The use of machine learning in the study of suicidal and non-suicidal self-injurious thoughts and behaviors: A systematic review

Suicide attempt risk predicts inconsistent self-reported suicide attempts: A machine learning approach using longitudinal data

INTRODUCTION

Inconsistent self-reports of lifetime suicide attempts (LSAs) are a major obstacle for accurate assessment of suicidal behavior. This study is the first to posit that adolescents at higher risk report LSAs more consistently than those at lower risk, revealing a link between suicide attempt risk and consistent reporting.

METHODS

A machine learning model was trained with 70 % of the baseline assessment data of a longitudinal sample of Norwegian adolescents (n = 10,739). The model was used to estimate the LSA risk score for the remaining 30 % of the testing dataset. The relationship between these baseline risk scores and the consistency of reporting LSAs was assessed using a 2-year follow-up reassessment of the testing dataset.

RESULTS

Internalizing problems, optimism about the future, conduct problems, substance use, and disordered eating were important factors associated with suicide attempt risk. Of the participants, 63.41 % had inconsistent self-reports at the two-year follow-up. Adolescents who consistently reported LSAs had significantly higher scores of suicide attempt risk at baseline. Two logistic regression analyses confirmed an association between suicide attempt risk and inconsistent self-reported LSAs and showed that sex (being male), and lower levels of depression and conduct problems significantly predicted such inconsistencies. Those who inconsistently reported LSAs were more likely than the others to be classified by the model as false negatives at the baseline risk assessment due to their lower estimated risk scores.

LIMITATIONS

Suicide attempts were measured with a single item in this study.

CONCLUSION

These risk factors support the theory of adolescent suicidality (TAS) and could improve suicide attempt risk assessment. Inconsistent self-reported LSAs signal lower suicide attempt risk.

Psychological pain and sociodemographic factors classified suicide attempt and non-suicidal self-injury in adolescents

This study aimed to utilize machine learning to explore the psychological similarities and differences between suicide attempt (SA) and non-suicidal self-injury (NSSI), with a particular focus on the role of psychological pain. A total of 2385 middle school students were recruited using cluster sampling. The random forest algorithm was used with 25 predictors to develop classification models of SA and NSSI, respectively, and to estimate the importance scores of each predictor. Based on these scores and related theories, shared risk factors (control feature set) and distinct risk factors (distinction feature set) were selected and tested to distinguish between NSSI and SA. The machine learning algorithm exhibited fair to good performance in classifying SA history [Area Under Curves (AUCs): 0.65-0.87] and poor performance in classifying NSSI history (AUC: 0.61-0.68). The distinction feature set comprised pain avoidance, family togetherness, and deviant peer affiliation, while the control feature set included pain arousal, painful feelings, and crisis events. The distinction feature set slightly but stably outperformed the control feature set in classifying SA from NSSI. The three-dimensional psychological pain model, especially pain avoidance, might play a dominant role in understanding the similarities and differences between SA and NSSI.

Predicting post-treatment symptom severity for adults receiving psychological therapy in routine care for generalised anxiety disorder: a machine learning approach

Approximately half of generalised anxiety disorder (GAD) patients do not recover from first-line treatments, and no validated prediction models exist to inform individuals or clinicians of potential treatment benefits. This study aimed to develop and validate an accurate and explainable prediction model of post-treatment GAD symptom severity. Data from adults receiving treatment for GAD in eight Improving Access to Psychological Therapies (IAPT) services (n=15,859) were separated into training, validation and holdout datasets. Thirteen machine learning algorithms were compared using 10-fold cross-validation, against two simple clinically relevant comparison models. The best-performing model was tested on the holdout dataset and model-specific explainability measures identified the most important predictors. A Bayesian Additive Regression Trees model out-performed all comparison models (MSE=16.54 [95 % CI=15.58; 17.51]; MAE=3.19; R²=0.33, including a single predictor linear regression model: MSE=20.70 [95 % CI=19.58; 21.82]; MAE=3.94; R²=0.14). The five most important predictors were: PHQ-9 anhedonia, GAD-7 annoyance/irritability, restlessness and fear items, then the referral-assessment waiting time. The best-performing model accurately predicted post-treatment GAD symptom severity using only pre-treatment data, outperforming comparison models that approximated clinical judgement and remaining within the GAD-7 error of measurement and minimal clinically important differences. This model could inform treatment decision-making and provide desired information to clinicians and patients receiving treatment for GAD.

Role of machine learning algorithms in suicide risk prediction: a systematic review-meta analysis of clinical studies

OBJECTIVE

Suicide is a complex and multifactorial public health problem. Understanding and addressing the various factors associated with suicide is crucial for prevention and intervention efforts. Machine learning (ML) could enhance the prediction of suicide attempts.

METHOD

A systematic review was performed using PubMed, Scopus, Web of Science and SID databases. We aim to evaluate the performance of ML algorithms and summarize their effects, gather relevant and reliable information to synthesize existing evidence, identify knowledge gaps, and provide a comprehensive list of the suicide risk factors using mixed method approach.

RESULTS

Forty-one studies published between 2011 and 2022, which matched inclusion criteria, were chosen as suitable. We included studies aimed at predicting the suicide risk by machine learning algorithms except natural language processing (NLP) and image processing. The neural network (NN) algorithm exhibited the lowest accuracy at 0.70, whereas the random forest demonstrated the highest accuracy, reaching 0.94. The study assessed the COX and random forest models and observed a minimum area under the curve (AUC) value of 0.54. In contrast, the XGBoost classifier yielded the highest AUC value, reaching 0.97. These specific AUC values emphasize the algorithm-specific performance in capturing the trade-off between sensitivity and specificity for suicide risk prediction. Furthermore, our investigation identified several common suicide risk factors, including age, gender, substance abuse, depression, anxiety, alcohol consumption, marital status, income, education, and occupation. This comprehensive analysis contributes valuable insights into the multifaceted nature of suicide risk, providing a foundation for targeted preventive strategies and intervention efforts.

CONCLUSIONS

The effectiveness of ML algorithms and their application in predicting suicide risk has been controversial. There is a need for more studies on these algorithms in clinical settings, and the related ethical concerns require further clarification.

Potential suicide risk among the college student population: machine learning approaches for identifying predictors and different students' risk profiles

BACKGROUND

Suicide is one of the leading causes of death among young people and university students. Research has identified numerous socio-demographic, relational, and clinical factors as potential predictors of suicide risk, and machine learning techniques have emerged as promising ways to improve risk assessment.

OBJECTIVE

This cross-sectional observational study aimed at identifying predictors and college student profiles associated with suicide risk through a machine learning approach.

METHODS

A total of 3102 students were surveyed regarding potential suicide risk, socio-demographic characteristics, academic career, and physical/mental health and well-being. The classification tree technique and the multiple correspondence analysis were applied to define students' profiles in terms of suicide risk and to detect the main predictors of such a risk.

RESULTS

Among the participating students, 7% showed high potential suicide risk and 3.8% had a history of suicide attempts. Psychological distress and use of alcohol/substance were prominent predictors of suicide risk contributing to define the profile of high risk of suicide: students with significant psychological distress, and with medium/high-risk use of alcohol and psychoactive substances. Conversely, low psychological distress and low-risk use of alcohol and substances, together with religious practice, represented the profile of students with low risk of suicide.

