Predicting mental health problems in adolescence using machine learning techniques

Peer Relationships Are a Direct Cause of the Adolescent Mental Health Crisis: Interpretable Machine Learning Analysis of 2 Large Cohort Studies

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.

Machine learning approaches for classifying major depressive disorder using biological and neuropsychological markers: A meta-analysis

Traditional diagnostic methods for major depressive disorder (MDD), which rely on subjective assessments, may compromise diagnostic accuracy. In contrast, machine learning models have the potential to classify and diagnose MDD more effectively, reducing the risk of misdiagnosis associated with conventional methods. The aim of this meta-analysis is to evaluate the overall classification accuracy of machine learning models in MDD and examine the effects of machine learning algorithms, biomarkers, diagnostic comparison groups, validation procedures, and participant age on classification performance. As of September 2024, a total of 176 studies were ultimately included in the meta-analysis, encompassing a total of 60,926 participants. A random-effects model was applied to analyze the extracted data, resulting in an overall classification accuracy of 0.825 (95 % CI [0.810; 0.839]). Convolutional neural networks significantly outperformed support vector machines (SVM) when using electroencephalography and magnetoencephalography data. Additionally, SVM demonstrated significantly better performance with functional magnetic resonance imaging data compared to graph neural networks and gaussian process classification. The sample size was negatively correlated to classification accuracy. Furthermore, evidence of publication bias was also detected. Therefore, while this study indicates that machine learning models show high accuracy in distinguishing MDD from healthy controls and other psychiatric disorders, further research is required before these findings can be generalized to large-scale clinical practice.

Predicting children's emotional and behavioral difficulties at age five using pregnancy and newborn risk factors: Evidence from the UK Household Longitudinal Study

Childhood emotional and behavioral difficulties have a profound impact on later life outcomes, making it crucial to identify early-life risk factors that predict emotional and behavioral difficulties. However, much of the existing research has concentrated on diagnosing, rather than predicting, emotional and behavioral difficulties, and has often focused on adolescents rather than younger children. This study employs machine learning (ML) techniques to construct an interpretable predictive model using data from the UK Household Longitudinal Study, aiming to identify key risk factors that influence children's emotional and behavioral difficulties during childhood. We examined maternal habits during pregnancy and parent-reported data on birth, breastfeeding and regulatory problems during the newborn stage. Our findings highlighted lack of breastfeeding, low birthweight and maternal smoking during pregnancy as the three most significant predictors of emotional behavioral difficulties. Other important factors were related to infant regulatory problems. Heterogeneity analysis revealed gender differences in predictive power, with maternal smoking during pregnancy being a stronger predictor for boys, and the amount of fussing in infancy having a greater impact on girls. This study highlights the importance of comprehensive prenatal and postnatal care, advocates for early screening of emotional and behavioral difficulties, and calls for gender-specific approaches in assessing and addressing emotional and behavioral difficulties in children.

Suicide risk prediction for Korean adolescents based on machine learning

Traditional clinical risk assessment tools proved inadequate for reliably identifying individuals at high risk for suicidal behavior. As a result, machine learning (ML) techniques have become progressively incorporated into psychiatric care. This study evaluates the predictive capability of national survey data, which includes factors such as lifestyle behaviors and mental health indicators, in forecasting adolescent suicidal behavior. The predictive performance of six ML models-Logistic Regression (LR), Decision Tree (DT), Support Vector Machine (SVM), Gradient Boosting Machine (GBM), Extremely Randomized Trees (ET), and Distributed Random Forest (DRF)-was systematically compared. Employed SHapley Additive exPlanations (SHAP) values and Permutation Feature Importance (PFI) for interpretability analysis, and ultimately utilized interaction analysis to examine the complex interrelationships among key variables associated with suicide risk. Both the Expert Consultation Method (ECM) and Random Forest-Based Filter Feature Selection (RFFS) datasets revealed that the GBM model achieved the best results, with a predictive accuracy (ACC) of 88%, sensitivity (SENS) of 97%, specificity (SPEC) of 26%, positive predictive value (PPV) of 90%, negative predictive value (NPV) of 56%, and an area under the curve (AUC) of 83%. Feature importance analysis identified stress and depression as the most significant determinants of suicidal ideation and behavior in middle and high school students, respectively. Multivariate interaction effect analysis further revealed that, at higher levels of depression, lower anxiety levels were significantly correlated with a reduced probability of suicide risk. Additionally, a positive association between stress and anxiety was observed. Overall, the integration of advanced computational techniques with national survey data moderately enhances the accuracy of suicide risk prediction, providing a strong empirical foundation for early intervention in adolescent suicidal behavior.

Machine learning for prediction of childhood mental health problems in social care

BACKGROUND

Rates of childhood mental health problems are increasing in the UK. Early identification of childhood mental health problems is challenging but critical to children's future psychosocial development. This is particularly important for children with social care contact because earlier identification can facilitate earlier intervention. Clinical prediction tools could improve these early intervention efforts.

AIMS

Characterise a novel cohort consisting of children in social care and develop effective machine learning models for prediction of childhood mental health problems.

METHOD

We used linked, de-identified data from the Secure Anonymised Information Linkage Databank to create a cohort of 26 820 children in Wales, UK, receiving social care services. Integrating health, social care and education data, we developed several machine learning models aimed at predicting childhood mental health problems. We assessed the performance, interpretability and fairness of these models.

RESULTS

Risk factors strongly associated with childhood mental health problems included age, substance misuse and being a looked after child. The best-performing model, a gradient boosting classifier, achieved an area under the receiver operating characteristic curve of 0.75 (95% CI 0.73-0.78). Assessments of algorithmic fairness showed potential biases within these models.

CONCLUSIONS

Machine learning performance on this prediction task was promising. Predictive performance in social care settings can be bolstered by linking diverse routinely collected data-sets, making available a range of heterogenous risk factors relating to clinical, social and environmental exposures.

Using fuzzy decision support to create a positive mental health environment for preschoolers

The preschool period is a crucial time for behavioural and social-emotional development and the cultivation of mental well-being. Preschoolers may be affected by various traumatic problems. During this process, preschoolers may develop hazardous behaviours such as defiance, aggression, speech delays, difficulty socializing, and emotional dysregulation. To assess their mental health before starting school, preschoolers need early detection, intervention, and assessment. However, data shortages, heterogeneity, privacy issues, model interpretability, and generalization restrictions hamper the review process. This study sought to improve toddlers' behaviour by creating an effective decision-making mechanism. This study uses a fuzzy decision support (FDS) system using fuzzy rules and a degree of membership function to overcome the obstacles. Fuzzified data from the Preschool Pediatric Symptom Checklist (PPSC) was utilized to study preschoolers' behavior. Follow guidelines to decrease uncertainty to get a fuzzy set value. Afterwards, de-fuzzification was done according to the membership level needed to make effective mental health decisions. The FDS process identifies the relationship between a child's behaviour and attention level with maximum accuracy (97.98%), specificity (96.79%), sensitivity (97.08%), and minimum error (0.28). Behavioural prediction helps improve preschoolers' mental health and activities effectively. The system's excellence was analyzed using different metrics, ensuring 96.79% specificity and 97.98% accuracy. The dataset used in this study may lack sufficient diversity, limiting the generalizability of the findings across different socio-economic, cultural, and demographic groups. Future work should explore integrating real-time data collection methods like wearable devices or mobile applications to gather more comprehensive and dynamic behavioural data.

Longitudinal analysis of adolescents at high risk of depression: Prediction models

BACKGROUND

This study aimed to develop a machine-learning-based predictive model to identify adolescents at high risk of depression using longitudinal analysis to determine changes in risk factors over time.

METHODS

This longitudinal study used 4 years of data from the Korea Child and Youth Panel Survey (2018-2021). The classification of high-risk depression was the outcome variable, with predictors categorized into general characteristics and personal, family, and school factors. The machine learning algorithms used in the analysis included logistic regression, support vector machine, decision tree, random forest, and extreme gradient boosting.

RESULTS

Among the 1833 adolescents classified as having a low risk of depression during the initial survey year, 27.8 % were identified as being at a high risk of depression over the subsequent 3 years. The extreme gradient boosting algorithm yielded the best performance with an area under the curve of 0.9302. The key predictors identified included violent tendencies, self-esteem, sleep duration, gender, and coercive parenting style.

CONCLUSION

A machine-learning-based predictive model for identifying adolescents at high risk of depression was developed. These findings provide a foundation for early screening and the development of intervention programs and policies aimed at mitigating adolescent depression risk.

