, achieving excellent discriminative performance in identifying suicidal ideation. Depressive symptoms, hopelessness, and emotional dysregulation emerged as the strongest positive predictors, while perceived family support demonstrated a robust protective effect. SHAP-based analyses revealed substantial heterogeneity in risk patterns, identifying multiple psychological profiles characterized by internalizing distress, affective instability, interpersonal disconnection, and contextual stress. These profiles explained individual predictions with high transparency, demonstrating that similar levels of suicidal ideation risk can arise from distinct configurations of psychological and social factors. Conclusion: The findings indicate that explainable artificial intelligence can simultaneously achieve high predictive accuracy and meaningful psychological interpretability in youth suicide risk assessment. SHAP-based machine learning offers a powerful framework for identifying individualized risk profiles.

Objective: The objective of this study was to identify and interpret high-risk substance use profiles among youth by applying explainable machine learning models that integrate psychological, familial, peer, and sociodemographic factors. Methods and Materials: A cross-sectional study design was employed with a large, community-based sample of adolescents and young adults recruited from educational institutions and youth organizations in Ireland. Participants completed standardized self-report measures assessing substance use behaviors, psychological characteristics, family and peer contexts, and demographic factors. Supervised machine learning models, including regularized logistic regression and ensemble-based algorithms, were trained to classify high-risk substance use status. Model performance was evaluated using cross-validated inferential metrics, including area under the receiver operating characteristic curve, sensitivity, specificity, and balanced accuracy. Explainable artificial intelligence techniques based on SHapley Additive exPlanations were used to interpret both global predictor importance and individual-level risk patterns. Findings: Inferential analyses demonstrated that ensemble machine learning models significantly outperformed linear models in classifying high-risk substance use, with the highest-performing model achieving excellent discrimination and sensitivity. Explainability analyses revealed that peer substance use norms, impulsivity, parental monitoring, sensation seeking, and emotional dysregulation exerted statistically meaningful and nonlinear effects on risk classification. Distinct high-risk profiles were identified, including socially driven risk, emotionally vulnerable risk, sensation-seeking–dominant risk, and structurally disadvantaged risk, each characterized by unique constellations of predictors with differential contributions to model output. Conclusion: The findings indicate that explainable machine learning models can accurately and transparently identify heterogeneous high-risk substance use profiles among youth, offering a robust and interpretable framework for advancing early detection, targeted prevention, and data-informed public health decision-making.

No abstract available

Background: As adolescent suicide rates continue to rise, innovation in risk identification is warranted. Machine learning can identify suicidal individuals based on their language samples. This feasibility pilot was conducted to explore this technology’s use in adolescent therapy sessions and assess machine learning model performance. Method: Natural language processing machine learning models to identify level of suicide risk using a smartphone app were tested in outpatient therapy sessions. Data collection included language samples, depression and suicidality standardized scale scores, and therapist impression of the client’s mental state. Previously developed models were used to predict suicidal risk. Results: 267 interviews were collected from 60 students in eight schools by ten therapists, with 29 students indicating suicide or self-harm risk. During external validation, models were trained on suicidal speech samples collected from two separate studies. We found that support vector machines (AUC: 0.75; 95% CI: 0.69–0.81) and logistic regression (AUC: 0.76; 95% CI: 0.70–0.82) lead to good discriminative ability, with an extreme gradient boosting model performing the best (AUC: 0.78; 95% CI: 0.72–0.84). Conclusion: Voice collection technology and associated procedures can be integrated into mental health therapists’ workflow. Collected language samples could be classified with good discrimination using machine learning methods.

No abstract available

The society is currently witnessing an unprecedented growth in the incidence of mental disorders, with an estimated 300 million people suffering from depression globally. People with high life satisfaction tend to suffer fewer mental health issues. The large volume of data generated on social network platforms enables us to detect hidden patterns in data and obtain new insights. This work aims to (a) explore the relationship between life satisfaction and depression in social network users, using Facebook as an example, and (b) develop a multilevel predictive model to detect users with depression. We trained a set of predictive models on datasets from myPersonality project including 2,085 participants who took the Satisfaction with Life Scale and 614 users who submitted the Centre for Epidemiological Study Depression (CES-D) scale. The resulting multilevel model establishes a negative correlation between life satisfaction and depression, and it can also improve the accuracy of a predictive model using depressive labels alone.

In recent times, depression becomes a major issue, which causes suicide, particularly among youngsters. During the Coronavirus disease (COVID-19) epidemic, many organizations suggested social distance and quarantine actions, which cause major attention to the mental health and depression of each individual. Most population express their emotions by using modern social media technologies like Twitter, Facebook, etc. By considering user tweets, a Long Short-Term Memory (LSTM)-based classifier was designed that learns rich attributes such as psychological, contextual, cognitive, person-level and distress-dependent n-gram attributes of each user for depression severity level prediction. But, it did not learn the social network structural properties of the most prominent communities, which influences the prediction outcomes. Hence in this article, a novel model is developed to predict the user’s depression severity levels by considering the social network structure of most prominent communities and influence measures. At first, it analyses the physical characteristics of the most prominent groups based on their balanced local and global power distribution. Then, the influential users and communities are identified along with the rich group of attributes. Moreover, those attributes are provided to the LSTM classifier for the user’s depression severity level prediction during the COVID-19 epidemic. Finally, the investigational outcomes exhibit that the presented model attains 93.53% accuracy and 0.4376 Root Mean Square Error (RMSE) contrasted with the conventional classifiers to estimate the user’s depression severity level.

Mental health issues are widely accepted as one of the most prominent health challenges in the world, with over 300 million people currently suffering from depression alone. With massive volumes of user-generated data on social networking platforms, researchers are increasingly using machine learning to determine whether this content can be used to detect mental health problems in users. This study aims to investigate whether training a predictive model with multiple instance learning (MIL) via Long Short-Term Memory (LSTM) and gated recurrent unit (GRU) can improve the performance of a predictive model to detect social network users with depression. The power of MIL is to learn from user-level labels to identify post-level labels. By combining every possibility of posts label category, it can generate temporal posting profiles which can then be used to classify users with depression. This study highlights that training a MIL model via LSTM and GRU can improve the accuracy of a MIL model trained with convolutional neural networks.

No abstract available

Background Major depressive disorder (MDD) or depression is among the most prevalent psychiatric disorders, affecting more than 300 million people globally. Early detection is critical for rapid intervention, which can potentially reduce the escalation of the disorder. Objective This study used data from social media networks to explore various methods of early detection of MDDs based on machine learning. We performed a thorough analysis of the dataset to characterize the subjects’ behavior based on different aspects of their writings: textual spreading, time gap, and time span. Methods We proposed 2 different approaches based on machine learning singleton and dual. The former uses 1 random forest (RF) classifier with 2 threshold functions, whereas the latter uses 2 independent RF classifiers, one to detect depressed subjects and another to identify nondepressed individuals. In both cases, features are defined from textual, semantic, and writing similarities. Results The evaluation follows a time-aware approach that rewards early detections and penalizes late detections. The results show how a dual model performs significantly better than the singleton model and is able to improve current state-of-the-art detection models by more than 10%. Conclusions Given the results, we consider that this study can help in the development of new solutions to deal with the early detection of depression on social networks.