CONCLUSIONS

Machine learning techniques could hold promise for assessing suicide risk in college students, potentially leading to the development of more effective prevention programs. These programs should address both risk and protective factors and be tailored to students' needs and to the different categories of risk.

Utility of Single Items within the Suicidal Behaviors Questionnaire-Revised (SBQ-R): A Bayesian Network Approach and Relative Importance Analysis

The Suicidal Behaviors Questionnaire-Revised (SBQ-R) comprises four content-specific items widely used to assess the history of suicide-related thoughts, plans or attempts, frequency of suicidal ideation, communication of intent to die by suicide and self-reported likelihood of a suicide attempt. Each item focuses on a specific parameter of the suicide-related thoughts and behaviors construct. Past research has primarily focused on the total score. This study used Bayesian network modeling and relative importance analyses on SBQ-R data from 1160 U.S. and 1141 Chinese undergraduate students. The Bayesian network analysis results showed that Item 1 is suitable for identifying other parameters of the suicide-related thoughts and behaviors construct. The results of the relative importance analysis further highlighted the relevancy of each SBQ-R item score when examining evidence for suicide-related thoughts and behaviors. These findings provided empirical support for using the SBQ-R item scores to understand the performances of different suicide-related behavior parameters. Further, they demonstrated the potential value of examining individual item-level responses to offer clinically meaningful insights. To conclude, the SBQ-R allows for the evaluation of each critical suicide-related thought and behavior parameter and the overall suicide risk.

Using recursive partitioning to predict presence and severity of suicidal ideation amongst college students

OBJECTIVE

Predicting the presence and severity of suicidal ideation in college students is important, as deaths by suicide amongst young adults have increased in the past 20 years.

PARTICIPANTS

We recruited college students (N = 5494) from ten universities across eight states.

METHOD

Participants answered three questionnaires related to lifetime and past month suicidal ideation, and an indicator of suicidal ideation in a DSM-5 symptom measure. We used recursive partitioning to predict the presence, absence, and severity, of suicidal ideation.

RESULTS

Recursive partitioning models varied in their accuracy and performance. The best-performing model consisted of predictors and outcomes measured by the DSM-5 Level 1 Cross-Cutting Symptom Measure. Sexual orientation was also an important predictor in most models.

CONCLUSIONS

A single measure of DSM-5 symptom severity may help universities understand suicide severity to promote targeted interventions. Though further work is needed, as similar scaling amongst predictors could have influenced the model.

Using cross-validation methods to select time series models: Promises and pitfalls

Vector autoregressive (VAR) modelling is widely employed in psychology for time series analyses of dynamic processes. However, the typically short time series in psychological studies can lead to overfitting of VAR models, impairing their predictive ability on unseen samples. Cross-validation (CV) methods are commonly recommended for assessing the predictive ability of statistical models. However, it is unclear how the performance of CV is affected by characteristics of time series data and the fitted models. In this simulation study, we examine the ability of two CV methods, namely,10-fold CV and blocked CV, in estimating the prediction errors of three time series models with increasing complexity (person-mean, AR, and VAR), and evaluate how their performance is affected by data characteristics. We then compare these CV methods to the traditional methods using the Akaike (AIC) and Bayesian (BIC) information criteria in their accuracy of selecting the most predictive models. We find that CV methods tend to underestimate prediction errors of simpler models, but overestimate prediction errors of VAR models, particularly when the number of observations is small. Nonetheless, CV methods, especially blocked CV, generally outperform the AIC and BIC. We conclude our study with a discussion on the implications of the findings and provide helpful guidelines for practice.

Prediction of non-suicidal self-injury in adolescents at the family level using regression methods and machine learning

BACKGROUND

Adolescent non-suicidal self-injury (NSSI) is a major public health issue. Family factors are significantly associated with NSSI in adolescents, while studies on forecasting NSSI at the family level are still limited. In addition to regression methods, machine learning (ML) techniques have been recommended to improve the accuracy of family-level risk prediction for NSSI.

METHODS

Using a dataset of 7967 students and their primary caregivers from a cross-sectional study, logistic regression model and random forest model were used to test the forecasting accuracy of NSSI predictions at the family level. Cross-validation was used to assess model prediction performance, including the area under the receiver operator curve (AUC), precision, Brier score, accuracy, sensitivity, specificity, positive predictive value and negative predictive value.

RESULTS

The top three important family-related predictors within the random forest algorithm included family function (importance:42.66), family conflict (importance:42.18), and parental depression (importance:27.21). The most significant family-related risk predictors and protective predictors identified by the logistic regression model were family history of mental illness (OR:2.25) and help-seeking behaviors of mental distress from parents (OR:0.65), respectively. The AUCs of the two models, logistic regression and random forest, were 0.852 and 0.835, respectively.

LIMITATIONS

The key limitation is that this cross-sectional survey only enabled the authors to examine predictors that were considered to be proximal rather than distal.

CONCLUSIONS

These findings highlight the significance of family-related factors in forecasting NSSI in adolescents. Combining both conventional statistical methods and ML methods to improve risk assessment of NSSI at the family level deserves attention.

Predicting adolescent suicidal behavior following inpatient discharge using structured and unstructured data

BACKGROUND

The objective was to develop and assess performance of an algorithm predicting suicide-related ICD codes within three months of psychiatric discharge.

METHODS

This prognostic study used a retrospective cohort of EHR data from 2789 youth (12 to 20 years old) hospitalized in a safety net institution in the Northeastern United States. The dataset combined structured data with unstructured data obtained through natural language processing of clinical notes. Machine learning approaches compared gradient boosting to random forest analyses.

RESULTS

Area under the ROC and precision-recall curve were 0.88 and 0.17, respectively, for the final Gradient Boosting model. The cutoff point of the model-generated predicted probabilities of suicide that optimally classified the individual as high risk or not was 0.009. When applying the chosen cutoff (0.009) to the hold-out testing set, the model correctly identified 8 positive cases out of 10, and 418 negative cases out 548. The corresponding performance metrics showed 80 % sensitivity, 76 % specificity, 6 % PPV, 99 % NPV, F-1 score of 0.11, and an accuracy of 76 %.

LIMITATIONS

The data in this study comes from a single health system, possibly introducing bias in the model's algorithm. Thus, the model may have underestimated the incidence of suicidal behavior in the study population. Further research should include multiple system EHRs.

CONCLUSIONS

These performance metrics suggest a benefit to including both unstructured and structured data in design of predictive algorithms for suicidal behavior, which can be integrated into psychiatric services to help assess risk.

Predicting Lifetime Suicide Attempts in a Community Sample of Adolescents Using Machine Learning Algorithms

Suicide is a major global health concern and a prominent cause of death in adolescents. Previous research on suicide prediction has mainly focused on clinical or adult samples. To prevent suicides at an early stage, however, it is important to screen for risk factors in a community sample of adolescents. We compared the accuracy of logistic regressions, elastic net regressions, and gradient boosting machines in predicting suicide attempts by 17-year-olds in the Millennium Cohort Study (N = 7,347), combining a large set of self- and other-reported variables from different categories. Both machine learning algorithms outperformed logistic regressions and achieved similar balanced accuracies (.76 when using data 3 years before the self-reported lifetime suicide attempts and .85 when using data from the same measurement wave). We identified essential variables that should be considered when screening for suicidal behavior. Finally, we discuss the usefulness of complex machine learning models in suicide prediction.