A feasibility study of a preventative, transdiagnostic intervention for mental health problems in adolescence: building resilience through socioemotional training (ReSET)

BACKGROUND

Adolescence is a developmental period during which an estimated 75% of mental health problems emerge (Solmi et al. in Mol Psychiat 27:281-295, 2022). This paper reports a feasibility study of a novel indicated, preventative, transdiagnostic, school-based intervention: Building Resilience Through Socioemotional Training (ReSET). The intervention addresses two domains thought to be causally related to mental health problems during adolescence: social relationships and emotion processing. Social relationships were targeted using principles from interpersonal psychotherapy, while emotion processing was targeted using cognitive-emotional training focused on three areas of emotion processing: Emotion perception, emotion regulation and interoception. The aims of this feasibility study were to (i) assess the acceptability of integrating group-based psychotherapy with individual cognitive-emotional training, (ii) evaluate the feasibility of our recruitment measures, and (iii) assess the feasibility of delivering our research measures.

METHODS

The feasibility study involved 41 adolescents, aged 12-14, who were randomly assigned to receive the ReSET intervention or their school's usual mental health and wellbeing provision.

RESULTS

Qualitative data from intervention participants suggested the programme was experienced as a cohesive intervention, with participants able to draw on a combination of skills. Further, the cognitive-training tasks were received positively (with the exception of the interoception training task). The recruitment and research measures were successfully delivered in the school-based setting, with 97.5% retention of participants from baseline to post-intervention assessment. Qualitative data was overwhelmingly positive regarding the benefits to participants who had completed the intervention. Moreover, there was only limited data missingness.

CONCLUSIONS

We conclude that a trial of the ReSET intervention in a school setting is feasible. We discuss the implications of the feasibility study with regard to optimising school-based interventions and adaptations made in preparation for a full-scale randomised controlled trial, now underway.

A Machine-Learning-Based Approach to Informing Student Admission Decisions

University resources are limited, and strategic admission management is required in certain fields that have high application volumes but limited available study places. Student admission processes need to select an appropriate number of applicants to ensure the optimal enrollment while avoiding over- or underenrollment. The traditional approach often relies on the enrollment yields from previous years, assuming fixed admission probabilities for all applicants and ignoring statistical uncertainty, which can lead to suboptimal decisions. In this study, we propose a novel machine-learning-based approach to improving student admission decisions. Trained on historical application data, this approach predicts the number of enrolled applicants conditionally based on the number of admitted applicants, incorporates the statistical uncertainty of these predictions, and derives the probability of the number of enrolled applicants being larger or smaller than the available study places. The application of this approach is illustrated using empirical application data from a German university. In this illustration, first, several machine learning models were trained and compared. The best model was selected. This was then applied to applicant data for the next year to estimate the individual enrollment probabilities, which were aggregated to predict the number of applicants enrolled and the probability of this number being larger or smaller than the available study places. When this approach was compared with the traditional approach using fixed enrollment yields, the results showed that the proposed approach enables data-driven adjustments to the number of admitted applicants, ensuring controlled risk of over- and underenrollment.

Research on prediction model of adolescent suicide and self-injury behavior based on machine learning algorithm

Objective

To explore the risk factors that affect adolescents' suicidal and self-injurious behaviors and to construct a prediction model for adolescents' suicidal and self-injurious behaviors based on machine learning algorithms.

Methods

Stratified cluster sampling was used to select high school students in Chongqing, yielding 3,000 valid questionnaires. Based on whether students had engaged in suicide or self-injury, they were categorized into a suicide/self-injury group (n=78) and a non-suicide/self-injury group (n=2,922). Gender, age, insomnia, and mental illness data were compared between the two groups, and a logistic regression model was used to analyze independent risk factors for adolescent suicidal and self-injurious behavior. Six methods-multi-level perceptron, random forest, K-nearest neighbor, support vector machine, logistic regression, and extreme gradient boosting-were used to build predictive models. Various model indicators for suicidal and self-injurious behavior were compared across the six algorithms using a confusion matrix to identify the optimal model.

Result

In the self-injury and suicide groups, the proportions of male adolescents, late adolescence, insomnia, and mental illness were significantly higher than in the non-suicide and self-injury groups (p <0.05). Compared with the non-suicidal self-injury group, this group also showed significantly increased scores in cognitive subscales, impulsivity, psychoticism, introversion-extroversion, neuroticism, interpersonal sensitivity, depression, anxiety, hostility, terror, and paranoia (p <0.05). These statistically significant variables were analyzed in a logistic regression model, revealing that gender, impulsivity, psychoticism, neuroticism, interpersonal sensitivity, depression, and paranoia are independent risk factors for adolescent suicide and self-injury. The logistic regression model achieved the highest sensitivity and specificity in predicting adolescent suicide and self-injury behavior (0.9948 and 0.9981, respectively). Performance of the random forest, multi-level perceptron, and extreme gradient models was acceptable, while the K-nearest neighbor algorithm and support vector machine performed poorly.

Conclusion

The detection rate of suicidal and self-injurious behaviors is higher in women than in men. Adolescents displaying impulsiveness, psychoticism, neuroticism, interpersonal sensitivity, depression, and paranoia have a greater likelihood of engaging in such behaviors. The machine learning model for classifying and predicting adolescent suicide and self-injury risk effectively identifies these behaviors, enabling targeted interventions.

Guardian-BERT: Early detection of self-injury and suicidal signs with language technologies in electronic health reports

Mental health disorders, including non-suicidal self-injury (NSSI) and suicidal behavior, represent a growing global concern. Early detection of these conditions is crucial for timely intervention and prevention of adverse outcomes. In this study, we present Guardian-BERT (Guardian-Bidirectional Encoder Representations from Transformers), a novel approach for the early detection of NSSI and suicidal behavior in electronic health records (EHRs) using natural language processing (NLP) techniques for the Spanish language. Guardian-BERT employs a dual-domain adaptation strategy based on a pre-trained language model. The initial adaptation phase involves training on EHR discharge reports, enabling the model to learn the structure and linguistic patterns typical of clinical text. A second adaptation phase, using EHRs from the Psychiatry department of another hospital, refines the model's understanding of the specialized terminology and nuanced expressions used by mental health professionals. Empirical results show that Guardian-BERT outperforms existing pre-trained models and other supervised methods in detecting NSSI and suicidal behavior. The model achieves a more balanced trade-off between precision and recall, resulting in superior F-measure scores. Specifically, Guardian-BERT attains an F-measure of 0.95 for NSSI detection and 0.89 for suicidal behavior prediction. In addition to predictive performance, we investigated risk factors associated with these mental health conditions, identifying influences such as adverse personal circumstances and emotional distress. This analysis serves two key purposes: enhancing the interpretability of individual predictions by linking them to relevant risk factors, and enabling broader research through patient stratification and temporal studies of risk factor evolution. Our findings indicate that language technologies like Guardian-BERT offer valuable support for healthcare professionals by facilitating early detection and prevention of mental health disorders. Furthermore, the integration of risk factor analysis provides critical insights into the underlying conditions, improving both the explainability and clinical utility of predictive systems.

The application of artificial intelligence in the field of mental health: a systematic review

INTRODUCTION

The integration of artificial intelligence in mental health care represents a transformative shift in the identification, treatment, and management of mental disorders. This systematic review explores the diverse applications of artificial intelligence, emphasizing both its benefits and associated challenges.

METHODS

A comprehensive literature search was conducted across multiple databases based on Preferred Reporting Items for Systematic Reviews and Meta-Analyses, including ProQuest, PubMed, Scopus, and Persian databases, resulting in 2,638 initial records. After removing duplicates and applying strict selection criteria, 15 articles were included for analysis.

RESULTS

The findings indicate that AI enhances early detection and intervention for mental health conditions. Various studies highlighted the effectiveness of AI-driven tools, such as chatbots and predictive modeling, in improving patient engagement and tailoring interventions. Notably, tools like the Wysa app demonstrated significant improvements in user-reported mental health symptoms. However, ethical considerations regarding data privacy and algorithm transparency emerged as critical challenges.

DISCUSSION

While the reviewed studies indicate a generally positive trend in AI applications, some methodologies exhibited moderate quality, suggesting room for improvement. Involving stakeholders in the creation of AI technologies is essential for building trust and tackling ethical issues. Future studies should aim to enhance AI methods and investigate their applicability across various populations.

CONCLUSION

This review underscores the potential of AI to revolutionize mental health care through enhanced accessibility and personalized interventions. However, careful consideration of ethical implications and methodological rigor is essential to ensure the responsible deployment of AI technologies in this sensitive field.