Social networks influence health-related behavior, such as obesity and smoking. While researchers have studied social networks as a driver for diffusion of influences and behavior, it is less understood how the structure or topology of the network, in itself, impacts an individual’s health behavior and wellness state. In this paper, we investigate whether the structure or topology of a social network offers additional insight and predictability on an individual’s health and wellness. We develop a method called the Network-Driven health predictor (NetCARE) that leverages features representative of social network structure. Using a large longitudinal data set of students enrolled in the NetHealth study at the University of Notre Dame, we show that the NetCARE method improves the overall prediction performance over the baseline models—that use demographics and physical attributes—by 38%, 65%, 55%, and 54% for the wellness states—stress, happiness, positive attitude, and self-assessed health—considered in this paper.

Background Mental health problems have become increasingly prevalent in the past decade. With the advance of Web 2.0 technologies, social media present a novel platform for Web users to form online health groups. Members of online health groups discuss health-related issues and mutually help one another by anonymously revealing their mental conditions, sharing personal experiences, exchanging health information, and providing suggestions and support. The conversations in online health groups contain valuable information to facilitate the understanding of their mutual help behaviors and their mental health problems. Objective We aimed to characterize the conversations in a major online health group for major depressive disorder (MDD) patients in a popular Chinese social media platform. In particular, we intended to explain how Web users discuss depression-related issues from the perspective of the social networks and linguistic patterns revealed by the members’ conversations. Methods Social network analysis and linguistic analysis were employed to characterize the social structure and linguistic patterns, respectively. Furthermore, we integrated both perspectives to exploit the hidden relations between them. Results We found an intensive use of self-focus words and negative affect words. In general, group members used a higher proportion of negative affect words than positive affect words. The social network of the MDD group for depression possessed small-world and scale-free properties, with a much higher reciprocity ratio and clustering coefficient value as compared to the networks of other social media platforms and classic network models. We observed a number of interesting relationships, either strong correlations or convergent trends, between the topological properties and linguistic properties of the MDD group members. Conclusions (1) The MDD group members have the characteristics of self-preoccupation and negative thought content, according to Beck’s cognitive theory of depression; (2) the social structure of the MDD group is much stickier than those of other social media groups, indicating the tendency of mutual communications and efficient spread of information in the MDD group; and (3) the linguistic patterns of MDD members are associated with their topological positions in the social network.

The detection of depression in social media posts is crucial due to the increasing prevalence of mental health issues. Traditional machine learning algorithms often fail to capture intricate textual patterns, limiting their effectiveness in identifying depression. Existing studies have explored various approaches to this problem but often fall short in terms of accuracy and robustness. To address these limitations, this research proposes a neural network architecture leveraging transformer-based models combined with metadata and linguistic markers. The study employs DistilBERT, extracting information from the last four layers of the transformer, applying learned weights, and averaging them to create a rich representation of the input text. This representation, augmented by metadata and linguistic markers, enhances the model’s comprehension of each post. Dropout layers prevent overfitting, and a Multilayer Perceptron (MLP) is used for final classification. Data augmentation techniques, inspired by the Easy Data Augmentation (EDA) methods, are also employed to improve model performance. Using BERT, random insertion and substitution of phrases generate additional training data, focusing on balancing the dataset by augmenting underrepresented classes. The proposed model achieves weighted Precision, Recall, and F1-scores of 84.26%, 84.18%, and 84.15%, respectively. The augmentation techniques significantly enhance model performance, increasing the weighted F1-score from 72.59% to 84.15%.

Depression, a prevalent mental health concern, requires timely identification and intervention. Automating the early stage identification of depression cues within social media text has become critically important. Existing methods for depression identification through text obtained from social media have shown promising results; however, these methods do not address the graphlike nature of social media data. To date, there has been little work addressing this problem by appropriately modeling social media data. This article aims to effectively harness the power of multiembedding techniques and graph attention networks (GATs) for depression identification. To overcome these limitations, the current study presents a novel methodology that combines multiple token-level embeddings, including BERT, RoBERTa, and DeBERTa, with GATs to leverage the graphlike nature of social media data and detect depression cues from such text. A dataset comprising 100 000 tweets has been curated using data from publicly available annotated tweet datasets. This dataset maintains a balanced distribution between depressive and nondepressive text samples. This was followed by a rigorous cleaning and preprocessing pipeline. Next, these data were transformed into a numeric feature matrix using multiple-word embeddings, which enabled the treatment of tweets as nodes in a graphlike structure. This graph was used for training a GAT model with multiple self-attentional layers, culminating in a linear layer for mapping traits to binary classes. The model presented in the current study achieves a precision of 97.2%, a recall of 96.4%, and an F1 score of 96.7% on a benchmark dataset.

Over 320 million people are suffering from depression worldwide. Depression is one of the common mental health disorders. By its nature, depression can reoccur. People suffering from depression tend to lose interest, have low mood, feel hopeless, or have social isolation. At its worst, depression can lead to suicide. So far, there are a few numbers of studies investigating deep learning techniques to classify social network users with depression. Most of the studies used classical machine learning techniques e.g., regression, support vector machine, or decision trees. This paper aims to develop a deep learning predictive model to classify users with depression. Because depression is a recurrent disease, it is interesting in finding unusual patterns in user-generated content over time. Social network posts over time were extracted for time series data. The predictive model for the classification was obtained from deep learning techniques.

This study proposes a joint identification algorithm based on multi-source data fusion and Graph Neural Networks (GNN) to simultaneously detect adolescent internet addiction and psychological disorders. With the rapid development of internet technology, adolescent internet addiction has become increasingly prominent, and there exists a complex bidirectional relationship between internet addiction and psychological abnormalities. Traditional single-source data modeling methods struggle to accurately capture these intricate psychological and behavioral patterns. To address this, the study integrates multi-source heterogeneous data, including online behavior logs, psychological assessment data, physiological signals, and social relationships, to construct a Heterogeneous Graph Neural Network (H-GNN) model. Through a multi-task learning mechanism, the model simultaneously predicts the risks of internet addiction and psychological disorders. Experimental results demonstrate that the model outperforms traditional machine learning and deep learning models in terms of accuracy and AUC, effectively capturing the complex interactions between adolescent behaviors, psychological states, and social relationships. Ablation studies further validate the critical role of psychological features in prediction, and the model's potential in real-world applications is highlighted.

Depression has an intense impact on individuals, yet many cases go undiagnosed. Thus, it is imperative to design an effective model for the automated diagnosis of depression. However, existing methods do not adequately capture contextual information in a clinical interview. Inspired by the depression diagnosis process, we propose a new perspective on detecting depression as a dialog information extraction task. Specifically, this article constructs a heterogeneous graph that models the participant’s depression state and uses the graph attention network to aggregate the pieces of depressive clues. In addition, we use the focal loss as a loss function for dealing with class imbalance by reshaping the standard cross-entropy loss. Experimental results demonstrate that our proposed model depression state extraction with heterogeneous graph attention neural network (DSE-HGAT) surpasses the baseline models on the Distress Analysis Interview Corpus-Wizard of Oz (DAIC-WOZ) dataset. Meanwhile, agreement analysis between our proposed model and the gold standard shows that it is moderate ( $k$ = 0.528, $p 0.05$ ). Overall, our model is very effective in identifying depression in the clinical interview transcript, which has the potential to assist doctors with medical conditions.