The use of machine learning on administrative and survey data to predict suicidal thoughts and behaviors: a systematic review

Background

Machine learning is a promising tool in the area of suicide prevention due to its ability to combine the effects of multiple risk factors and complex interactions. The power of machine learning has led to an influx of studies on suicide prediction, as well as a few recent reviews. Our study distinguished between data sources and reported the most important predictors of suicide outcomes identified in the literature.

Objective

Our study aimed to identify studies that applied machine learning techniques to administrative and survey data, summarize performance metrics reported in those studies, and enumerate the important risk factors of suicidal thoughts and behaviors identified.

Methods

A systematic literature search of PubMed, Medline, Embase, PsycINFO, Web of Science, Cumulative Index to Nursing and Allied Health Literature (CINAHL), and Allied and Complementary Medicine Database (AMED) to identify all studies that have used machine learning to predict suicidal thoughts and behaviors using administrative and survey data was performed. The search was conducted for articles published between January 1, 2019 and May 11, 2022. In addition, all articles identified in three recently published systematic reviews (the last of which included studies up until January 1, 2019) were retained if they met our inclusion criteria. The predictive power of machine learning methods in predicting suicidal thoughts and behaviors was explored using box plots to summarize the distribution of the area under the receiver operating characteristic curve (AUC) values by machine learning method and suicide outcome (i.e., suicidal thoughts, suicide attempt, and death by suicide). Mean AUCs with 95% confidence intervals (CIs) were computed for each suicide outcome by study design, data source, total sample size, sample size of cases, and machine learning methods employed. The most important risk factors were listed.

Results

The search strategy identified 2,200 unique records, of which 104 articles met the inclusion criteria. Machine learning algorithms achieved good prediction of suicidal thoughts and behaviors (i.e., an AUC between 0.80 and 0.89); however, their predictive power appears to differ across suicide outcomes. The boosting algorithms achieved good prediction of suicidal thoughts, death by suicide, and all suicide outcomes combined, while neural network algorithms achieved good prediction of suicide attempts. The risk factors for suicidal thoughts and behaviors differed depending on the data source and the population under study.

Conclusion

The predictive utility of machine learning for suicidal thoughts and behaviors largely depends on the approach used. The findings of the current review should prove helpful in preparing future machine learning models using administrative and survey data.

Systematic review registration

https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42022333454 identifier CRD42022333454.

Unveiling Adolescent Suicidality: Holistic Analysis of Protective and Risk Factors Using Multiple Machine Learning Algorithms

Adolescent suicide attempts are on the rise, presenting a significant public health concern. Recent research aimed at improving risk assessment for adolescent suicide attempts has turned to machine learning. But no studies to date have examined the performance of stacked ensemble algorithms, which are more suitable for low-prevalence conditions. The existing machine learning-based research also lacks population-representative samples, overlooks protective factors and their interplay with risk factors, and neglects established theories on suicidal behavior in favor of purely algorithmic risk estimation. The present study overcomes these shortcomings by comparing the performance of a stacked ensemble algorithm with a diverse set of algorithms, performing a holistic item analysis to identify both risk and protective factors on a comprehensive data, and addressing the compatibility of these factors with two competing theories of suicide, namely, The Interpersonal Theory of Suicide and The Strain Theory of Suicide. A population-representative dataset of 173,664 Norwegian adolescents aged 13 to 18 years (mean = 15.14, SD = 1.58, 50.5% female) with a 4.65% rate of reported suicide attempt during the past 12 months was analyzed. Five machine learning algorithms were trained for suicide attempt risk assessment. The stacked ensemble model significantly outperformed other algorithms, achieving equal sensitivity and a specificity of 90.1%, AUC of 96.4%, and AUCPR of 67.5%. All algorithms found recent self-harm to be the most important indicator of adolescent suicide attempt. Exploratory factor analysis suggested five additional risk domains, which we labeled internalizing problems, sleep disturbance, disordered eating, lack of optimism regarding future education and career, and victimization. The identified factors provided stronger support for The Interpersonal Theory of Suicide than for The Strain Theory of Suicide. An enhancement to The Interpersonal Theory based on the risk and protective factors identified by holistic item analysis is presented.

Prediction of Repeated Self-Harm in Six Months: Comparison of Traditional Psychometrics With Random Forest Algorithm

Suicidal risk has been a significant mental health problem. However, the predictive ability for repeated self-harm (SH) has not improved over the past decades. This study thus aimed to explore a potential tool with theoretical accommodation and clinical application by employing traditional logistic regression (LR) and newly developed machine learning, random forest algorithm (RF). Starting with 89 items from six commonly used scales (i.e., proximal suicide risk factors) as preliminary predictors, both LR and RF resulted in a better solution with much fewer items in two phases of item selections and analyses, with prediction accuracy 88.6% and 79.8%, respectively. A combination with 12 selected items, named LR-12, well predicted repeated self-harm in 6-month follow-up with satisfactory performance (AUC = 0.84, 95% CI: 0.76-0.92; cut-off point by 1/2 with sensitivity 81.1% and specificity 74.0%). The psychometrically appealing LR-12 could be used as a screening scale for suicide risk assessment.

Machine Learning-Based Prediction of Suicidality in Adolescents With Allergic Rhinitis: Derivation and Validation in 2 Independent Nationwide Cohorts

BACKGROUND

Given the additional risk of suicide-related behaviors in adolescents with allergic rhinitis (AR), it is important to use the growing field of machine learning (ML) to evaluate this risk.

OBJECTIVE

This study aims to evaluate the validity and usefulness of an ML model for predicting suicide risk in patients with AR.

METHODS

We used data from 2 independent survey studies, Korea Youth Risk Behavior Web-based Survey (KYRBS; n=299,468) for the original data set and Korea National Health and Nutrition Examination Survey (KNHANES; n=833) for the external validation data set, to predict suicide risks of AR in adolescents aged 13 to 18 years, with 3.45% (10,341/299,468) and 1.4% (12/833) of the patients attempting suicide in the KYRBS and KNHANES studies, respectively. The outcome of interest was the suicide attempt risks. We selected various ML-based models with hyperparameter tuning in the discovery and performed an area under the receiver operating characteristic curve (AUROC) analysis in the train, test, and external validation data.

RESULTS

The study data set included 299,468 (KYRBS; original data set) and 833 (KNHANES; external validation data set) patients with AR recruited between 2005 and 2022. The best-performing ML model was the random forest model with a mean AUROC of 84.12% (95% CI 83.98%-84.27%) in the original data set. Applying this result to the external validation data set revealed the best performance among the models, with an AUROC of 89.87% (sensitivity 83.33%, specificity 82.58%, accuracy 82.59%, and balanced accuracy 82.96%). While looking at feature importance, the 5 most important features in predicting suicide attempts in adolescent patients with AR are depression, stress status, academic achievement, age, and alcohol consumption.

CONCLUSIONS

This study emphasizes the potential of ML models in predicting suicide risks in patients with AR, encouraging further application of these models in other conditions to enhance adolescent health and decrease suicide rates.