Artificial intelligence in mental health care: a systematic review of diagnosis, monitoring, and intervention applications

Artificial intelligence (AI) has been recently applied to different mental health illnesses and healthcare domains. This systematic review presents the application of AI in mental health in the domains of diagnosis, monitoring, and intervention. A database search (CCTR, CINAHL, PsycINFO, PubMed, and Scopus) was conducted from inception to February 2024, and a total of 85 relevant studies were included according to preestablished inclusion criteria. The AI methods most frequently used were support vector machine and random forest for diagnosis, machine learning for monitoring, and AI chatbot for intervention. AI tools appeared to be accurate in detecting, classifying, and predicting the risk of mental health conditions as well as predicting treatment response and monitoring the ongoing prognosis of mental health disorders. Future directions should focus on developing more diverse and robust datasets and on enhancing the transparency and interpretability of AI models to improve clinical practice.

Enhancing prediction of major depressive disorder onset in adolescents: A machine learning approach

Major Depressive Disorder (MDD) is a prevalent mental health condition that often begins in adolescence, with significant long-term implications. Indicated prevention programs targeting adolescents with mild symptoms have shown efficacy, yet the methods for identifying at-risk individuals need improvement. This study aims to evaluate the utility of Partial Least Squares Regression (PLSR) in predicting the onset of MDD among non-depressed adolescents, compared to traditional screening methods. The study recruited 1462 Portuguese adolescents aged 13-16, who were assessed using various self-report measures and followed for two years. Participants were randomly divided into training (70%, N = 1023) and testing (30%, N = 439) samples. PLSR models were developed to predict the occurrence of a major depressive episode (MDE) within two years, using 331 variables. The model's performance was compared to the Children's Depression Inventory (CDI) in predicting MDE onset. The best-fitting PLSR model with two components explained 19.1% and 16.9% of the variance in the training and testing samples, respectively, significantly outperforming the CDI, which explained 7.7% of the variance. The area under the ROC curve was 0.78 for PLSR, compared to 0.71 for CDI. An empirically derived cut-off point was used to create dichotomous risk categories, and it showed a significant difference in MDE rates between predicted high-risk and low-risk groups. The balanced accuracy of the PLSR model was 0.77, compared to 0.65 for the CDI method. The PLSR model effectively identified adolescents at risk for developing MDD, demonstrating superior predictive power over the CDI. This study supports the potential utility of ML techniques (e.g., PLSR) in enhancing early identification and prevention efforts for adolescent depression.

Predicting adolescent psychopathology from early life factors: A machine learning tutorial

Objective

The successful implementation and interpretation of machine learning (ML) models in epidemiological studies can be challenging without an extensive programming background. We provide a didactic example of machine learning for risk prediction in this study by determining whether early life factors could be useful for predicting adolescent psychopathology.

Methods

In total, 9643 adolescents ages 9-10 from the Adolescent Brain and Cognitive Development (ABCD) Study were included in ML analysis to predict high Child Behavior Checklist (CBCL) scores (i.e., t-scores ≥ 60). ML models were constructed using a series of predictor combinations (prenatal, family history, sociodemographic) across 5 different algorithms. We assessed ML performance through sensitivity, specificity, F1-score, and area under the curve (AUC) metrics.

Results

A total of 1267 adolescents (13.1 %) were found to have high CBCL scores. The best performing algorithms were elastic net and gradient boosted trees. The best performing elastic net models included prenatal and family history factors (Sensitivity 0.654, Specificity 0.713; AUC 0.742, F1-score 0.401) and prenatal, family, history, and sociodemographic factors (Sensitivity 0.668, Specificity 0.704; AUC 0.745, F1-score 0.402). Across all 5 ML algorithms, family history factors (e.g., either parent had nervous breakdowns, trouble holding jobs/fights/police encounters, and counseling for mental issues) and sociodemographic covariates (e.g., maternal age, child's sex, caregiver income and caregiver education) tended to be better predictors of adolescent psychopathology. The most important prenatal predictors were unplanned pregnancy, birth complications, and pregnancy complications.

Conclusion

Our results suggest that inclusion of prenatal, family history, and sociodemographic factors in ML models can generate moderately accurate predictions of adolescent psychopathology. Issues associated with model overfitting, hyperparameter tuning, and system seed setting should be considered throughout model training, testing, and validation. Future early risk predictions models may improve with the inclusion of additional relevant covariates.

Prediction of adolescent depression from prenatal and childhood data from ALSPAC using machine learning

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.

Prediction of Clinical Outcomes in Psychotic Disorders Using Artificial Intelligence Methods: A Scoping Review

BACKGROUND

Psychotic disorders are major psychiatric disorders that can impact multiple domains including physical, social, and psychological functioning within individuals with these conditions. Being able to better predict the outcomes of psychotic disorders will allow clinicians to identify illness subgroups and optimize treatment strategies in a timely manner.

OBJECTIVE

In this scoping review, we aimed to examine the accuracy of the use of artificial intelligence (AI) methods in predicting the clinical outcomes of patients with psychotic disorders as well as determine the relevant predictors of these outcomes.

METHODS

This review was guided by the PRISMA Guidelines for Scoping Reviews. Seven electronic databases were searched for relevant published articles in English until 1 February 2024.

RESULTS

Thirty articles were included in this review. These studies were mainly conducted in the West (63%) and Asia (37%) and published within the last 5 years (83.3%). The clinical outcomes included symptomatic improvements, illness course, and social functioning. The machine learning models utilized data from various sources including clinical, cognitive, and biological variables such as genetic, neuroimaging measures. In terms of main machine learning models used, the most common approaches were support vector machine, random forest, logistic regression, and linear regression models. No specific machine learning approach outperformed the other approaches consistently across the studies, and an overall range of predictive accuracy was observed with an AUC from 0.58 to 0.95. Specific predictors of clinical outcomes included demographic characteristics (gender, socioeconomic status, accommodation, education, and employment); social factors (activity level and interpersonal relationships); illness features (number of relapses, duration of relapses, hospitalization rates, cognitive impairments, and negative and disorganization symptoms); treatment (prescription of first-generation antipsychotics, high antipsychotic doses, clozapine, use of electroconvulsive therapy, and presence of metabolic syndrome); and structural and functional neuroimaging abnormalities, especially involving the temporal and frontal brain regions.

CONCLUSIONS

The current review highlights the potential and need to further refine AI and machine learning models in parsing out the complex interplay of specific variables that contribute to the clinical outcome prediction of psychotic disorders.

Machine Learning, Deep Learning, and Data Preprocessing Techniques for Detecting, Predicting, and Monitoring Stress and Stress-Related Mental Disorders: Scoping Review

BACKGROUND

Mental stress and its consequent mental health disorders (MDs) constitute a significant public health issue. With the advent of machine learning (ML), there is potential to harness computational techniques for better understanding and addressing mental stress and MDs. This comprehensive review seeks to elucidate the current ML methodologies used in this domain to pave the way for enhanced detection, prediction, and analysis of mental stress and its subsequent MDs.

OBJECTIVE

This review aims to investigate the scope of ML methodologies used in the detection, prediction, and analysis of mental stress and its consequent MDs.

METHODS

Using a rigorous scoping review process with PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) guidelines, this investigation delves into the latest ML algorithms, preprocessing techniques, and data types used in the context of stress and stress-related MDs.

RESULTS

A total of 98 peer-reviewed publications were examined for this review. The findings highlight that support vector machine, neural network, and random forest models consistently exhibited superior accuracy and robustness among all ML algorithms examined. Physiological parameters such as heart rate measurements and skin response are prevalently used as stress predictors due to their rich explanatory information concerning stress and stress-related MDs, as well as the relative ease of data acquisition. The application of dimensionality reduction techniques, including mappings, feature selection, filtering, and noise reduction, is frequently observed as a crucial step preceding the training of ML algorithms.

CONCLUSIONS

The synthesis of this review identified significant research gaps and outlines future directions for the field. These encompass areas such as model interpretability, model personalization, the incorporation of naturalistic settings, and real-time processing capabilities for the detection and prediction of stress and stress-related MDs.