Objective: This study aimed to examine the nonlinear and interactive effects of loneliness, family communication quality, and digital media dependency on depressive symptoms among Indonesian adolescents using neural network modeling. Methods and Materials: A cross-sectional design was implemented with 684 secondary school students aged 14–18 years selected via multistage cluster sampling from urban regions of Indonesia. Participants completed validated measures assessing depressive symptoms, loneliness, family communication quality, and digital media dependency. Data were analyzed using a multilayer perceptron neural network constructed in Python with TensorFlow. The dataset was partitioned into training, validation, and test sets. Model performance was evaluated using root mean square error, mean absolute error, and coefficient of determination. Predictor contributions and interaction effects were interpreted using SHAP values and partial dependence analyses. Findings: The neural network demonstrated high predictive accuracy (R² = .79 on the test set). Loneliness emerged as the strongest predictor of depressive symptoms, followed by digital media dependency and family communication quality. Significant nonlinear interaction effects were observed, indicating that the combination of high loneliness, poor family communication, and elevated digital media dependency produced the highest levels of depressive symptoms. Family communication quality exerted a strong buffering effect that attenuated the impact of loneliness and digital dependency on depression. Conclusion: Adolescent depression is shaped by complex, interactive psychosocial and digital factors that are effectively captured through neural network modeling. Strengthening family communication and promoting healthy digital engagement may substantially reduce depressive risk among adolescents.

College students' mental health has become a critical global concern, with increasing prevalence of depression and anxiety significantly affecting academic performance. Traditional monitoring approaches rely on periodic surveys and clinical interviews, which are subjective, resource-intensive, and detect problems only after escalation. This paper proposes a spatiotemporal graph neural network framework for real-time monitoring of students' mental health status. The method constructs dynamic social networks incorporating academic, behavioral, and interaction data, employing graph neural networks with temporal modeling to capture spatial dependencies among connected students and temporal evolution of individual psychological states. By integrating Graph Attention Networks for spatial modeling with Gated Recurrent Units for temporal learning, our framework enables early detection of mental health deterioration and identifies at-risk populations before crises occur. Experiments on real-world data from 10,000 diverse students (distributed across freshmen 25%, sophomores 28%, juniors 24%, seniors 23%, representing 8 colleges including engineering, sciences, humanities, and business) over 12 months demonstrate AUC of 0.89 and accuracy of 0.90, significantly outperforming five baselines. Temporal analysis reveals critical periods including exams and transitions requiring enhanced support, providing actionable insights for campus psychological services.

BACKGROUND Adolescent depression shows high clinical heterogeneity. Brain functional networks serve as a powerful tool for investigating neural mechanisms underlying depression profiles. A key challenge is to characterize how variation in brain functional organization links to behavioral features and psychosocial environmental influences. METHODS We recruited 80 adolescents with major depressive disorder (MDD) and 42 healthy controls (HCs). First, we estimated the differences in functional connectivity of resting-state networks (RSN) between the two groups. Then, we used sparse canonical correlation analysis to characterize patterns of associations between RSN connectivity and symptoms, cognition, and psychosocial environmental factors in MDD adolescents. Clustering analysis was applied to stratify patients into homogenous subtypes according to these brain-behavior-environment associations. RESULTS MDD adolescents showed significantly hyperconnectivity between the ventral attention and cingulo-opercular networks compared with HCs. We identified one reliable pattern of covariation between RSN connectivity and clinical/environmental features in MDD adolescents. In this pattern, psychosocial factors, especially the interpersonal and family relationships, were major contributors to variation in connectivity of salience, cingulo-opercular, ventral attention, subcortical and somatosensory-motor networks. Based on this association, we categorized patients into two subgroups which showed different environment and symptoms characteristics, and distinct connectivity alterations. These differences were covered up when the patients were taken as a whole group. CONCLUSION This study identified the environmental exposures associated with specific functional networks in MDD youths. Our findings emphasize the importance of the psychosocial context in assessing brain function alterations in adolescent depression and have the potential to promote targeted treatment and precise prevention.

Adolescent depression is a significant public health concern in low- and middle-income countries, including Kenya, where limited screening capacity contributes to under diagnosis. This study developed and validated a Bayesian Neural Network for predicting depression risk among adolescents, leveraging probabilistic inference to capture predictive uncertainty and complex nonlinear relationships. Secondary data for 2,192 adolescents aged 12-18 years were obtained from the Open Science Framework, incorporating psychosocial, demographic, and mental health measures. Significant predictors were identified using chi-square tests and point-biserial correlations, followed by forward feature selection. The proposed Bayesian Neural Network employed a three-hidden-layer feedforward architecture with ReLU and sigmoid activations, Gaussian priors on weights, and Bayes-by-Backprop Variational inference. Model performance was benchmarked against a Random Forest classifier using cross-validation. Results indicate that anxiety, loneliness, perceived social support, gratitude, positive youth development, academic self-perception, gender, academic form, financial status, parental education, and age are significant predictors of depression. The model achieved superior performance, with an accuracy of 80.13%, F1 score of 0.876, recall of 0.965, ROC-AUC of 0.813, and PR-AUC of 0.913, outperforming the Random Forest in most metrics except precision. Calibration analysis yielded a low Brier score (0.0693), indicating well-calibrated probabilistic predictions. These findings demonstrate the suitability of Bayesian Neural Networks for adolescent depression risk screening in resource-constrained settings, where uncertainty-aware decision support is critical.

BACKGROUND This study aims to develop an artificial neural network (ANN) prediction model incorporating random forest (RF) screening ability for predicting the risk of depression in adolescents and identifies key risk factors to provide a new approach for primary care screening of depression among adolescents. METHODS The data were from a large cross-sectional study conducted in China from July to September 2021, enrolling 8635 adolescents aged 10-17 with their parents. We used the Patient health questionnaire (PHQ-9) to rate adolescent depression symptoms, using scales and single-item questions to collect demographic information and other variables. Initial model variables screening used the RF importance assessment, followed by building prediction model using the screened variables through the ANN. RESULTS The rate of depression symptoms in adolescents was 24.6%, and the depression risk prediction model was built based on 70% of the training set and 30% of the test set. Ten variables were included in the final prediction model with a model accuracy of 85.03%, AUC of 0.892, specificity of 89.79%, and sensitivity of 70.81%. The top 10 significant factors of depression risk were adolescent rumination, adolescent self-esteem, adolescent mobile phone addiction, peer victimization, care in parenting styles, overprotection in parenting styles, academic pressure, conflict in parent-child relationship, parental rumination, and relationship between parents. CONCLUSIONS The ANN model based on the RF effectively identifies depression risk in adolescents and provides a methodological reference for large-scale primary screening. Cross-sectional studies and single-item scales limit further improvements in model accuracy.

There is a significant correlation between depression, verbal behavior, and facial expressions. By analyzing patients’ audio and facial visuals, depression assessments can be conducted. However, existing work is predominantly based on single modalities. Additionally, acquiring a sufficient amount of labeled data in clinical settings is challenging and costly. To leverage multimodal audio-visual data while addressing the issue of lacking trainable labeled data, we propose an audiovisual multimodal semi-supervised depression detection model based on Graph Neural Networks (AVS-GNN). This model first extracts dual-modality temporal information from audio features and facial visual features of patients and obtains modality-specific high-level embedding representations through graph representation learning. Subsequently, it utilizes graph-based contrastive unsupervised learning to capture consistency information between pairs of unlabeled samples across different modalities and to facilitate cross-modal interactions. We specifically designed a hybrid weighted pseudo-labeling strategy to assign high-confidence pseudo-labels to unlabeled data and further retraining the model. Experiments on two depression datasets show that this model outperforms baseline methods across all evaluation metrics.