A Randomized Control Trial of a Digital Health Tool for Safer Firearm and Medication Storage for Patients with Suicide Risk

Most patients with suicide risk do not receive recommendations to reduce access to lethal means due to a variety of barriers (e.g., lack of provider time, training). Determine if highly efficient population-based EHR messaging to visit the Lock to Live (L2L) decision aid impacts patient-reported storage behaviors. Randomized trial. Integrated health care system serving Denver, CO. Served by primary care or mental health specialty clinic in the 75-99.5th risk percentile on a suicide attempt or death prediction model. Lock to Live (L2L) is a web-based decision aid that incorporates patients' values into recommendations for safe storage of lethal means, including firearms and medications. Anonymous survey that determined readiness to change: pre-contemplative (do not believe in safe storage), contemplative (believe in safe storage but not doing it), preparation (planning storage changes) or action (safely storing). There were 21,131 patients randomized over a 6-month period with a 27% survey response rate. Many (44%) had access to a firearm, but most of these (81%) did not use any safe firearm storage behaviors. Intervention patients were more likely to be categorized as preparation or action compared to controls for firearm storage (OR = 1.30 (1.07-1.58)). When examining action alone, there were no group differences. There were no statistically significant differences for any medication storage behaviors. Selection bias in those who responded to survey. Efficiently sending an EHR invitation message to visit L2L encouraged patients with suicide risk to consider safer firearm storage practices, but a stronger intervention is needed to change storage behaviors. Future studies should evaluate whether combining EHR messaging with provider nudges (e.g., brief clinician counseling) changes storage behavior.ClinicalTrials.gov: NCT05288517.

Sequential autoencoders for feature engineering and pretraining in major depressive disorder risk prediction

Objectives

We evaluated autoencoders as a feature engineering and pretraining technique to improve major depressive disorder (MDD) prognostic risk prediction. Autoencoders can represent temporal feature relationships not identified by aggregate features. The predictive performance of autoencoders of multiple sequential structures was evaluated as feature engineering and pretraining strategies on an array of prediction tasks and compared to a restricted Boltzmann machine (RBM) and random forests as a benchmark.

Materials and Methods

We study MDD patients from Vanderbilt University Medical Center. Autoencoder models with Attention and long-short-term memory (LSTM) layers were trained to create latent representations of the input data. Predictive performance was evaluated temporally by fitting random forest models to predict future outcomes with engineered features as input and using autoencoder weights to initialize neural network layers. We evaluated area under the precision-recall curve (AUPRC) trends and variation over the study population's treatment course.

Results

The pretrained LSTM model improved predictive performance over pretrained Attention models and benchmarks in 3 of 4 outcomes including self-harm/suicide attempt (AUPRCs, LSTM pretrained = 0.012, Attention pretrained = 0.010, RBM = 0.009, random forest = 0.005). The use of autoencoders for feature engineering had varied results, with benchmarks outperforming LSTM and Attention encodings on the self-harm/suicide attempt outcome (AUPRCs, LSTM encodings = 0.003, Attention encodings = 0.004, RBM = 0.009, random forest = 0.005).

Discussion

Improvement in prediction resulting from pretraining has the potential for increased clinical impact of MDD risk models. We did not find evidence that the use of temporal feature encodings was additive to predictive performance in the study population. This suggests that predictive information retained by model weights may be lost during encoding. LSTM pretrained model predictive performance is shown to be clinically useful and improves over state-of-the-art predictors in the MDD phenotype. LSTM model performance warrants consideration of use in future related studies.

Conclusion

LSTM models with pretrained weights from autoencoders were able to outperform the benchmark and a pretrained Attention model. Future researchers developing risk models in MDD may benefit from the use of LSTM autoencoder pretrained weights.

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.

Using electronic health record data to predict future self-harm or suicidal ideation in young people treated by child and youth mental health services

INTRODUCTION

Identifying young people who are at risk of self-harm or suicidal ideation (SHoSI) is a priority for mental health clinicians. We explore the utility of routinely collected data in developing a tool to aid early identification of those at risk.

METHOD

We used electronic health records of 4610 young people aged 5-19 years who were treated by Child and Youth Mental Health Services (CYMHS) in greater Brisbane, Australia. Two Lasso models were trained to predict the risk of future SHoSI in young people currently rated SHoSI; and those who were not.

RESULTS

For currently non-SHoSI children, an Area Under the Receiver Operating Characteristics (AUC) of 0.78 was achieved. Those with the highest risk were 4.97 (CI 4.35-5.66) times more likely to be categorized as SHoSI in the future. For current SHoSI children, the AUC was 0.62.

CONCLUSION

A prediction model with fair overall predictive power for currently non-SHoSI children was generated. Predicting persistence for SHoSI was more difficult. The electronic health records alone were not sufficient to discriminate at acceptable levels and may require adding unstructured data such as clinical notes. To optimally predict SHoSI models need to be tested and validated separately for those young people with varying degrees of risk.

Predicting suicide attempts among Norwegian adolescents without using suicide-related items: a machine learning approach

Introduction

Research on the classification models of suicide attempts has predominantly depended on the collection of sensitive data related to suicide. Gathering this type of information at the population level can be challenging, especially when it pertains to adolescents. We addressed two main objectives: (1) the feasibility of classifying adolescents at high risk of attempting suicide without relying on specific suicide-related survey items such as history of suicide attempts, suicide plan, or suicide ideation, and (2) identifying the most important predictors of suicide attempts among adolescents.

Methods

Nationwide survey data from 173,664 Norwegian adolescents (ages 13-18) were utilized to train a binary classification model, using 169 questionnaire items. The Extreme Gradient Boosting (XGBoost) algorithm was fine-tuned to classify adolescent suicide attempts, and the most important predictors were identified.

Results

XGBoost achieved a sensitivity of 77% with a specificity of 90%, and an AUC of 92.1% and an AUPRC of 47.1%. A coherent set of predictors in the domains of internalizing problems, substance use, interpersonal relationships, and victimization were pinpointed as the most important items related to recent suicide attempts.

Conclusion

This study underscores the potential of machine learning for screening adolescent suicide attempts on a population scale without requiring sensitive suicide-related survey items. Future research investigating the etiology of suicidal behavior may direct particular attention to internalizing problems, interpersonal relationships, victimization, and substance use.

Interpersonal Well-Being and Suicidal Outcomes in a Nationally Representative Study of Adolescents: A Translational Study

Adolescent suicide continues to rise despite burgeoning research on interpersonal risk for suicide. This may reflect challenges in applying developmental psychopathology research into clinical settings. In response, the present study used a translational analytic plan to examine indices of social well-being most accurate and statistically fair for indexing adolescent suicide. Data from the National Comorbidity Survey Replication Adolescent Supplement were used. Adolescents aged 13-17 (N = 9,900) completed surveys on traumatic events, current relationships, and suicidal thoughts and attempts. Both frequentist (e.g., receiver operating characteristics) and Bayesian (e.g., Diagnostic Likelihood Ratios; DLRs) techniques provided insight into classification, calibration, and statistical fairness. Final algorithms were compared to a machine learning-informed algorithm. Overall, parental care and family cohesion best classified suicidal ideation, while these indices and school engagement best classified attempts. Multi-indicator algorithms suggested adolescents at high risk across these indices were approximately 3-times more likely to engage in ideation (DLR = 3.26) and 5-times more likely to engage in attempts (DLR = 4.53). Although equitable for attempts, models for ideation underperformed in non-White adolescents. Supplemental, machine learning-informed algorithms performed similarly, suggesting non-linear and interactive effects did not improve model performance. Future directions for interpersonal theories for suicide are discussed and clinical implications for suicide screening are demonstrated.