Building machine learning prediction models for well-being using predictors from the exposome and genome in a population cohort

Effective personalized well-being interventions require the ability to predict who will thrive or not, and the understanding of underlying mechanisms. Here, using longitudinal data of a large population cohort (the Netherlands Twin Register, collected 1991-2022), we aim to build machine learning prediction models for adult well-being from the exposome and genome, and identify the most predictive factors (N between 702 and 5874). The specific exposome was captured by parent and self-reports of psychosocial factors from childhood to adulthood, the genome was described by polygenic scores, and the general exposome was captured by linkage of participants' postal codes to objective, registry-based exposures. Not the genome (R 2 = -0.007 [-0.026-0.010]), but the general exposome (R 2 = 0.047 [0.015-0.076]) and especially the specific exposome (R 2 = 0.702 [0.637-0.753]) were predictive of well-being in an independent test set. Adding the genome (P = 0.334) and general exposome (P = 0.695) independently or jointly (P = 0.029) beyond the specific exposome did not improve prediction. Risk/protective factors such as optimism, personality, social support and neighborhood housing characteristics were most predictive. Our findings highlight the importance of longitudinal monitoring and promises of different data modalities for well-being prediction.

Machine Learning Techniques to Predict Mental Health Diagnoses: A Systematic Literature Review

Introduction

This study aims to investigate the potential of machine learning in predicting mental health conditions among college students by analyzing existing literature on mental health diagnoses using various machine learning algorithms.

Methods

The research employed a systematic literature review methodology to investigate the application of deep learning techniques in predicting mental health diagnoses among students from 2011 to 2024. The search strategy involved key terms, such as "deep learning," "mental health," and related terms, conducted on reputable repositories like IEEE, Xplore, ScienceDirect, SpringerLink, PLOS, and Elsevier. Papers published between January, 2011, and May, 2024, specifically focusing on deep learning models for mental health diagnoses, were considered. The selection process adhered to PRISMA guidelines and resulted in 30 relevant studies.

Results

The study highlights Convolutional Neural Networks (CNN), Random Forest (RF), Support Vector Machine (SVM), Deep Neural Networks, and Extreme Learning Machine (ELM) as prominent models for predicting mental health conditions. Among these, CNN demonstrated exceptional accuracy compared to other models in diagnosing bipolar disorder. However, challenges persist, including the need for more extensive and diverse datasets, consideration of heterogeneity in mental health condition, and inclusion of longitudinal data to capture temporal dynamics.

Conclusion

This study offers valuable insights into the potential and challenges of machine learning in predicting mental health conditions among college students. While deep learning models like CNN show promise, addressing data limitations and incorporating temporal dynamics are crucial for further advancements.

Factors influencing psychological distress among breast cancer survivors using machine learning techniques

Breast cancer is the most commonly diagnosed cancer among women worldwide. Breast cancer patients experience significant distress relating to their diagnosis and treatment. Managing this distress is critical for improving the lifespan and quality of life of breast cancer survivors. This study aimed to assess the level of distress in breast cancer survivors and analyze the variables that significantly affect distress using machine learning techniques. A survey was conducted with 641 adult breast cancer patients using the National Comprehensive Cancer Network Distress Thermometer tool. Participants identified various factors that caused distress. Five machine learning models were used to predict the classification of patients into mild and severe distress groups. The survey results indicated that 57.7% of the participants experienced severe distress. The top-three best-performing models indicated that depression, dealing with a partner, housing, work/school, and fatigue are the primary indicators. Among the emotional problems, depression, fear, worry, loss of interest in regular activities, and nervousness were determined as significant predictive factors. Therefore, machine learning models can be effectively applied to determine various factors influencing distress in breast cancer patients who have completed primary treatment, thereby identifying breast cancer patients who are vulnerable to distress in clinical settings.

Prediction of Junior High School Students' Problematic Internet Use: The Comparison of Neural Network Models and Linear Mixed Models in Longitudinal Study

Purpose

With the rise of big data, deep learning neural networks have garnered attention from psychology researchers due to their ability to process vast amounts of data and achieve superior model fitting. We aim to explore the predictive accuracy of neural network models and linear mixed models in tracking data when subjective variables are predominant in the field of psychology. We separately analyzed the predictive accuracy of both models and conduct a comparative study to further investigate. Simultaneously, we utilized the neural network model to examine the influencing factors of problematic internet usage and its temporal changes, attempting to provide insights for early interventions in problematic internet use.

Patients and Methods

This study compared longitudinal data of junior high school students using both a linear mixed model and a neural network model to ascertain the efficacy of these two methods in processing psychological longitudinal data.

Results

The neural network model exhibited significantly smaller errors compared to the linear mixed model. Furthermore, the outcomes from the neural network model revealed that, when analyzing data from a single time point, the influences of seventh grade better predicted Problematic Internet Use in ninth grade. And when analyzing data from multiple time points, the influences of sixth, seventh, and eighth grades more accurately predicted Problematic Internet Use in ninth grade.

Conclusion

Neural network models surpass linear mixed models in precision when predicting and analyzing longitudinal data. Furthermore, the influencing factors in lower grades provide more accurate predictions of Problematic Internet Use in higher grades. The highest prediction accuracy is attained through the utilization of data from multiple time points.

Evaluating Machine Learning Stability in Predicting Depression and Anxiety Amidst Subjective Response Errors

Major Depressive Disorder (MDD) and Generalized Anxiety Disorder (GAD) pose significant burdens on individuals and society, necessitating accurate prediction methods. Machine learning (ML) algorithms utilizing electronic health records and survey data offer promising tools for forecasting these conditions. However, potential bias and inaccuracies inherent in subjective survey responses can undermine the precision of such predictions. This research investigates the reliability of five prominent ML algorithms-a Convolutional Neural Network (CNN), Random Forest, XGBoost, Logistic Regression, and Naive Bayes-in predicting MDD and GAD. A dataset rich in biomedical, demographic, and self-reported survey information is used to assess the algorithms' performance under different levels of subjective response inaccuracies. These inaccuracies simulate scenarios with potential memory recall bias and subjective interpretations. While all algorithms demonstrate commendable accuracy with high-quality survey data, their performance diverges significantly when encountering erroneous or biased responses. Notably, the CNN exhibits superior resilience in this context, maintaining performance and even achieving enhanced accuracy, Cohen's kappa score, and positive precision for both MDD and GAD. This highlights the CNN's superior ability to handle data unreliability, making it a potentially advantageous choice for predicting mental health conditions based on self-reported data. These findings underscore the critical importance of algorithmic resilience in mental health prediction, particularly when relying on subjective data. They emphasize the need for careful algorithm selection in such contexts, with the CNN emerging as a promising candidate due to its robustness and improved performance under data uncertainties.

Machine learning-based prediction for self-harm and suicide attempts in adolescents

This study aimed to use machine learning (ML) models to predict the risk of self-harm and suicide attempts in adolescents. We conducted secondary analysis of cross-sectional data from the Longitudinal Study of Australian Children dataset. Several key variables at the age of 14-15 years were used to predict self-harm or suicide attempt at 16-17 years. Random forest classification models were used to select the optimal subset of predictors and subsequently make predictions. Among 2809 participants, 296 (10.54%) reported an act of self-harm and 145 (5.16%) reported attempting suicide at least once in the past 12 months. The area under the receiver operating curve was fair for self-harm (0.7397) and suicide attempt (0.7220), which outperformed the prediction strategy solely based on prior suicide or self-harm attempt (AUC: 0.6). The most important factors identified were similar, and included depressed feelings, strengths and difficulties questionnaire scores, perceptions of self, and school- and parent-related factors. The random forest classification algorithm, an ML technique, can effectively select the optimal subset of predictors from hundreds of variables to forecast the risks of suicide and self-harm among adolescents. Further research is needed to validate the utility and scalability of ML techniques in mental health research.

Predicting risk of suicidal behaviour after initiation of selective serotonin reuptake inhibitors in children, adolescents and young adults: protocol for development and validation of clinical prediction models

INTRODUCTION

There is concern regarding suicidal behaviour risk during selective serotonin reuptake inhibitor (SSRI) treatment among the young. A clinically useful model for predicting suicidal behaviour risk should have high predictive performance in terms of discrimination and calibration; transparency and ease of implementation are desirable.

METHODS AND ANALYSIS

Using Swedish national registers, we will identify individuals initiating an SSRI aged 8-24 years 2007-2020. We will develop: (A) a model based on a broad set of predictors, and (B) a model based on a restricted set of predictors. For the broad predictor model, we will consider an ensemble of four base models: XGBoost (XG), neural net (NN), elastic net logistic regression (EN) and support vector machine (SVM). The predictors with the greatest contribution to predictive performance in the base models will be determined. For the restricted predictor model, clinical input will be used to select predictors based on the top predictors in the broad model, and inputted in each of the XG, NN, EN and SVM models. If any show superiority in predictive performance as defined by the area under the receiver-operator curve, this model will be selected as the final model; otherwise, the EN model will be selected. The training and testing samples will consist of data from 2007 to 2017 and from 2018 to 2020, respectively. We will additionally assess the final model performance in individuals receiving a depression diagnosis within 90 days before SSRI initiation.The aims are to (A) develop a model predicting suicidal behaviour risk after SSRI initiation among children and youths, using machine learning methods, and (B) develop a model with a restricted set of predictors, favouring transparency and scalability.