Precise remote evaluation of both suicide risk and psychiatric disorders is critical for suicide prevention as well as for psychiatric well-being. Using questionnaires is an alternative to labor-intensive diagnostic interviews in a large general population, but previous models for predicting suicide attempts suffered from low sensitivity. We developed and validated a deep graph neural network model that increased the prediction sensitivity of suicide risk in young adults (n = 17,482 for training; n = 14,238 for testing) using multi-dimensional questionnaires and suicidal ideation within 2 weeks as the prediction target. The best model achieved a sensitivity of 76.3%, specificity of 83.4%, and an area under curve of 0.878 (95% confidence interval, 0.855–0.899). We demonstrated that multi-dimensional deep features covering depression, anxiety, resilience, self-esteem, and clinico-demographic information contribute to the prediction of suicidal ideation. Our model might be useful for the remote evaluation of suicide risk in the general population of young adults for specific situations such as the COVID-19 pandemic.

The global prevalence of depression necessitates the application of technological solutions, particularly sensor-based systems, to augment scarce resources for early diagnostic purposes. In this study, we use benchmark datasets that contain multimodal data including video, audio, and transcribed text. To address depression detection as a chronic long-term disorder reflected by temporal behavioral patterns, we propose a novel framework that segments videos into utterance-level instances using GRU for contextual representation, and then constructs graphs where utterance embeddings serve as nodes connected through dual relationships capturing both chronological development and intermittent relevant information. Graph neural networks are employed to learn multi-dimensional edge relationships and align multimodal representations across different temporal dependencies. Our approach achieves superior performance with an MAE of 5.25 and RMSE of 6.75 on AVEC2014, and CCC of 0.554 and RMSE of 4.61 on AVEC2019, demonstrating significant improvements over existing methods that focus primarily on momentary expressions.

No abstract available

No abstract available

No abstract available

Depression is a prevalent mental disorder that imposes a significant public health burden worldwide. Although multimodal detection methods have shown potential, existing techniques still face two critical bottlenecks: (i) insufficient integration of global patterns and local fluctuations in long-sequence modeling and (ii) static fusion strategies that fail to dynamically adapt to the complementarity and redundancy among modalities. To address these challenges, this paper proposes a dynamic multimodal depression detection framework, DynMultiDep, which combines multi-scale temporal modeling with an adaptive fusion mechanism. The core innovations of DynMultiDep lie in its Multi-scale Temporal Experts Module (MTEM) and Dynamic Multimodal Fusion module (DynMM). On one hand, MTEM employs Mamba experts to extract long-term trend features and utilizes local-window Transformers to capture short-term dynamic fluctuations, achieving adaptive fusion through a long-short routing mechanism. On the other hand, DynMM introduces modality-level and fusion-level dynamic decision-making, selecting critical modality paths and optimizing cross-modal interaction strategies based on input characteristics. The experimental results demonstrate that DynMultiDep outperforms existing state-of-the-art methods in detection performance on two widely used large-scale depression datasets.

Passively detecting depression is important to achieve universal screening. As text and call logs have been overlooked as potential passive screening modalities, we construct time series from these text and call logs using the number of communications, number of contacts, and average length of communications. We calculate these values for every 4, 6, 12, and 24 hours. We leverage a time series feature extraction library to extract features for machine learning models. Our results show that outgoing texts are more predictive of depression than incoming texts but incoming calls are more predictive of depression than outgoing calls. Specifically, we are able to achieve an F1 score of 0.72 with average text length and an F1 score of 0.65 with call duration. This detailed exploration into the ability of text and call logs to screen for depression will help guide future research in this domain.

Background Psychosocial distress among youth is a major public health issue characterized by disruptions in cognitive control processing. Using the National Institute of Mental Health's Research Domain Criteria (RDoC) framework, we quantified multidimensional neural oscillatory markers of psychosocial distress serving cognitive control in youth. Methods The sample consisted of 39 peri-adolescent participants who completed the NIH Toolbox Emotion Battery (NIHTB-EB) and the Eriksen flanker task during magnetoencephalography (MEG). A psychosocial distress index was computed with exploratory factor analysis using assessments from the NIHTB-EB. MEG data were analyzed in the time-frequency domain and peak voxels from oscillatory maps depicting the neural cognitive interference effect were extracted for voxel time series analyses to identify spontaneous and oscillatory aberrations in dynamics serving cognitive control as a function of psychosocial distress. Further, we quantified the relationship between psychosocial distress and dynamic functional connectivity between regions supporting cognitive control. Results The continuous psychosocial distress index was strongly associated with validated measures of pediatric psychopathology. Theta-band neural cognitive interference was identified in the left dorsolateral prefrontal cortex (dlPFC) and middle cingulate cortex (MCC). Time series analyses of these regions indicated that greater psychosocial distress was associated with elevated spontaneous activity in both the dlPFC and MCC and blunted theta oscillations in the MCC. Finally, we found that stronger phase coherence between the dlPFC and MCC was associated with greater psychosocial distress. Conclusions Greater psychosocial distress was marked by alterations in spontaneous and oscillatory theta activity serving cognitive control, along with hyperconnectivity between the dlPFC and MCC.

. Major Depressive Disorder (MDD) has been known as one of the most prevalent mental disorders whose symptoms can be observed from changes in facial behaviors. Previous studies had attempted to build Machine Learning (ML) models to assess depression severity using such features but few have utilized these models to determine key facial behaviors for MDD. In this study, we used video data to assess the severity of MDD and determine important features based on three approaches (XGBoost, Spearman’s correlation, and t-test). In addition, there is the Facial Action Coding System (FACS) framework that allows visual data such as changes in facial behavior to be modeled as time series data. The results show that the XGBoost model obtained the best results when trained using features selected through the t-test statistical method with 5.387 MAE, 6.266 RMSE, and 0.042 R 2 . The majority of the important features consist of Action Unit (AU) and Features 3D around the regions of the left eye, right cheek, and lip area. However, the majority of the important features discovered from the three approaches, are the first derivatives of the 3D facial landmark coordinates of the cheeks, eyes, and ∗ Corresponding

Social media platforms like Twitter generate massive amounts of multimodal content, such as text and images, which makes them a great source for learning about people's mental health. If we can detect signs of depression early from what users share online, it could help support timely interventions and assist clinical screening. Social media analysis provides useful opportunities for early depression detection. This work evaluates multimodal architectures that utilise various BERT text encoders with CNN image encoders to detect depressive expression in tweets. Our dataset consists of manually collected Twitter posts that were labeled by experts and crowd annotators. The best-performing model, RoBERTa+ResNet50, as per the accuracy, F1-score, and ROC-AUC. These results either match or surpass the performance of recent multimodal approaches. Our contributions include a newly labeled multimodal Twitter dataset, a thorough evaluation across transformer-CNN pairings for depression detection and a thorough analysis of model performance.

Our approach to automatically summarizing online mental health posts could help counselors by reducing their reading time, enabling quicker and more effective support for individuals seeking mental health assistance. Neural text summarization methods demonstrate promising performance owing to their strong pre-training procedure. Random token/span masking technique is often relied upon by existing pre-trained language models; an approach that overlooks the importance of content when learning word representations. In an attempt to rectify this, we propose using source and summary alignments as a saliency signal to enhance the pre-training strategy of language model for better representation learning of important content, paving the way for a positive impact on the model fine-tuning phase. Our experiments on a mental health-related dataset for user post summarization MentSum reveal improved performance, as evidenced by human evaluation metrics, surpassing the current state-of-the-art system.