Assessment of machine learning algorithms in national data to classify the risk of self-harm among young adults in hospital: A retrospective study

BACKGROUND

Self-harm is one of the most common presentations at accident and emergency departments in the UK and is a strong predictor of suicide risk. The UK Government has prioritised identifying risk factors and developing preventative strategies for self-harm. Machine learning offers a potential method to identify complex patterns with predictive value for the risk of self-harm.

METHODS

National data in the UK Mental Health Services Data Set were isolated for patients aged 18-30 years who started a mental health hospital admission between Aug 1, 2020 and Aug 1, 2021, and had been discharged by Jan 1, 2022. Data were obtained on age group, gender, ethnicity, employment status, marital status, accommodation status and source of admission to hospital and used to construct seven machine learning models that were used individually and as an ensemble to predict hospital stays that would be associated with a risk of self-harm.

OUTCOMES

The training dataset included 23 808 items (including 1081 episodes of self-harm) and the testing dataset 5951 items (including 270 episodes of self-harm). The best performing algorithms were the random forest model (AUC-ROC 0.70, 95%CI:0.66-0.74) and the ensemble model (AUC-ROC 0.77 95%CI:0.75-0.79).

INTERPRETATION

Machine learning algorithms could predict hospital stays with a high risk of self-harm based on readily available data that are routinely collected by health providers and recorded in the Mental Health Services Data Set. The findings should be validated externally with other real-world, prospective data.

FUNDING

This study was supported by the Midlands and Lancashire Commissioning Support Unit.

Using machine learning with intensive longitudinal data to predict depression and suicidal ideation among medical interns over time

BACKGROUND

Use of intensive longitudinal methods (e.g. ecological momentary assessment, passive sensing) and machine learning (ML) models to predict risk for depression and suicide has increased in recent years. However, these studies often vary considerably in length, ML methods used, and sources of data. The present study examined predictive accuracy for depression and suicidal ideation (SI) as a function of time, comparing different combinations of ML methods and data sources.

METHODS

Participants were 2459 first-year training physicians (55.1% female; 52.5% White) who were provided with Fitbit wearable devices and assessed daily for mood. Linear [elastic net regression (ENR)] and non-linear (random forest) ML algorithms were used to predict depression and SI at the first-quarter follow-up assessment, using two sets of variables (daily mood features only, daily mood features + passive-sensing features). To assess accuracy over time, models were estimated iteratively for each of the first 92 days of internship, using data available up to that point in time.

RESULTS

ENRs using only the daily mood features generally had the best accuracy for predicting mental health outcomes, and predictive accuracy within 1 standard error of the full 92 day models was attained by weeks 7-8. Depression at 92 days could be predicted accurately (area under the curve >0.70) after only 14 days of data collection.

CONCLUSIONS

Simpler ML methods may outperform more complex methods until passive-sensing features become better specified. For intensive longitudinal studies, there may be limited predictive value in collecting data for more than 2 months.

A machine learning analysis of risk and protective factors of suicidal thoughts and behaviors in college students

OBJECTIVE

To identify robust and reproducible factors associated with suicidal thoughts and behaviors (STBs) in college students.

METHODS

356 first-year university students completed a large battery of demographic and clinically-relevant self-report measures during the first semester of college and end-of-year (n = 228). Suicide Behaviors Questionnaire-Revised (SBQ-R) assessed STBs. A machine learning (ML) pipeline using stacking and nested cross-validation examined correlates of SBQ-R scores.

RESULTS

9.6% of students were identified at significant STBs risk by the SBQ-R. The ML algorithm explained 28.3% of variance (95%CI: 28-28.5%) in baseline SBQ-R scores, with depression severity, social isolation, meaning and purpose in life, and positive affect among the most important factors. There was a significant reduction in STBs at end-of-year with only 1.8% of students identified at significant risk.

CONCLUSION

Analyses replicated known factors associated with STBs during the first semester of college and identified novel, potentially modifiable factors including positive affect and social connectedness.

Suicide risk classification with machine learning techniques in a large Brazilian community sample

Even though suicide is a relatively preventable poor outcome, its prediction remains an elusive task. The main goal of this study was to develop machine learning classifiers to identify increased suicide risk in Brazilians with common mental disorders. With the use of clinical and sociodemographic baseline data (n = 4039 adult participants) from a large Brazilian community sample, we developed several models (Elastic Net, Random Forests, Naïve Bayes, and ensemble) for the classification of increased suicide risk among individuals with common mental disorders. 1120 participants (27.7%) presented increased suicide risk. The Random Forests model achieved the best AUC ROC (0.814), followed by Naive Bayes (0.798) and Elastic Net (0.773). Sensitivity varied from 0.922 (Naive Bayes) to 0.630 (Random Forests), while specificity varied from 0.792 (Random Forests) to 0.473 (Naive Bayes). The ensemble model presented an AUC ROC of 0.811, sensitivity of 0.899, and specificity of 0.510. Features representing depression symptoms were the most relevant for the classification of increased suicide risk. Some of our models presented good performance metrics in the classification of increased suicide risk in the investigated sample, which can provide the means to early preventive interventions.

Predictors of suicide ideation among South Korean adolescents: A machine learning approach

BACKGROUND

The current study developed a predictive model for suicide ideation among South Korean (Korean) adolescents using a comprehensive set of factors across demographic, physical and mental health, academic, social, and behavioral domains. The aim of this study was to address the pressing public health concerns of adolescent suicide in Korea and the methodological limitations of suicidal research.

METHODS

This study used machine learning methods (decision tree, logistic regression, naive Bayes classifier) to improve the accuracy of predicting suicidal ideation and related factors among a nationally representative sample of Korean middle school students (N = 6666).

RESULTS

Factors within all domains, including demographic characteristics, physical and mental health, and academic, social, and behavioral, were important in predicting suicidal thoughts among Korean adolescents, with mental health being the most important factor.

LIMITATIONS

The predictive model of the current research does not infer causality, and there may have been some loss of information due to measurement issues.

CONCLUSIONS

Study results provide insights for taking a multidimensional approach when identifying adolescents at risk of suicide, which may be used to further address their needs through intervention programs within the school setting. Considering the cultural stigma attached to disclosing suicidal ideation and behavior, the current study proposes the need for a preventive screening process based on the observation and assessment of adolescents' general characteristics and experiences in everyday life.

Predicting short-term suicidal thoughts in adolescents using machine learning: developing decision tools to identify daily level risk after hospitalization

BACKGROUND

Mobile technology offers unique opportunities for monitoring short-term suicide risk in daily life. In this study of suicidal adolescent inpatients, theoretically informed risk factors were assessed daily following discharge to predict near-term suicidal ideation and inform decision algorithms for identifying elevations in daily level risk, with implications for real-time suicide-focused interventions.