ETHICS AND DISSEMINATION

The research is approved by the Swedish Ethical Review Authority (2020-06540). We will disseminate findings by publishing in peer-reviewed open-access journals, and presenting at international conferences.

Predicting Adolescent Mental Health Outcomes Across Cultures: A Machine Learning Approach

Adolescent mental health problems are rising rapidly around the world. To combat this rise, clinicians and policymakers need to know which risk factors matter most in predicting poor adolescent mental health. Theory-driven research has identified numerous risk factors that predict adolescent mental health problems but has difficulty distilling and replicating these findings. Data-driven machine learning methods can distill risk factors and replicate findings but have difficulty interpreting findings because these methods are atheoretical. This study demonstrates how data- and theory-driven methods can be integrated to identify the most important preadolescent risk factors in predicting adolescent mental health. Machine learning models examined which of 79 variables assessed at age 10 were the most important predictors of adolescent mental health at ages 13 and 17. These models were examined in a sample of 1176 families with adolescents from nine nations. Machine learning models accurately classified 78% of adolescents who were above-median in age 13 internalizing behavior, 77.3% who were above-median in age 13 externalizing behavior, 73.2% who were above-median in age 17 externalizing behavior, and 60.6% who were above-median in age 17 internalizing behavior. Age 10 measures of youth externalizing and internalizing behavior were the most important predictors of age 13 and 17 externalizing/internalizing behavior, followed by family context variables, parenting behaviors, individual child characteristics, and finally neighborhood and cultural variables. The combination of theoretical and machine-learning models strengthens both approaches and accurately predicts which adolescents demonstrate above average mental health difficulties in approximately 7 of 10 adolescents 3-7 years after the data used in machine learning models were collected.

Machine learning techniques for identifying mental health risk factor associated with schoolchildren cognitive ability living in politically violent environments

Introduction

Mental health and cognitive development are critical aspects of a child's overall well-being; they can be particularly challenging for children living in politically violent environments. Children in conflict areas face a range of stressors, including exposure to violence, insecurity, and displacement, which can have a profound impact on their mental health and cognitive development.

Methods

This study examines the impact of living in politically violent environments on the mental health and cognitive development of children. The analysis was conducted using machine learning techniques on the 2014 health behavior school children dataset, consisting of 6373 schoolchildren aged 10-15 from public and United Nations Relief and Works Agency schools in Palestine. The dataset included 31 features related to socioeconomic characteristics, lifestyle, mental health, exposure to political violence, social support, and cognitive ability. The data was balanced and weighted by gender and age.

Results

This study examines the impact of living in politically violent environments on the mental health and cognitive development of children. The analysis was conducted using machine learning techniques on the 2014 health behavior school children dataset, consisting of 6373 schoolchildren aged 10-15 from public and United Nations Relief and Works Agency schools in Palestine. The dataset included 31 features related to socioeconomic characteristics, lifestyle, mental health, exposure to political violence, social support, and cognitive ability. The data was balanced and weighted by gender and age.

Discussion

The findings can inform evidence-based strategies for preventing and mitigating the detrimental effects of political violence on individuals and communities, highlighting the importance of addressing the needs of children in conflict-affected areas and the potential of using technology to improve their well-being.

Does multidimensional daily information predict the onset of myopia? A 1-year prospective cohort study

PURPOSE

This study aimed to develop an interpretable machine learning model to predict the onset of myopia based on individual daily information.

METHOD

This study was a prospective cohort study. At baseline, non-myopia children aged 6-13 years old were recruited, and individual data were collected through interviewing students and parents. One year after baseline, the incidence of myopia was evaluated based on visual acuity test and cycloplegic refraction measurement. Five algorithms, Random Forest, Support Vector Machines, Gradient Boosting Decision Tree, CatBoost and Logistic Regression were utilized to develop different models and their performance was validated by area under curve (AUC). Shapley Additive exPlanations was applied to interpret the model output on the individual and global level.

RESULT

Of 2221 children, 260 (11.7%) developed myopia in 1 year. In univariable analysis, 26 features were associated with the myopia incidence. Catboost algorithm had the highest AUC of 0.951 in the model validation. The top 3 features for predicting myopia were parental myopia, grade and frequency of eye fatigue. A compact model using only 10 features was validated with an AUC of 0.891.

CONCLUSION

The daily information contributed reliable predictors for childhood's myopia onset. The interpretable Catboost model presented the best prediction performance. Oversampling technology greatly improved model performance. This model could be a tool in myopia preventing and intervention that can help identify children who are at risk of myopia, and provide personalized prevention strategies based on contributions of risk factors to the individual prediction result.

Methodological and Quality Flaws in the Use of Artificial Intelligence in Mental Health Research: Systematic Review

BACKGROUND

Artificial intelligence (AI) is giving rise to a revolution in medicine and health care. Mental health conditions are highly prevalent in many countries, and the COVID-19 pandemic has increased the risk of further erosion of the mental well-being in the population. Therefore, it is relevant to assess the current status of the application of AI toward mental health research to inform about trends, gaps, opportunities, and challenges.

OBJECTIVE

This study aims to perform a systematic overview of AI applications in mental health in terms of methodologies, data, outcomes, performance, and quality.

METHODS

A systematic search in PubMed, Scopus, IEEE Xplore, and Cochrane databases was conducted to collect records of use cases of AI for mental health disorder studies from January 2016 to November 2021. Records were screened for eligibility if they were a practical implementation of AI in clinical trials involving mental health conditions. Records of AI study cases were evaluated and categorized by the International Classification of Diseases 11th Revision (ICD-11). Data related to trial settings, collection methodology, features, outcomes, and model development and evaluation were extracted following the CHARMS (Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies) guideline. Further, evaluation of risk of bias is provided.

RESULTS

A total of 429 nonduplicated records were retrieved from the databases and 129 were included for a full assessment-18 of which were manually added. The distribution of AI applications in mental health was found unbalanced between ICD-11 mental health categories. Predominant categories were Depressive disorders (n=70) and Schizophrenia or other primary psychotic disorders (n=26). Most interventions were based on randomized controlled trials (n=62), followed by prospective cohorts (n=24) among observational studies. AI was typically applied to evaluate quality of treatments (n=44) or stratify patients into subgroups and clusters (n=31). Models usually applied a combination of questionnaires and scales to assess symptom severity using electronic health records (n=49) as well as medical images (n=33). Quality assessment revealed important flaws in the process of AI application and data preprocessing pipelines. One-third of the studies (n=56) did not report any preprocessing or data preparation. One-fifth of the models were developed by comparing several methods (n=35) without assessing their suitability in advance and a small proportion reported external validation (n=21). Only 1 paper reported a second assessment of a previous AI model. Risk of bias and transparent reporting yielded low scores due to a poor reporting of the strategy for adjusting hyperparameters, coefficients, and the explainability of the models. International collaboration was anecdotal (n=17) and data and developed models mostly remained private (n=126).

CONCLUSIONS

These significant shortcomings, alongside the lack of information to ensure reproducibility and transparency, are indicative of the challenges that AI in mental health needs to face before contributing to a solid base for knowledge generation and for being a support tool in mental health management.

Single classifier vs. ensemble machine learning approaches for mental health prediction

Early prediction of mental health issues among individuals is paramount for early diagnosis and treatment by mental health professionals. One of the promising approaches to achieving fully automated computer-based approaches for predicting mental health problems is via machine learning. As such, this study aims to empirically evaluate several popular machine learning algorithms in classifying and predicting mental health problems based on a given data set, both from a single classifier approach as well as an ensemble machine learning approach. The data set contains responses to a survey questionnaire that was conducted by Open Sourcing Mental Illness (OSMI). Machine learning algorithms investigated in this study include Logistic Regression, Gradient Boosting, Neural Networks, K-Nearest Neighbours, and Support Vector Machine, as well as an ensemble approach using these algorithms. Comparisons were also made against more recent machine learning approaches, namely Extreme Gradient Boosting and Deep Neural Networks. Overall, Gradient Boosting achieved the highest overall accuracy of 88.80% followed by Neural Networks with 88.00%. This was followed by Extreme Gradient Boosting and Deep Neural Networks at 87.20% and 86.40%, respectively. The ensemble classifier achieved 85.60% while the remaining classifiers achieved between 82.40 and 84.00%. The findings indicate that Gradient Boosting provided the highest classification accuracy for this particular mental health bi-classification prediction task. In general, it was also demonstrated that the prediction results produced by all of the machine learning approaches studied here were able to achieve more than 80% accuracy, thereby indicating a highly promising approach for mental health professionals toward automated clinical diagnosis.