This study uses machine-learning and text-mining techniques to classify social work case records to better distinguish among adolescents with four types of problems—delinquency, abuse, self-injury, and mental health problems—and identify the differential factors. We selected 573 cases recorded from a local social service organization in Shanghai, China, with 279 delinquent cases, 76 abused cases, 37 self-injured cases, and 181 cases with mental health problems. We utilized the Term Frequency-Inverse Document Frequency (TF-IDF) method to extract keywords, and trained three classification models: Naive Bayes, Decision Tree, and Random Forest. The Decision Tree model outperformed the other models with a precision of 0.9295. Based on the analysis of co-occurring keywords, we further found that adolescent delinquency was associated with early dropout from school, migrant working status, lack of parental guardianship, and negative peer influence; adolescent abuse was associated with unstable family structure and weak family support; adolescent self-injury was primarily associated with depression; and mental health problems were associated with grandparenting, low social-economic status, and transition periods. This classification model can guide tailored services and interventions, and be used as early warning system to mitigate potential risks to adolescents.

Objective: The objective of this study was to develop and interpret an explainable artificial intelligence model to predict adolescents’ psychological adjustment from coping styles, social support quality, and life stress exposure in a representative sample of Iraqi secondary school students. Methods and Materials: This cross-sectional study was conducted among 612 adolescents aged 14–18 years recruited from public secondary schools in Baghdad, Najaf, and Basra using multistage cluster sampling. Participants completed validated measures assessing psychological adjustment, coping styles, perceived social support quality, and life stress exposure. Data were preprocessed through normalization, missing-value imputation, and outlier screening. Machine learning models including random forest, extreme gradient boosting, and multilayer perceptron neural networks were trained using five-fold cross-validation. Model interpretability was achieved through SHapley Additive exPlanations, permutation feature importance, and partial dependence analyses. Findings: The XGBoost model demonstrated the highest predictive performance (R² = 0.76, RMSE = 0.29, MAE = 0.22). Social support quality emerged as the strongest positive predictor of psychological adjustment, followed by problem-focused coping. Life stress exposure exerted a substantial negative effect. Avoidance coping significantly predicted poorer adjustment, whereas emotion-focused coping displayed nonlinear effects depending on stress levels. Interaction analysis revealed that high social support significantly buffered the adverse effects of life stress on psychological adjustment. Conclusion: The findings demonstrate that adolescents’ psychological adjustment is governed by complex nonlinear interactions among coping strategies, social support, and stress exposure, and that explainable artificial intelligence offers a powerful framework for modeling these processes with high predictive accuracy and theoretical transparency.

This paper presents a system aimed at preventing the serious social issue of adolescent suicide by identifying at-risk adolescents early and providing appropriate support. To achieve this, we propose machine learning models based on data from the Korea Youth Risk Behavior Web-Based Survey (KYRBWS), targeting South Korean adolescents and considering various factors to predict suicide attempts. The models incorporate variables such as adolescents' behavioral patterns, emotional states, and family and social support systems. We applied Random Forest for feature selection to identify important features from the original dataset. The core algorithms include Logistic Regression, Random Forest, and XGBoost. To ensure robust performance, we employed K-fold cross-validation. Additionally, to address the challenge of data imbalance, we applied various sampling techniques to develop reliable, high-performance models for predicting suicide attempts. By detecting adolescents at high risk of suicide attempts, we aim to contribute to suicide prevention efforts and promote adolescent health and well-being.

The application of Social and Emotional Learning (SEL) activities in classrooms is associated with a positive impact on students. Specifically, the development of social and emotional competencies of students leads to improvements in school performance, social relationships, well-being, and mental health. Various SEL methodologies have been designed to support the development of such competencies but do not focus too much on the assessment part. In our work on this manuscript, we detail the development of a Knowledge Graph, called EmoSocioGraph, that supports the application of a SEL methodology and helps teachers to gain insights and adjust accordingly the SEL activities in the classroom to achieve better results. EmoSocioGraph is well aligned with the open-access Emotional Intelligence model, called EmoSocio, that represents the social and emotional competencies of students at the individual and group levels. It is also aligned with a systemic SEL methodology for the development of such competencies that takes advantage of open-source software tools. The main entities and relationships of EmoSocioGraph are detailed, while indicative usage and analysis scenarios are provided based on data produced by applying the systemic SEL methodology in K12 schools in Greece and Spain.

Abstract Background Converging evidence indicates an adolescent mental health crisis in Western societies that has developed and exacerbated over the past decade. The proposed driving factors of this trend include more screen time, physical inactivity, and social isolation, but their causal influence on mental health is insufficiently understood. Objective The objective of this study is to test whether and based on which predictor variables the development of mental health in adolescents in the last decade can be predicted and to better understand the causal chain of factors at work. Methods We implemented an interpretable machine learning pipeline based on gradient boosting regression with repeated cross-validation to assess the development of mental health throughout adolescence in members of 2 longitudinal cohort studies, the British Millenium cohort (MC; n=8599) and the German Health Interview and Examination Survey for Children and Adolescents (KiGGS) cohort (n=1212). In total, 144 (MC) and 102 (KiGGS) predictors assessed at the age of around 13.8 years (MC) and 11.6 years (KiGGS) were used to assess mental health at the ages of around 16.7 years (MC) and 16.4 years (KiGGS). Based on these predictive models, we used permutation-based feature importance analyses to identify relevant predictors and predictor domains. Moreover, we performed partial dependence analyses in a causal inference framework to determine the direct effects of physical inactivity, screen time, and peer problems on the development of mental health. Results The average cross-validated Pearson correlation coefficient (r) between predicted and true mental health in late adolescence was 0.614 (MC) and 0.466 (KiGGS). Feature importance analyses indicated a strong impact of preexisting mental health and weaker impacts of sex (female as a risk factor), physical health (chronic disease as a risk factor), lifestyle, and socioeconomic and family factors (eg, low parental education, income, and mental health as risk factors). Causal inference analyses suggested a strong direct effect of peer relationships, but only a small direct effect of physical inactivity and a very small direct effect of screen time. Conclusions Mental health development during adolescence can be assessed by a combination of variables from early adolescence. Peer problems represent an important direct cause of mental health development, and their deterioration may contribute to the current mental health crisis.

Background and objectives Adolescent depression and anxiety are becoming increasingly prevalent in China, with rates reaching 20%–30%, driven largely by intense academic pressure and the cultural tendency toward somatization. Traditional screening tools, such as the Patient Health Questionnaire-9 (PHQ-9) and Generalized Anxiety Disorder-7 (GAD-7), often suffer from subjective bias, recall errors, and underreporting due to social stigma. This study developed an AI-based intelligent diagnosis system (IDS) using multitask deep learning to non-intrusively predict comorbid depression and anxiety severity based on the spontaneous textual expressions of Chinese adolescents. Methods Textual responses from approximately 1,275 adolescents were collected and labeled with clinician-assessed PHQ-9 and GAD-7 scores. Preprocessing involved jieba segmentation and variational autoencoder (VAE)-based data augmentation to address class imbalance, resulting in an expanded test set of 308 samples. The IDS architecture utilizes a Chinese-optimized BERT encoder with self-attention and dual-feature fusion (combining pooled [CLS] tokens and global pooling) to extract shared representations. These are processed through multitask heads for regression (MSE loss) and classification (weighted cross-entropy). The model was trained using an 8:1:1 split with AdamW optimization, cosine annealing, and regularization, supported by ablation studies to validate individual components. Results On the test set, the IDS achieved Pearson correlation coefficients of 0.706 for PHQ-9 and 0.693 for GAD-7, with AUC values of 0.877 and 0.902, respectively. Binary classification yielded F1-scores of 0.762 (PHQ-9) and 0.863 (GAD-7). Ablation analysis confirmed that the multitask learning framework improved F1-scores by 6.2%–7.8% and reduced MSE by 14.2%–18.4%. Furthermore, adaptations for somatization and data augmentation for severe cases significantly enhanced the system’s sensitivity. Conclusion The IDS offers a robust, culturally sensitive, and scalable tool for adolescent mental health screening. By outperforming single-task baselines, it provides a proactive, privacy-preserving alternative to traditional self-reports. Future research will focus on longitudinal validation, multimodal integration, and ethical deployment strategies to maximize the system’s utility in educational and clinical settings.