METHODS

Adolescents (N = 78; 67.9% female) completed brief surveys texted daily for 4 weeks after discharge (n = 1621 observations). Using multi-level classification and regression trees (CARTSs) with repeated 5-fold cross-validation, we tested (a) a simple prediction model incorporating previous-day scores for each of 10 risk factors, and (b) a more complex model incorporating, for each of these factors, a time-varying person-specific mean over prior days together with deviation from that mean. Models also incorporated missingness and contextual (study week, day of the week) indicators. The outcome was the presence/absence of next-day suicidal ideation.

RESULTS

The best-performing model (cross-validated AUC = 0.86) was a complex model that included ideation duration, hopelessness, burdensomeness, and self-efficacy to refrain from suicidal action. An equivalent model that excluded ideation duration had acceptable overall performance (cross-validated AUC = 0.78). Models incorporating only previous-day scores, with and without ideation duration (cross-validated AUC of 0.82 and 0.75, respectively), showed relatively weaker performance.

CONCLUSIONS

Results suggest that specific combinations of dynamic risk factors assessed in adolescents' daily life have promising utility in predicting next-day suicidal thoughts. Findings represent an important step in the development of decision tools identifying short-term risk as well as guiding timely interventions sensitive to proximal elevations in suicide risk in daily life.

Binge Eating, Purging, and Restriction Symptoms: Increasing Accuracy of Prediction Using Machine Learning

Eating disorders are severe mental illnesses characterized by the hallmark behaviors of binge eating, restriction, and purging. These disordered eating behaviors carry extreme impairment and medical complications, regardless of eating disorder diagnosis. Despite the importance of these disordered behaviors to every eating disorder diagnosis, our current models are not able to accurately predict behavior occurrence. The current study utilized machine learning to develop longitudinal predictive models of binge eating, purging, and restriction in an eating disorder sample (N = 60) using real-time intensive longitudinal data. Participants completed four daily assessments of eating disorder symptoms and emotions for 25 days on a smartphone (total data points per participant = 100). Using data, we were able to compute highly accurate prediction models for binge eating, restriction, and purging (.76-.96 accuracy). The ability to accurately predict the occurrence of binge eating, restriction, and purging has crucial implications for the development of preventative interventions for the eating disorders. Machine learning models may be able to accurately predict onset of problematic psychiatric behaviors leading to preventative interventions designed to disrupt engagement in such behaviors.

Using Machine Learning to Predict Cognitive Impairment Among Middle-Aged and Older Chinese: A Longitudinal Study

Objective: To explore the predictive value of machine learning in cognitive impairment, and identify important factors for cognitive impairment. Methods: A total of 2,326 middle-aged and elderly people completed questionnaire, and physical examination evaluation at baseline, Year 2, and Year 4 follow-ups. A random forest machine learning (ML) model was used to predict the cognitive impairment at Year 2 and Year 4 longitudinally. Based on Year 4 cross-sectional data, the same method was applied to establish a prediction model and verify its longitudinal prediction accuracy for cognitive impairment. Meanwhile, the ability of random forest and traditional logistic regression model to longitudinally predict 2-year and 4-year cognitive impairment was compared. Results: Random forest models showed high accuracy for all outcomes at Year 2, Year 4, and cross-sectional Year 4 [AUC = 0.81, 0.79, 0.80] compared with logistic regression [AUC = 0.61, 0.62, 0.70]. Baseline physical examination (e.g., BMI, Blood pressure), biomarkers (e.g., cholesterol), functioning (e.g., functional limitations), demography (e.g., age), and emotional status (e.g., depression) characteristics were identified as the top ten important predictors of cognitive impairment. Conclusion: ML algorithms could enhance the prediction of cognitive impairment among the middle-aged and older Chinese for 4 years and identify essential risk markers.

An explainable predictive model for suicide attempt risk using an ensemble learning and Shapley Additive Explanations (SHAP) approach

Machine learning approaches have been used to develop suicide attempt predictive models recently and have been shown to have a good performance. However, those proposed models have difficulty interpreting and understanding why an individual has suicidal attempts. To overcome this issue, the identification of features such as risk factors in predicting suicide attempts is important for clinicians to make decisions. Therefore, the aim of this study is to propose an explainable predictive model to predict and analyse the importance of features for suicide attempts. This model can also provide explanations to improve the clinical understanding of suicide attempts. Two complex ensemble learning models, namely Random Forest and Gradient Boosting with an explanatory model (SHapley Additive exPlanations (SHAP)) have been constructed. The models are used for predictive interpretation and understanding of the importance of the features. The experiment shows that both models with SHAP are able to interpret and understand the nature of an individual's predictions with suicide attempts. However, compared with Random Forest, the results show that Gradient Boosting with SHAP achieves higher accuracy and the analyses found that history of suicide attempts, suicidal ideation, and ethnicity as the main predictors for suicide attempts.

NSSI questionnaires revisited: A data mining approach to shorten the NSSI questionnaires

BACKGROUND AND OBJECTIVE

Non-suicidal self-injury (NSSI) is a psychological disorder that the sufferer consciously damages their body tissues, often too severe that requires intensive care medicine. As some individuals hide their NSSI behaviors, other people can only identify them if they catch them while injuring, or via dedicated questionnaires. However, questionnaires are long and tedious to answer, thus the answers might be inconsistent. Hence, in this study for the first time, we abstracted a larger questionnaire (of 662 items in total) to own only 22 items (questions) via data mining techniques. Then, we trained several machine learning algorithms to classify individuals based on their answers into two classes.

METHODS

Data from 277 previously-questioned participants is used in several data mining methods to select features (questions) that highly represent NSSI, then 245 different people were asked to participate in an online test to validate those features via machine learning methods.

RESULTS

The highest accuracy and F1 score of the selected features-via the Genetics algorithm-are 80.0% and 74.8% respectively for a Random Forest algorithm. Cronbach's alpha of the online test (validation on the selected features) is 0.82. Moreover, results suggest that an MLP can classify participants into two classes of NSSI Positive and NSSI Negative with 83.6% accuracy and 83.7% F1-score based on the answers to only 22 questions.

CONCLUSION

While previously psychologists used many combined questionnaires to see whether someone is involved in NSSI, via various data mining methods, the present study showed that only 22 questions are enough to predict if someone is involved or not. Then different machine learning algorithms were utilized to classify participants based on their NSSI behaviors, among which, an MLP with 10 hidden layers had the best performance.

Unravelling psychiatric heterogeneity and predicting suicide attempts in women with trauma-related dissociation using artificial intelligence

Background: Suicide is a leading cause of death, and rates of attempted suicide have increased during the COVID-19 pandemic. The under-diagnosed psychiatric phenotype of dissociation is associated with elevated suicidal self-injury; however, it has largely been left out of attempts to predict and prevent suicide.Objective: We designed an artificial intelligence approach to identify dissociative patients and predict prior suicide attempts in an unbiased, data-driven manner.Method: Participants were 30 controls and 93 treatment-seeking female patients with posttraumatic stress disorder (PTSD) and various levels of dissociation, including some with the PTSD dissociative subtype and some with dissociative identity disorder (DID).Results: Unsupervised learning models identified patients along a spectrum of dissociation. Moreover, supervised learning models accurately predicted prior suicide attempts with an F1 score up to 0.83. DID had the highest risk of prior suicide attempts, and distinct subtypes of dissociation predicted suicide attempts in PTSD and DID.Conclusions: These findings expand our understanding of the dissociative phenotype and underscore the urgent need to assess for dissociation to identify individuals at high-risk of suicidal self-injury.