Prediction and comparison of psychological health during COVID-19 among Indian population and Rajyoga meditators using machine learning algorithms

Issues of providing mental health support to people with emerging or current mental health disorders are becoming a significant concern throughout the world. One of the biggest effects of digital psychiatry during COVID-19 is its capacity for early identification and forecasting of a person's mental health decline resulting in chronic mental health issues. Therefore, through this study aims at addressing the hological problems by identifying people who are more likely to acquire mental health issues induced by COVID-19 epidemic. To achieve this goal, this study includes 1) Rajyoga practitioners' perceptions of psychological effects, levels of anxiety, stress, and depression are compared to those of the non practitioners 2) Predictions of mental health disorders such as stress, anxiety and depression using machine learning algorithms using the online survey data collected from Rajyoga meditators and general the population. Decision tree, random forest, naive bayeBayespport vector machine and K nearest neighbor algorithms were used for the prediction as they have been shown to be more accurate for predicting psychological disorders. The support vector machine showed the highest accuracy among all other algorithms. The f1 score was also the highest for support vector machine.

Applied machine learning to identify differential risk groups underlying externalizing and internalizing problem behaviors trajectories: A case study using a cohort of Asian American children

BACKGROUND

Internalizing and externalizing problems account for over 75% of the mental health burden in children and adolescents in the US, with higher burden among minority children. While complex interactions of multilevel factors are associated with these outcomes and may enable early identification of children in higher risk, prior research has been limited by data and application of traditional analysis methods. In this case example focused on Asian American children, we address the gap by applying data-driven statistical and machine learning methods to study clusters of mental health trajectories among children, investigate optimal predictions of children at high-risk cluster, and identify key early predictors.

METHODS

Data from the US Early Childhood Longitudinal Study 2010-2011 were used. Multilevel information provided by children, families, teachers, schools, and care-providers were considered as predictors. Unsupervised machine learning algorithm was applied to identify groups of internalizing and externalizing problems trajectories. For prediction of high-risk group, ensemble algorithm, Superlearner, was implemented by combining several supervised machine learning algorithms. Performance of Superlearner and candidate algorithms, including logistic regression, was assessed using discrimination and calibration metrics via crossvalidation. Variable importance measures along with partial dependence plots were utilized to rank and visualize key predictors.

FINDINGS

We found two clusters suggesting high- and low-risk groups for both externalizing and internalizing problems trajectories. While Superlearner had overall best discrimination performance, logistic regression had comparable performance for externalizing problems but worse for internalizing problems. Predictions from logistic regression were not well calibrated compared to those from Superlearner, however they were still better than few candidate algorithms. Important predictors identified were combination of test scores, child factors, teacher rated scores, and contextual factors, which showed non-linear associations with predicted probabilities.

CONCLUSIONS

We demonstrated the application of data-driven analytical approach to predict mental health outcomes among Asian American children. Findings from the cluster analysis can inform critical age for early intervention, while prediction analysis has potential to inform intervention programing prioritization decisions. However, to better understand external validity, replicability, and value of machine learning in broader mental health research, more studies applying similar analytical approach is needed.

Leading Predictors of COVID-19-Related Poor Mental Health in Adult Asian Indians: An Application of Extreme Gradient Boosting and Shapley Additive Explanations

During the COVID-19 pandemic, an increase in poor mental health among Asian Indians was observed in the United States. However, the leading predictors of poor mental health during the COVID-19 pandemic in Asian Indians remained unknown. A cross-sectional online survey was administered to self-identified Asian Indians aged 18 and older (N = 289). Survey collected information on demographic and socio-economic characteristics and the COVID-19 burden. Two novel machine learning techniques-eXtreme Gradient Boosting and Shapley Additive exPlanations (SHAP) were used to identify the leading predictors and explain their associations with poor mental health. A majority of the study participants were female (65.1%), below 50 years of age (73.3%), and had income ≥ $75,000 (81.0%). The six leading predictors of poor mental health among Asian Indians were sleep disturbance, age, general health, income, wearing a mask, and self-reported discrimination. SHAP plots indicated that higher age, wearing a mask, and maintaining social distancing all the time were negatively associated with poor mental health while having sleep disturbance and imputed income levels were positively associated with poor mental health. The model performance metrics indicated high accuracy (0.77), precision (0.78), F1 score (0.77), recall (0.77), and AUROC (0.87). Nearly one in two adults reported poor mental health, and one in five reported sleep disturbance. Findings from our study suggest a paradoxical relationship between income and poor mental health; further studies are needed to confirm our study findings. Sleep disturbance and perceived discrimination can be targeted through tailored intervention to reduce the risk of poor mental health in Asian Indians.

Machine learning for autism spectrum disorder diagnosis using structural magnetic resonance imaging: Promising but challenging

Autism spectrum disorder (ASD) is a complex neurodevelopmental disorder that affects approximately 1% of the population and causes significant burdens. ASD's pathogenesis remains elusive; hence, diagnosis is based on a constellation of behaviors. Structural magnetic resonance imaging (sMRI) studies have shown several abnormalities in volumetric and geometric features of the autistic brain. However, inconsistent findings prevented most contributions from being translated into clinical practice. Establishing reliable biomarkers for ASD using sMRI is crucial for the correct diagnosis and treatment. In recent years, machine learning (ML) and specifically deep learning (DL) have quickly extended to almost every sector, notably in disease diagnosis. Thus, this has led to a shift and improvement in ASD diagnostic methods, fulfilling most clinical diagnostic requirements. However, ASD discovery remains difficult. This review examines the ML-based ASD diagnosis literature over the past 5 years. A literature-based taxonomy of the research landscape has been mapped, and the major aspects of this topic have been covered. First, we provide an overview of ML's general classification pipeline and the features of sMRI. Next, representative studies are highlighted and discussed in detail with respect to methods, and biomarkers. Finally, we highlight many common challenges and make recommendations for future directions. In short, the limited sample size was the main obstacle; Thus, comprehensive data sets and rigorous methods are necessary to check the generalizability of the results. ML technologies are expected to advance significantly in the coming years, contributing to the diagnosis of ASD and helping clinicians soon.

Prediction of Mental Health Problem Using Annual Student Health Survey: A Machine Learning Approach (Preprint)

BACKGROUND

One of the reasons why students go to counseling is being called on based on self-reported health survey results. However, there is no concordant standard for such calls.

OBJECTIVE

This study aims to develop a machine learning (ML) model to predict students' mental health problems in 1 year and the following year using the health survey's content and answering time (response time, response time stamp, and answer date).

METHODS

Data were obtained from the responses of 3561 (62.58%) of 5690 undergraduate students from University A in Japan (a national university) who completed the health survey in 2020 and 2021. We performed 2 analyses; in analysis 1, a mental health problem in 2020 was predicted from demographics, answers for the health survey, and answering time in the same year, and in analysis 2, a mental health problem in 2021 was predicted from the same input variables as in analysis 1. We compared the results from different ML models, such as logistic regression, elastic net, random forest, XGBoost, and LightGBM. The results with and without answering time conditions were compared using the adopted model.

RESULTS

On the basis of the comparison of the models, we adopted the LightGBM model. In this model, both analyses and conditions achieved adequate performance (eg, Matthews correlation coefficient [MCC] of with answering time condition in analysis 1 was 0.970 and MCC of without answering time condition in analysis 1 was 0.976; MCC of with answering time condition in analysis 2 was 0.986 and that of without answering time condition in analysis 2 was 0.971). In both analyses and in both conditions, the response to the questions about campus life (eg, anxiety and future) had the highest impact (Gain 0.131-0.216; Shapley additive explanations 0.018-0.028). Shapley additive explanations of 5 to 6 input variables from questions about campus life were included in the top 10. In contrast to our expectation, the inclusion of answering time-related variables did not exhibit substantial improvement in the prediction of students' mental health problems. However, certain variables generated based on the answering time are apparently helpful in improving the prediction and affecting the prediction probability.