The rapid growth of social media has higher concerns about its potential influence on mental health. Although social media may effect in a social connection, emotional support, etc; it has also been associated with bad significances, including anxiety, depression and stress. Nonetheless,  issue of determining particular mental health effects of social media behavior is not widely researched. The goal of this research is to model the predictions of depression, anxiety, and stress with the experience of deep-learning models trained on social media data and make comparisons with the predictions of other traditional statistical tools such as the regression analysis. The sample size was 500 participants of ages 18-30, including data about the use of social media (duration spent using social media, information interaction type: active and passive and interaction with the contents) and self-reported mental health results. The research used deep learning theories. It was revealed that passive use of social media (or scrolling with no action) was most associated with being linked with negative mental health outcomes whereas active use had some less strong effects. Furthermore, the models based on deep learning were more effective in comparison with the traditional ones, as the latter proved to have less predictive accuracy. This can be the result of the considerable improvement of the prediction of mental health outcomes with the assistance of the social media action as application of the deep learning models could be evaluated. The researcher furnished data on the effects of the trends of the social media usage on the mental health and need to facilitate the preservation of an active and healthy usage. The further studies need to concentrate on longitudinal research and use multimodal data to bias prediction model. The research paper is added to the developing sphere of computational mental health and suggests specific interventions via social media behaviors.

Student mental health has become an increasingly pressing issue across academic institutions worldwide, significantly influencing learning outcomes, emotional well-being, and interpersonal relationships within educational ecosystems. The complexity of this issue stems from its multifaceted and dynamic nature, shaped by various psychological, social, and environmental factors. Mental health challenges, including anxiety, depression, and stress, not only hinder academic performance but also disrupt students’ ability to engage meaningfully with peers and faculty, undermining the overall quality of the educational experience. Traditional approaches to assessing student mental health often rely on static surveys, periodic self-reports, or limited clinical evaluations, which are inherently insufficient in capturing the intricate, evolving, and multidimensional nature of mental well-being. Such methods frequently fail to account for critical factors such as daily stress variations, cumulative peer influences, and the temporal dynamics of mental health conditions. These limitations result in a lack of real-time insights, reducing the precision, personalization, and scalability of mental health interventions. Advancements in computational methodologies, particularly in the domains of artificial intelligence (AI), machine learning (ML), and multimodal data analytics, offer transformative potential in addressing these challenges. By integrating diverse data sources such as physiological signals, behavioral patterns, and social network interactions, computational approaches provide a holistic view of students’ mental states. These methods enable the development of adaptive and proactive intervention strategies, tailored to individual needs and responsive to real-time changes in mental health.

Peer interactions and influence affect adolescent physical activity (PA) and mental health. Social emotional learning (SEL) can be key to adolescent development and well-being. This study aimed to evaluate the ripple effects and preliminary effectiveness of EmpowerHER, an integrated PA and SEL pilot intervention, on changes in adolescent girls’ PA levels and friendship networks during an 8-week summer care program. Adolescents (n=47, ages 10–14 years) wore accelerometers to measure PA and completed surveys on psychological distress, social connectedness, and friendship nominations at the start and end of summer. EmpowerHER consisted of biweekly 90-min PA and SEL skill-building sessions over 4 weeks for 11 girls within this larger sample. Stochastic actor-oriented modeling (SIENA) analyzed selection and influence processes related to changes in PA engagement and socioemotional well-being. Girls participating in EmpowerHER significantly increased their PA levels (b = 1.57, SE = 0.73) and were more likely to send new friendship connections over time (b = 1.71, SE = 0.73). Higher social connectedness was associated with elevated PA (b = 0.31, SE = 0.12) and increased friendship connections (b = 0.39, SE = 0.18) across summer. The overall sample showed significant increases in daily MVPA (mean difference 9.04 min/day, p = .03) and friendship connections (p < .001) from start to end of summer. This exploratory study demonstrated preliminary efficacy of an integrated PA and SEL curriculum implemented through a summer care program. Participation bolstered objective PA levels and improved positive peer dynamics. The findings highlight the potential of holistic interventions addressing multiple aspects of adolescent health and development simultaneously.

BACKGROUND Adolescent mental health is influenced by various adverse environmental conditions. However, it remains unclear how these factors jointly affect adolescent depression. This study aimed to use network analysis to assess the associations between different environmental factors and depressive symptoms in adolescents and to identify key pathways between them. METHODS This study included 610 adolescents with depression from inpatient and outpatient units recruited between March 2020 and November 2021. The mean age was 14.86 ± 1.96, with no significant difference between males (n = 155, 15.10 ± 2.19) and females (n = 455, 14.78 ± 1.88). Depressive symptoms were measured using the Children's Depression Inventory, and individual risk environment factors included childhood trauma, social peer and family risk factors. Network features, including network centrality, stability, and bridge centrality, were investigated. RESULTS Anhedonia and self-esteem were found to be more central in depressive symptoms. Insult experiences from the social peer and emotional abuse experience from childhood were more central environmental factors. Childhood trauma experiences were more related to adolescent depressive symptoms compared to family and peer factors. Bridge analyses identified emotional abuse, emotional neglect and physical neglect as the main bridges linking environment risk to depressive symptoms. LIMITATIONS This was a cross-sectionally designed study, which limited its ability to examine longitudinal dynamic interactions between environmental factors and adolescent depressive symptoms. CONCLUSIONS Our findings suggested that childhood trauma experiences might have greater psychological impacts on adolescent depression than family and social peer environments, and should be considered as crucial targets for preventing severe depressive moods.

Objective Conventional scale-based diagnostic approaches are increasingly insufficient for addressing the growing mental health challenges among adolescents. Leveraging advances in artificial intelligence, this study aims to develop an accurate, efficient, and scalable model for early identification of adolescent depression risk using large-scale census data, and to identify key daily life factors associated with mental health outcomes. Methods Data were obtained from the 2021 National Survey of Children's Health, including 50,892 adolescents and 463 variables. Based on prior literature, 60 relevant variables were selected. Three progressively structured hypotheses concerning the relationships between adolescent depression and developmental environments were proposed. Machine learning models, including decision trees, XGBoost, support vector machines, and neural networks, were applied to predict depression risk. Mediation analysis was conducted to examine the pathways through which living conditions influence mental health. Results The optimal model demonstrated strong predictive performance, achieving an accuracy of 0.85 and an AUC exceeding 0.87. Feature importance analysis identified several key predictors. Mediation analysis indicated that living conditions exerted a direct effect of 0.225 on mental health, while physical activity and diet quality partially mediated this relationship. Conclusion Living conditions are critical indicators for early identification of adolescent depression risk. The use of nationwide census data enables timely screening and targeted intervention. Improving dietary habits and increasing physical activity may serve as effective preventive strategies for adolescent mental health disorders.