Prediction Models for Suicide Attempts among Adolescents Using Machine Learning Techniques

OBJECTIVE

Suicide attempts (SAs) in adolescents are difficult to predict although it is a leading cause of death among adolescents. This study aimed to develop and evaluate SA prediction models based on six different machine learning (ML) algorithms for Korean adolescents using data from online surveys.

METHODS

Data were extracted from the 2011-2018 Korea Youth Risk Behavior Survey (KYRBS), an ongoing annual national survey. The participants comprised 468,482 nationally representative adolescents from 400 middle and 400 high schools, aged 12 to 18. The models were trained using several classic ML methods and then tested on internal and external independent datasets; performance metrics were calculated. Data analysis was performed from March 2020 to June 2020.

RESULTS

Among the 468,482 adolescents included in the analysis, 15,012 cases (3.2%) were identified as having made an SA. Three features (suicidal ideation, suicide planning, and grade) were identified as the most important predictors. The performance of the six ML models on the internal testing dataset was good, with both the area under the receiver operating characteristic curve (AUROC) and area under the precision-recall curve (AUPRC) ranging from 0.92 to 0.94. Although the AUROC of all models on the external testing dataset (2018 KYRBS) ranged from 0.93 to 0.95, the AUPRC of the models was approximately 0.5.

CONCLUSION

The developed and validated SA prediction models can be applied to detect high risks of SA. This approach could facilitate early intervention in the suicide crisis and may ultimately contribute to suicide prevention for adolescents.

Frequency and correlates of suicidal ideation and behaviors in treatment-seeking Post-9/11 Veterans

OBJECTIVE

Post-9/11 U.S. veterans and servicemembers are at increased risk for suicide, indicating an important need to identify and mitigate suicidal ideation and behaviors in this population.

METHOD

Using data modeling techniques, we examined correlates of suicidal ideation and behavior at intake in 261 Post-9/11 veterans and servicemembers seeking mental health treatment.

RESULTS

Our sample endorsed high rates of suicidal ideation and behavior. Approximately 40% of our sample scored in a range on the Suicide Behaviors Questionnaire-Revised (SBQ-R), indicating high clinical risk for suicide. Results from multivariate analyses indicate that greater state and/or trait depression severity, greater anger and anger expression, less impulse control, and lower rank were consistently associated with suicidal ideation and behavior across our models. Negative posttraumatic thoughts about the self, gender, and military branch of service were also significantly associated with suicidal ideation and behavior.

CONCLUSIONS

Suicidal ideation and behaviors are common in veterans seeking mental health treatment. State and/or trait depression, anger and impulse control were predictors of increased risk for suicidal ideation and behavior across models. Consistencies and differences across models as well as limitations and practical implications for the findings are discussed.

The performance of machine learning models in predicting suicidal ideation, attempts, and deaths: A meta-analysis and systematic review

Research has posited that machine learning could improve suicide risk prediction models, which have traditionally performed poorly. This systematic review and meta-analysis evaluated the performance of machine learning models in predicting longitudinal outcomes of suicide-related outcomes of ideation, attempt, and death and examines outcome, data, and model types as potential covariates of model performance. Studies were extracted from PubMed, Web of Science, Embase, and PsycINFO. A bivariate mixed effects meta-analysis and meta-regression analyses were performed for studies using machine learning to predict future events of suicidal ideation, attempts, and/or deaths. Risk of bias was assessed for each study using an adaptation of the Prediction model Risk Of Bias Assessment Tool. Narrative review included 56 studies, and analyses examined 54 models from 35 studies. The models achieved a very good pooled AUC of 0.86, sensitivity of 0.66 (95% CI [0.60, 0.72)], and specificity of 0.87 (95% CI [0.84, 0.90]). Pooled AUCs for ideation, attempt, and death were similar at 0.88, 0.87, and 0.84 respectively. Model performance was highly varied; however, meta-regressions did not provide evidence that performance varied by outcome, data, or model types. Findings suggest that machine learning has the potential to improve suicide risk detection, with pooled estimates of machine learning performance comparing favourably to performance of traditional suicide prediction models. However, more studies with lower risk of bias are necessary to improve the application of machine learning in suicidology.

Suicide theory-guided natural language processing of clinical progress notes to improve prediction of veteran suicide risk: protocol for a mixed-method study

INTRODUCTION

The state-of-the-art 3-step Theory of Suicide (3ST) describes why people consider suicide and who will act on their suicidal thoughts and attempt suicide. The central concepts of 3ST-psychological pain, hopelessness, connectedness, and capacity for suicide-are among the most important drivers of suicidal behaviour but they are missing from clinical suicide risk prediction models in use at the US Veterans Health Administration (VHA). These four concepts are not systematically recorded in structured fields of VHA's electronic healthcare records. Therefore, this study will develop a domain-specific ontology that will enable automated extraction of these concepts from clinical progress notes using natural language processing (NLP), and test whether NLP-based predictors for these concepts improve accuracy of existing VHA suicide risk prediction models.

METHODS AND ANALYSIS

Our mixed-method study has an exploratory sequential design where a qualitative component (aim 1) will inform quantitative analyses (aims 2 and 3). For aim 1, subject matter experts will manually annotate progress notes of clinical encounters with veterans who attempted or died by suicide to develop a domain-specific ontology for the 3ST concepts. During aim 2, we will use NLP to machine-annotate clinical progress notes and derive longitudinal representations for each patient with respect to the presence and intensity of hopelessness, psychological pain, connectedness and capacity for suicide in temporal proximity of suicide attempts and deaths by suicide. These longitudinal representations will be evaluated during aim 3 for their ability to improve existing VHA prediction models of suicide and suicide attempts, STORM (Stratification Tool for Opioid Risk Mitigation) and REACHVET (Recovery Engagement and Coordination for Health - Veterans Enhanced Treatment).

ETHICS AND DISSEMINATION

Ethics approval for this study was granted by the Stanford University Institutional Review Board and the Research and Development Committee of the VA Palo Alto Health Care System. Results of the study will be disseminated through several outlets, including peer-reviewed publications and presentations at national conferences.

Artificial intelligence and its impact on the domains of universal health coverage, health emergencies and health promotion: An overview of systematic reviews

•

An overview of systematic reviews on the application of AI including 129 studies.

•

AI use is prominent in Universal Health Coverage, featuring image analysis in neoplasms.

•

Half of the reviews did not evaluate validation procedures nor reporting guidelines.

•

Risk of bias was only included un a third of the reviews.

•

There is not sufficient evidence to transfer AI to actual healthcare delivery.

Background

Artificial intelligence is fueling a new revolution in medicine and in the healthcare sector. Despite the growing evidence on the benefits of artificial intelligence there are several aspects that limit the measure of its impact in people’s health. It is necessary to assess the current status on the application of AI towards the improvement of people’s health in the domains defined by WHO’s Thirteenth General Programme of Work (GPW13) and the European Programme of Work (EPW), to inform about trends, gaps, opportunities, and challenges.

Objective

To perform a systematic overview of systematic reviews on the application of artificial intelligence in the people’s health domains as defined in the GPW13 and provide a comprehensive and updated map on the application specialties of artificial intelligence in terms of methodologies, algorithms, data sources, outcomes, predictors, performance, and methodological quality.