CONCLUSIONS

These results demonstrate the possibility of predicting mental health across years using health survey data. Demographic and behavioral data, including answering time, were effective as well as self-rating items. This model demonstrates the possibility of synergistically using the characteristics of health surveys and advantages of ML. These findings can improve health survey items and calling criteria.

Predicting relapse or recurrence of depression: systematic review of prognostic models

BACKGROUND

Relapse and recurrence of depression are common, contributing to the overall burden of depression globally. Accurate prediction of relapse or recurrence while patients are well would allow the identification of high-risk individuals and may effectively guide the allocation of interventions to prevent relapse and recurrence.

AIMS

To review prognostic models developed to predict the risk of relapse, recurrence, sustained remission, or recovery in adults with remitted major depressive disorder.

METHOD

We searched the Cochrane Library (current issue); Ovid MEDLINE (1946 onwards); Ovid Embase (1980 onwards); Ovid PsycINFO (1806 onwards); and Web of Science (1900 onwards) up to May 2021. We included development and external validation studies of multivariable prognostic models. We assessed risk of bias of included studies using the Prediction model risk of bias assessment tool (PROBAST).

RESULTS

We identified 12 eligible prognostic model studies (11 unique prognostic models): 8 model development-only studies, 3 model development and external validation studies and 1 external validation-only study. Multiple estimates of performance measures were not available and meta-analysis was therefore not necessary. Eleven out of the 12 included studies were assessed as being at high overall risk of bias and none examined clinical utility.

CONCLUSIONS

Due to high risk of bias of the included studies, poor predictive performance and limited external validation of the models identified, presently available clinical prediction models for relapse and recurrence of depression are not yet sufficiently developed for deploying in clinical settings. There is a need for improved prognosis research in this clinical area and future studies should conform to best practice methodological and reporting guidelines.

Prediction of the trajectories of depressive symptoms among children in the adolescent brain cognitive development (ABCD) study using machine learning approach

BACKGROUND

Depression often first emerges during adolescence and evidence shows that the long-term patterns of depressive symptoms over time are heterogeneous. It is meaningful to predict the trajectory of depressive symptoms in adolescents to find early intervention targets.

METHODS

Based on the Adolescent Brain Cognitive Development Study, we included 4962 participants aged 9-10 who were followed-up for 2 years. Trajectories of depressive symptoms were identified by Latent Class Growth Analyses (LCGA). Four types of machine learning models were built to predict the identified trajectories and to obtain variables with predictive value based on the best performance model.

RESULTS

Of all participants, 536 (10.80%) were classified as increasing, 269 (5.42%) as persistently high, 433 (8.73%) as decreasing, and 3724 (75.05%) as persistently low by LCGA. Gradient Boosting Machine (GBM) model got the highest discriminant performance. Sleep quality, parental emotional state and family financial adversities were the most important predictors and three resting state functional magnetic resonance imaging functional connectivity data were also helpful to distinguish trajectories.

LIMITATION

We only have depressive symptom scores at three time points. Some valuable predictors are not specific to depression. External validation is an important next step. These predictors should not be interpreted as etiology and some variables were reported by parents/caregivers.

CONCLUSION

Using GBM combined with baseline characteristics, the trajectories of depressive symptoms with two years among adolescents aged 9-10 years can be well predicted, which might further facilitate the identification of adolescents at high risk of depressive symptoms and development of effective early interventions.

Exploratory analysis using machine learning of predictive factors for falls in type 2 diabetes

We aimed to investigate the status of falls and to identify important risk factors for falls in persons with type 2 diabetes (T2D) including the non-elderly. Participants were 316 persons with T2D who were assessed for medical history, laboratory data and physical capabilities during hospitalization and given a questionnaire on falls one year after discharge. Two different statistical models, logistic regression and random forest classifier, were used to identify the important predictors of falls. The response rate to the survey was 72%; of the 226 respondents, there were 129 males and 97 females (median age 62 years). The fall rate during the first year after discharge was 19%. Logistic regression revealed that knee extension strength, fasting C-peptide (F-CPR) level and dorsiflexion strength were independent predictors of falls. The random forest classifier placed grip strength, F-CPR, knee extension strength, dorsiflexion strength and proliferative diabetic retinopathy among the 5 most important variables for falls. Lower extremity muscle weakness, elevated F-CPR levels and reduced grip strength were shown to be important risk factors for falls in T2D. Analysis by random forest can identify new risk factors for falls in addition to logistic regression.

Deep Learning-Based Text Emotion Analysis for Legal Anomie

Text emotion analysis is an effective way for analyzing the emotion of the subjects’ anomie behaviors. This paper proposes a text emotion analysis framework (called BCDF) based on word embedding and splicing. Bi-direction Convolutional Word Embedding Classification Framework (BCDF) can express the word vector in the text and embed the part of speech tagging information as a feature of sentence representation. In addition, an emotional parallel learning mechanism is proposed, which uses the temporal information of the parallel structure calculated by Bi-LSTM to update the storage information through the gating mechanism. The convolutional layer can better extract certain components of sentences (such as adjectives, adverbs, nouns, etc.), which play a more significant role in the expression of emotion. To take advantage of convolution, a Convolutional Long Short-Term Memory (ConvLSTM) network is designed to further improve the classification results. Experimental results show that compared with traditional LSTM model, the proposed text emotion analysis model has increased 3.3 and 10.9% F1 score on psychological and news text datasets, respectively. The proposed CBDM model based on Bi-LSTM and ConvLSTM has great value in practical applications of anomie behavior analysis.

Overview of CAPICE-Childhood and Adolescence Psychopathology: unravelling the complex etiology by a large Interdisciplinary Collaboration in Europe-an EU Marie Skłodowska-Curie International Training Network

The Roadmap for Mental Health and Wellbeing Research in Europe (ROAMER) identified child and adolescent mental illness as a priority area for research. CAPICE (Childhood and Adolescence Psychopathology: unravelling the complex etiology by a large Interdisciplinary Collaboration in Europe) is a European Union (EU) funded training network aimed at investigating the causes of individual differences in common childhood and adolescent psychopathology, especially depression, anxiety, and attention deficit hyperactivity disorder. CAPICE brings together eight birth and childhood cohorts as well as other cohorts from the EArly Genetics and Life course Epidemiology (EAGLE) consortium, including twin cohorts, with unique longitudinal data on environmental exposures and mental health problems, and genetic data on participants. Here we describe the objectives, summarize the methodological approaches and initial results, and present the dissemination strategy of the CAPICE network. Besides identifying genetic and epigenetic variants associated with these phenotypes, analyses have been performed to shed light on the role of genetic factors and the interplay with the environment in influencing the persistence of symptoms across the lifespan. Data harmonization and building an advanced data catalogue are also part of the work plan. Findings will be disseminated to non-academic parties, in close collaboration with the Global Alliance of Mental Illness Advocacy Networks-Europe (GAMIAN-Europe).

Mental Health Prediction Using Machine Learning: Taxonomy, Applications, and Challenges

The increase of mental health problems and the need for effective medical health care have led to an investigation of machine learning that can be applied in mental health problems. This paper presents a recent systematic review of machine learning approaches in predicting mental health problems. Furthermore, we will discuss the challenges, limitations, and future directions for the application of machine learning in the mental health field. We collect research articles and studies that are related to the machine learning approaches in predicting mental health problems by searching reliable databases. Moreover, we adhere to the PRISMA methodology in conducting this systematic review. We include a total of 30 research articles in this review after the screening and identification processes. Then, we categorize the collected research articles based on the mental health problems such as schizophrenia, bipolar disorder, anxiety and depression, posttraumatic stress disorder, and mental health problems among children. Discussing the findings, we reflect on the challenges and limitations faced by the researchers on machine learning in mental health problems. Additionally, we provide concrete recommendations on the potential future research and development of applying machine learning in the mental health field.

Listening to Mental Health Crisis Needs at Scale: Using Natural Language Processing to Understand and Evaluate a Mental Health Crisis Text Messaging Service

The current mental health crisis is a growing public health issue requiring a large-scale response that cannot be met with traditional services alone. Digital support tools are proliferating, yet most are not systematically evaluated, and we know little about their users and their needs. Shout is a free mental health text messaging service run by the charity Mental Health Innovations, which provides support for individuals in the UK experiencing mental or emotional distress and seeking help. Here we study a large data set of anonymised text message conversations and post-conversation surveys compiled through Shout. This data provides an opportunity to hear at scale from those experiencing distress; to better understand mental health needs for people not using traditional mental health services; and to evaluate the impact of a novel form of crisis support. We use natural language processing (NLP) to assess the adherence of volunteers to conversation techniques and formats, and to gain insight into demographic user groups and their behavioural expressions of distress. Our textual analyses achieve accurate classification of conversation stages (weighted accuracy = 88%), behaviours (1-hamming loss = 95%) and texter demographics (weighted accuracy = 96%), exemplifying how the application of NLP to frontline mental health data sets can aid with post-hoc analysis and evaluation of quality of service provision in digital mental health services.