ABSTRACT Objective We provide proof-of-principle for a mental health risk calculator advancing clinical utility of the irritability construct for identification of young children at high risk for common, early onsetting syndromes. Method Data were harmonized from two longitudinal early childhood subsamples (total N = 403; 50.1% Male; 66.7% Nonwhite; Mage = 4.3 years). The independent subsamples were clinically enriched via disruptive behavior and violence (Subsample 1) and depression (Subsample 2). In longitudinal models, epidemiologic risk prediction methods for risk calculators were applied to test the utility of the transdiagnostic indicator, early childhood irritability, in the context of other developmental and social-ecological indicators to predict risk of internalizing/externalizing disorders at preadolescence (Mage = 9.9 years). Predictors were retained when they improved model discrimination (area under the receiver operating characteristic curve [AUC] and integrated discrimination index [IDI]) beyond the base demographic model. Results Compared to the base model, the addition of early childhood irritability and adverse childhood experiences significantly improved the AUC (0.765) and IDI slope (0.192). Overall, 23% of preschoolers went on to develop a preadolescent internalizing/externalizing disorder. For preschoolers with both elevated irritability and adverse childhood experiences, the likelihood of an internalizing/externalizing disorder was 39–66%. Conclusions Predictive analytic tools enable personalized prediction of psychopathological risk for irritable young children, holding transformative potential for clinical translation.

ObjectiveMental health problems are among the most common and debilitating health conditions in young people. Early identification is key to make support available in a timely manner. We combined administrative and longitudinal cohort study data from England to test whether routinely collected school data could aid in population-wide screening efforts. MethodData came from the UK Millennium Cohort Study (MCS), a longitudinal birth cohort study of children born between 2000-2002. Cohort data was linked the National Pupil Database, with successful linkage for n=8,671 children attending primary school in England. School readiness was measured using the Early Years Foundation Stage Profile when children were 4-5 years old. The profile is based on teacher reports of children’s personal/emotional/social development; communication and language; numeracy, literary, physical development; understanding of the world; and creative development. Mental health problems were measured through parent report at ages 11, 14 and 17, and adolescent self-report at age 17. ResultsChildren rated as achieving less than ‘good’ school readiness at age 4-5 were at greater risk of experiencing a parent-reported mental health problem in adolescence, compared to children rated as achieving ‘good’ school readiness. School readiness had incremental predictive value over and above children’s demographic information that schools routinely collect, including relative age within class, ethnicity, household and area deprivation at age 5; as well as children’s special educational needs and disabilities at age 5. Children’s development in most domains of school readiness predicted mental health problems, including in areas that have not been studied much in previous research (such as creative development). Although school readiness predicted both externalising and internalising problems, associations tended to be larger for externalising problems. ConclusionOur findings illustrate how routinely collected school data could be used to improve the prediction of young people’s mental health problems at a population level, to inform prevention and treatment planning. The findings also illustrate the potential of administrative data to advance research on the development of psychopathology.

PURPOSE Research on adolescent social media use focuses on negative mental health outcomes, with less attention on potential positive outcomes. The current study addresses this limitation by investigating associations between adolescent social media use and both psychological well-being and psychopathology. METHODS Three US-based pediatric cohort sites participating in the National Institutes of Health Environmental influences on Child Health Outcomes study contributed cross-sectional survey data. Adolescents (13-18 years) self-reported the time spent and type of (active, passive) social media use, and their psychological well-being (Patient-Reported Outcome Measurement Information System [PROMIS] Life Satisfaction and Meaning and Purpose), psychopathology (Strengths and Difficulties Questionnaire and PROMIS Depressive Symptoms), and peer relationship quality (PROMIS Peer Relationships). We estimated associations between social media use and 4 mental health groups aligned to the dual factor model of mental health (high well-being/low psychopathology; high well-being/high psychopathology; low well-being/low psychopathology; low well-being/high psychopathology), and tested interactions with peer relationships. Models were adjusted for age, sex, race, ethnicity, and family income. RESULTS Participants (N = 963) were sociodemographically diverse (22% income ≤130% federal poverty level; 42% adolescents of color). Elastic net regressions revealed more hours using social media increased the probability of being in the high psychopathology/low well-being group; adolescents with poor peer relationships spending ≥7 hours/day on social media had the greatest risk of poor mental health. Positive peer relationships were the strongest predictor of positive mental health. DISCUSSION Peer relationships were the most meaningful contribution to adolescent mental health, and quality of social media use had little influence.

Millions of people all over the globe struggle with some form of mental illness, so it’s clear that mental health is an integral part of overall health and well-being. Individuals with mental health problems have much better results when they are diagnosed and treated early on. Results from applying machine learning algorithms for the early diagnosis and treatment of mental health issues have been promising in recent years. This article explores two machine learning algorithms for mental health prediction: K-nearest neighbors (KNN) and Recurrent Neural Networks (RNN). Data on mental health patients’ demographics, health, and long-term habits was included. Based on demographics and medical history, the KNN algorithm classified people as having or not having a mental health condition, and the RNN algorithm recognized early symptoms of mental health difficulties based on behavioral patterns over time. Age, gender, and a family history of mental disease were the most critical indicators of KNN’s 96.6% mental health prediction accuracy. Changes in sleeping and social interactions were the best predictors of mental health problems, which RNN detected 78.9% of the time. Machine learning can diagnose and treat mental health disorders early, according to one study. The study emphasizes the need for diverse and high-quality datasets for machine learning systems.

Background: Adolescent depression is highly prevalent in low- and middle-income countries (LMICs). Identifying top key risk factors is necessary to inform effective prevention program design. Machine learning (ML) offers a powerful approach to analyze complex multidomain of data to identify the most relevant predictors and estimate risks for mental health problems. This paper applies ML to study risks for adolescent depression to enhance adolescent depression prevention efforts in LMICs. Methods: Six ML approaches (e.g., Explainable Boosting Machine, random forests, and XGBoost) were applied to study the risks of depression. Data were drawn from a digital health intervention study conducted in Kenya (year 2024–2025, n = 269). Multiple domains of childhood and adolescent adversity and stress experiences were used to predict adolescent depression (using PHQ9-A). Findings: We found that ML was a valuable approach in the early identification of adolescents at risk for depression. Among the six ML approaches applied, the random forest approach outperformed other ML approaches, especially when multiple domains of risks were included. We also found that childhood adversity or home adversity alone were not strong predictors for depression. Adding adolescent stress experiences and community school adversity experiences significantly improves the accuracy and predictability of depression. Using the top-15 and top-20 ranking factors, we achieved 74.8% and 75.1% accuracy in depression prediction, which was similar to the accuracy when all 49 adverse/stress factors were included in the predictive model (78.3%). Conclusions: Innovative ML and modern predictive modeling approaches have the potential to transform modern preventive mental health care by better utilizing multidomain data to identify individuals at risk for developing depression early and identify top risk factors (for targeted individuals and/or populations). Findings from ML can inform tailored intervention design to better mitigate risks in order to prevent depression problem development. They can also inform the better utilization of resources to target high-need cases and key determinants, which is particularly useful for LMICs and low-resource settings. This paper illustrates an example of how to move toward this direction. Future research is needed to validate the approach.

Previous research has shown that both early-life stressors (e.g., adverse childhood experiences) and recent stress exposure (e.g., recent life events) may contribute to the onset of depressive symptoms. However, their combined predictive effect on depression remains unclear. Using data from 2440 Chinese college students, the present study employed nine machine learning algorithms to evaluate the joint predictive roles of childhood adversity and recent stressors and to identify the most influential predictors through interpretable analyses. Results indicated that models incorporating both types of stressors achieved moderate predictive performance for depressive symptoms (R2 = 0.192-0.280; RMSE = 3.903-4.134). Key predictors included the childhood experience of being 'frequently bullied', academic-related recent stressors (e.g., 'disliking school', 'academic pressure from family', 'heavy workload', 'exam frustration', 'pressure regarding further education'), difficulties in life adjustment (e.g., 'noticeable changes in daily routines'), and interpersonal challenges (e.g., 'romantic relationship problems'). These findings highlight the importance of considering stressors from different developmental stages and offer empirical insights for the early identification and intervention of depression in college students.

ABSTRACT This study introduces a student-centred informatics framework designed to predict the risk of depression among students, with the aim of supporting early mental health interventions and personalized educational support. The framework integrates two ensemble machine learning models Logistic Regression Classifier (LRC) and Extra Trees Classifier (ETC) optimized using Subtraction-Average-Based Optimizer (SABO) and Golf Optimization Algorithm (GOA) to enhance predictive accuracy and interpretability. The dataset, collected from a large and diverse student population, includes variables related to study stress, sleep duration, and academic satisfaction factors frequently linked to student mental health. A feature sensitivity analysis using ANOVA ranks these variables by influence, offering insights into potential intervention points. The optimized ETC model (ETSB) achieved an accuracy of 0.9638 (96.38%), significantly outperforming other single and hybrid models. Beyond technical performance, this framework provides a potential digital health tool for use in school health systems or student support platforms. By identifying at-risk individuals early, the study contributes to the development of more equitable, data-informed, and preventive mental health strategies within educational environments.

No abstract available

Research has shown a link between depression risk and how gamers form relationships with their in-game figure of representation, called avatar. This is reinforced by literature supporting that a gamer's connection to their avatar may provide broader insight into their mental health. Therefore, it has been argued that if properly examined, the bond between a person and their avatar may reveal information about their current or potential struggles with depression offline. To examine whether the connection with an individuals' avatars may reveal their risk for depression, longitudinal data from 565 adults/adolescents (Mage = 29.3 years, SD = 10.6) were evaluated twice (six months apart). Participants completed the User-Avatar-Bond [UAB] scale and Depression Anxiety Stress Scale to measure avatar bond and depression risk. A series of tuned and untuned artificial intelligence [AI] classifiers analyzed their responses concurrently and prospectively. This allowed the examination of whether user-avatar bond can provide cross-sectional and predictive information about depression risk. Findings revealed that AI models can learn to accurately and automatically identify depression risk cases, based on gamers' reported UAB, age, and length of gaming involvement, both at present and six months later. In particular, random forests outperformed all other AIs, while avatar immersion was shown to be the strongest training predictor. Study outcomes demonstrate that UAB can be translated into accurate, concurrent, and future, depression risk predictions via trained AI classifiers. Assessment, prevention, and practice implications are discussed in the light of these results.

Major Depressive Disorder (MDD) is a significant global health burden that contributes to disability and reduced quality of life. Its impact extends beyond individuals, placing emotional, social, and economic strain on families and healthcare systems worldwide. Despite its prevalence, MDD remains widely misunderstood, with limited mental health literacy and persistent stigma often preventing individuals from seeking help. This research explored the prediction of MDD utilising social media data via Natural Language Processing (NLP), Machine Learning (ML), and explainable Machine Learning (xML) techniques. The research aimed at identifying depressive indicators on X (formerly Twitter) and developing interpretable models for depression risk detection. The study’s methodology followed the Cross-Industry Standard Process for Data Mining (CRISP-DM) framework to ensure a systematic approach to data analysis. Data was collected via X’s API and processed using regex-based noise removal, normalisation, tokenisation, and lemmatisation. Symptoms were mapped to DSM-5-TR criteria at the post-level, with user-level MDD risk assessed based on symptom persistence over a two-week period. Risk levels were classified as No Risk, Monitor, and High Risk to facilitate early intervention. Six ML models were trained and tested, while the Synthetic Minority Over-sampling Technique (SMOTE) was applied to mitigate class imbalance. The dataset was partitioned into training and testing sets using an 80:20 split. ML models were evaluated, and the Extreme Gradient Boosting model outperformed the others. Extreme Gradient Boosting achieved an accuracy of 0.979, F1-score of 0.970, and ROC-AUC of 0.996, surpassing benchmark results reported in prior studies. Explainability techniques, such as LIME and tree-based feature importance, enhance model transparency and clinical interpretability. Depressed mood consistently emerged as the highest-weighted predictor across different models. The findings highlight the value of aligning ML models with validated diagnostic frameworks to improve trustworthiness and reduce false positives. Future research can expand beyond text-based analysis by incorporating multimodal features to broaden diagnostic depth.

Background In this study, machine learning was used to assess the prediction of the magnitude of depression changes in college students based on various psychological variable information. Methods A group of college students from a certain school completed two assessments in October 2021 and March 2022, respectively. We collected baseline levels of depression, demographic variables, parenting styles, college students’ mental health information, personality information, coping styles, SCL-90, and social support information. We applied logistic regression, random forest, support vector machine, and k-nearest neighbor machine learning methods to predict the magnitude of depression changes in college students. We selected the best-performing model and outputted the importance of features collected at different time points. Results Whether it is predicting the magnitude of positive changes or negative changes in depression, support vector machines (SVM) had the best prediction performance (with an accuracy of 89.4% for predicting negative changes in depression and an accuracy of 91.9% for predicting positive changes in depression). The baseline level of depression, father’s emotional expression, and mother’s emotional expression were all important predictors for predicting the negative and positive changes in depression among college students. Conclusion Machine learning models can predict the extent of depression changes in college students. The baseline level of depression, as well as the emotional state of both fathers and mothers, play a significant role in predicting the negative and positive changes associated with depression in college students. This provides new insights and methods for future psychological health research and practice.

Despite rising use of machine learning (ML) methods to detect depression within social media data, few are developed with and for adolescents. This is unfortunate, because adolescents may be more likely than adults to experience somatic than emotional symptoms and may be less likely to express emotions on social media. Accordingly, ML methods that focus on emotional symptoms may undercount adolescents at risk for depression. As a step toward developing an adolescent-centered ML method, we co-developed an interview guide with Latino adolescents to understand 1) social media norms for expressing somatic and emotional symptoms; and 2) identify potential signals of each. For the latter, we adopted a novel approach of asking interviewees to take on the “human classifier” role and tell us what they look for within social media data. Using framework analysis on 43 interviews with Latino adolescents, we find evidence suggesting norms prescribe more strongly against conveying emotional symptoms than somatic symptoms on social media. Additionally, rather than literal statements conveying they are experiencing depression, adolescents appear to use audiovisual cues to signal emotional symptoms and posting behavior (time of post, posting less) for somatic symptoms. Accordingly, norms may hinder opportunities for leveraging social media data to detect depression among adolescents, particularly when using ML methods that search for literal statements of depression or signals of emotional symptoms. Because peers tend to recognize depression in an adolescent earlier than medical experts, these findings suggest the need to develop and validate ML methods that incorporate a set of signals for somatic symptoms, particularly audiovisual cues and posting behavior. We discuss the benefits of “centering at the margins,” which is focusing on a population that is understudied within this domain, and the need for ML methods developed with adolescent input.