Machine Learning Algorithms to Distinguish Myocardial Perfusion SPECT Polar Maps

Myocardial perfusion imaging (MPI) plays an important role in patients with suspected and documented coronary artery disease (CAD). Machine Learning (ML) algorithms have been developed for many medical applications with excellent performance. This study used ML algorithms to discern normal and abnormal gated Single Photon Emission Computed Tomography (SPECT) images. We analyzed one thousand and seven polar maps from a database of patients referred to a university hospital for clinically indicated MPI between January 2016 and December 2018. These studies were reported and evaluated by two different expert readers. The image features were extracted from a specific type of polar map segmentation based on horizontal and vertical slices. A senior expert reading was the comparator (gold standard). We used cross-validation to divide the dataset into training and testing subsets, using data augmentation in the training set, and evaluated 04 ML models. All models had accuracy >90% and area under the receiver operating characteristics curve (AUC) >0.80 except for Adaptive Boosting (AUC = 0.77), while all precision and sensitivity obtained were >96 and 92%, respectively. Random Forest had the best performance (AUC: 0.853; accuracy: 0,938; precision: 0.968; sensitivity: 0.963). ML algorithms performed very well in image classification. These models were capable of distinguishing polar maps remarkably into normal and abnormal.

Prediction models for child and adolescent mental health: A systematic review of methodology and reporting in recent research

Background

There has been a rapid growth in the publication of new prediction models relevant to child and adolescent mental health. However, before their implementation into clinical services, it is necessary to appraise the quality of their methods and reporting. We conducted a systematic review of new prediction models in child and adolescent mental health, and examined their development and validation.

Method

We searched five databases for studies developing or validating multivariable prediction models for individuals aged 18 years old or younger from 1 January 2018 to 18 February 2021. Quality of reporting was assessed using the Transparent Reporting of a multivariable prediction models for Individual Prognosis Or Diagnosis checklist, and quality of methodology using items based on expert guidance and the PROBAST tool.

Results

We identified 100 eligible studies: 41 developing a new prediction model, 48 validating an existing model and 11 that included both development and validation. Most publications (k = 75) reported a model discrimination measure, while 26 investigations reported calibration. Of 52 new prediction models, six (12%) were for suicidal outcomes, 18 (35%) for future diagnosis, five (10%) for child maltreatment. Other outcomes included violence, crime, and functional outcomes. Eleven new models (21%) were developed for use in high-risk populations. Of development studies, around a third were sufficiently statistically powered (k = 16%, 31%), while this was lower for validation investigations (k = 12, 25%). In terms of performance, the discrimination (as measured by the C-statistic) for new models ranged from 0.57 for a tool predicting ADHD diagnosis in an external validation sample to 0.99 for a machine learning model predicting foster care permanency.

Conclusions

Although some tools have recently been developed for child and adolescent mental health for prognosis and child maltreatment, none can be currently recommended for clinical practice due to a combination of methodological limitations and poor model performance. New work needs to use ensure sufficient sample sizes, representative samples, and testing of model calibration.

Can the adverse childhood experiences (ACEs) checklist be utilized to predict emergency department visits among children and adolescents?

Background

Extensive literature has shown an association of Adverse Childhood Experiences (ACEs) with adverse health outcomes; however, its ability to predict events or stratify risks is less known. Individuals with mental illness and ACE exposure have been shown to visit emergency departments (ED) more often than those in the general population. This study thus examined the ability of the ACEs checklist to predict ED visits within the subsequent year among children and adolescents presenting to mental health clinics with pre-existing mental health issues.

Methods

The study analyzed linked data (n = 6100) from two databases provided by Alberta Health Services (AHS). The Regional Access and Intake System (RAIS 2016–2018) database provided data on the predictors (ACE items, age, sex, residence, mental health program type, and primary diagnosis) regarding children and adolescents (aged 0–17 years) accessing addiction and mental health services within Calgary Zone, and the National Ambulatory Care Reporting System (NACRS 2016–2019) database provided data on ED visits. A 25% random sample of the data was reserved for validation purposes. Two Least Absolute Shrinkage and Selection Operator (LASSO) logistic regression models, each employing a different method to tune the shrinkage parameter lambda (namely cross-validated and adaptive) and performing 10-fold cross-validation for a set of 100 lambdas in each model were examined.

Results

The adaptive LASSO model had a slightly better fit in the validation dataset than the cross-validated model; however, it still demonstrated poor discrimination (AUC 0.60, sensitivity 37.8%, PPV 49.6%) and poor calibration (over-triaged in low-risk and under-triaged in high-risk subgroups). The model’s poor performance was evident from an out-of-sample deviance ratio of − 0.044.

Conclusion

The ACEs checklist did not perform well in predicting ED visits among children and adolescents with existing mental health concerns. The diverse causes of ED visits may have hindered accurate predictions, requiring more advanced statistical procedures. Future studies exploring other machine learning approaches and including a more extensive set of childhood adversities and other important predictors may produce better predictions. Furthermore, despite highly significant associations being observed, ACEs may not be deterministic in predicting health-related events at the individual level, such as general ED use.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12874-021-01392-w.

Machine learning prediction of dropping out of outpatients with alcohol use disorders

BACKGROUND

Alcohol use disorder (AUD) is a chronic disease with a higher recurrence rate than that of other mental illnesses. Moreover, it requires continuous outpatient treatment for the patient to maintain abstinence. However, with a low probability of these patients to continue outpatient treatment, predicting and managing patients who might discontinue treatment becomes necessary. Accordingly, we developed a machine learning (ML) algorithm to predict which the risk of patients dropping out of outpatient treatment schemes.

METHODS

A total of 839 patients were selected out of 2,206 patients admitted for AUD in three hospitals under the Catholic Central Medical Center in Korea. We implemented six ML models-logistic regression, support vector machine, k-nearest neighbor, random forest, neural network, and AdaBoost-and compared the prediction performances thereof.

RESULTS

Among the six models, AdaBoost was selected as the final model for recommended use owing to its area under the receiver operating characteristic curve (AUROC) of 0.72. The four variables affecting the prediction based on feature importance were the length of hospitalization, age, residential area, and diabetes.

CONCLUSION

An ML algorithm was developed herein to predict the risk of patients with AUD in Korea discontinuing outpatient treatment. By testing and validating various machine learning models, we determined the best performing model, AdaBoost, as the final model for recommended use. Using this model, clinicians can manage patients with high risks of discontinuing treatment and establish patient-specific treatment strategies. Therefore, our model can potentially enable patients with AUD to successfully complete their treatments by identifying them before they can drop out.

Early Detection of Severe Functional Impairment Among Adolescents With Major Depression Using Logistic Classifier

Machine learning is about finding patterns and making predictions from raw data. In this study, we aimed to achieve two goals by utilizing the modern logistic regression model as a statistical tool and classifier. First, we analyzed the associations between Major Depressive Episode with Severe Impairment (MDESI) in adolescents with a list of broadly defined sociodemographic characteristics. Using findings from the logistic model, the second and ultimate goal was to identify the potential MDESI cases using a logistic model as a classifier (i.e., a predictive mechanism). Data on adolescents aged 12-17 years who participated in the National Survey on Drug Use and Health (NSDUH), 2011-2017, were pooled and analyzed. The logistic regression model revealed that compared with males and adolescents aged 12-13, females and those in the age groups of 14-15 and 16-17 had higher risk of MDESI. Blacks and Asians had lower risk of MDESI than Whites. Living in single-parent household, having less authoritative parents, having negative school experiences further increased adolescents' risk of having MDESI. The predictive model successfully identified 66% of the MDESI cases (recall rate) and accurately identified 72% of the MDESI and MDESI-free cases (accuracy rate) in the training data set. The rates of both recall and accuracy remained about the same (66 and 72%) using the test data. Results from this study confirmed that the logistic model, when used as a classifier, can identify potential cases of MDESI in adolescents with acceptable recall and reasonable accuracy rates. The algorithmic identification of adolescents at risk for depression may improve prevention and intervention.