Methods

A systematic search in MEDLINE, EMBASE, Cochrane and IEEEXplore was conducted between January 2015 and June 2021 to collect systematic reviews using a combination of keywords related to the domains of universal health coverage, health emergencies protection, and better health and wellbeing as defined by the WHO’s PGW13 and EPW. Eligibility criteria was based on methodological quality and the inclusion of practical implementation of artificial intelligence. Records were classified and labeled using ICD-11 categories into the domains of the GPW13. Descriptors related to the area of implementation, type of modeling, data entities, outcomes and implementation on care delivery were extracted using a structured form and methodological aspects of the included reviews studies was assessed using the AMSTAR checklist.

Results

The search strategy resulted in the screening of 815 systematic reviews from which 203 were assessed for eligibility and 129 were included in the review. The most predominant domain for artificial intelligence applications was Universal Health Coverage (N = 98) followed by Health Emergencies (N = 16) and Better Health and Wellbeing (N = 15). Neoplasms area on Universal Health Coverage was the disease area featuring most of the applications (21.7 %, N = 28). The reviews featured analytics primarily over both public and private data sources (67.44 %, N = 87). The most used type of data was medical imaging (31.8 %, N = 41) and predictors based on regions of interest and clinical data. The most prominent subdomain of Artificial Intelligence was Machine Learning (43.4 %, N = 56), in which Support Vector Machine method was predominant (20.9 %, N = 27). Regarding the purpose, the application of Artificial Intelligence I is focused on the prediction of the diseases (36.4 %, N = 47). With respect to the validation, more than a half of the reviews (54.3 %, N = 70) did not report a validation procedure and, whenever available, the main performance indicator was the accuracy (28.7 %, N = 37). According to the methodological quality assessment, a third of the reviews (34.9 %, N = 45) implemented methods for analysis the risk of bias and the overall AMSTAR score below was 5 (4.01 ± 1.93) on all the included systematic reviews.

Conclusion

Artificial intelligence is being used for disease modelling, diagnose, classification and prediction in the three domains of GPW13. However, the evidence is often limited to laboratory and the level of adoption is largely unbalanced between ICD-11 categoriesand diseases. Data availability is a determinant factor on the developmental stage of artificial intelligence applications. Most of the reviewed studies show a poor methodological quality and are at high risk of bias, which limits the reproducibility of the results and the reliability of translating these applications to real clinical scenarios. The analyzed papers show results only in laboratory and testing scenarios and not in clinical trials nor case studies, limiting the supporting evidence to transfer artificial intelligence to actual care delivery.

Consumer Perspectives on the Use of Artificial Intelligence Technology and Automation in Crisis Support Services: Mixed Methods Study

BACKGROUND

Emerging technologies, such as artificial intelligence (AI), have the potential to enhance service responsiveness and quality, improve reach to underserved groups, and help address the lack of workforce capacity in health and mental health care. However, little research has been conducted on the acceptability of AI, particularly in mental health and crisis support, and how this may inform the development of responsible and responsive innovation in the area.

OBJECTIVE

This study aims to explore the level of support for the use of technology and automation, such as AI, in Lifeline's crisis support services in Australia; the likelihood of service use if technology and automation were implemented; the impact of demographic characteristics on the level of support and likelihood of service use; and reasons for not using Lifeline's crisis support services if technology and automation were implemented in the future.

METHODS

A mixed methods study involving a computer-assisted telephone interview and a web-based survey was undertaken from 2019 to 2020 to explore expectations and anticipated outcomes of Lifeline's crisis support services in a nationally representative community sample (n=1300) and a Lifeline help-seeker sample (n=553). Participants were aged between 18 and 93 years. Quantitative descriptive analysis, binary logistic regression models, and qualitative thematic analysis were conducted to address the research objectives.

RESULTS

One-third of the community and help-seeker participants did not support the collection of information about service users through technology and automation (ie, via AI), and approximately half of the participants reported that they would be less likely to use the service if automation was introduced. Significant demographic differences were observed between the community and help-seeker samples. Of the demographics, only older age predicted being less likely to endorse technology and automation to tailor Lifeline's crisis support service and use such services (odds ratio 1.48-1.66, 99% CI 1.03-2.38; P<.001 to P=.005). The most common reason for reluctance, reported by both samples, was that respondents wanted to speak to a real person, assuming that human counselors would be replaced by automated robots or machine services.

CONCLUSIONS

Although Lifeline plans to always have a real person providing crisis support, help-seekers automatically fear this will not be the case if new technology and automation such as AI are introduced. Consequently, incorporating innovative use of technology to improve help-seeker outcomes in such services will require careful messaging and assurance that the human connection will continue.

Structured data vs. unstructured data in machine learning prediction models for suicidal behaviors: A systematic review and meta-analysis

Suicide remains a leading cause of preventable death worldwide, despite advances in research and decreases in mental health stigma through government health campaigns. Machine learning (ML), a type of artificial intelligence (AI), is the use of algorithms to simulate and imitate human cognition. Given the lack of improvement in clinician-based suicide prediction over time, advancements in technology have allowed for novel approaches to predicting suicide risk. This systematic review and meta-analysis aimed to synthesize current research regarding data sources in ML prediction of suicide risk, incorporating and comparing outcomes between structured data (human interpretable such as psychometric instruments) and unstructured data (only machine interpretable such as electronic health records). Online databases and gray literature were searched for studies relating to ML and suicide risk prediction. There were 31 eligible studies. The outcome for all studies combined was AUC = 0.860, structured data showed AUC = 0.873, and unstructured data was calculated at AUC = 0.866. There was substantial heterogeneity between the studies, the sources of which were unable to be defined. The studies showed good accuracy levels in the prediction of suicide risk behavior overall. Structured data and unstructured data also showed similar outcome accuracy according to meta-analysis, despite different volumes and types of input data.

Childhood adversities and suicidal thoughts and behaviors among first-year college students: results from the WMH-ICS initiative

PURPOSE

To investigate the associations of childhood adversities (CAs) with lifetime onset and transitions across suicidal thoughts and behaviors (STB) among incoming college students.

METHODS

Web-based self-report surveys administered to 20,842 incoming college students from nine countries (response rate 45.6%) assessed lifetime suicidal ideation, plans and attempts along with seven CAs: parental psychopathology, three types of abuse (emotional, physical, sexual), neglect, bully victimization, and dating violence. Logistic regression estimated individual- and population-level associations using CA operationalizations for type, number, severity, and frequency.

RESULTS

Associations of CAs with lifetime ideation and the transition from ideation to plan were best explained by the exact number of CA types (OR range 1.32-52.30 for exactly two to seven CAs). Associations of CAs with a transition to attempts were best explained by the frequency of specific CA types (scaled 0-4). Attempts among ideators with a plan were significantly associated with all seven CAs (OR range 1.16-1.59) and associations remained significant in adjusted analyses with the frequency of sexual abuse (OR = 1.42), dating violence (OR = 1.29), physical abuse (OR = 1.17) and bully victimization (OR = 1.17). Attempts among ideators without plan were significantly associated with frequency of emotional abuse (OR = 1.29) and bully victimization (OR = 1.36), in both unadjusted and adjusted analyses. Population attributable risk simulations found 63% of ideation and 30-47% of STB transitions associated with CAs.

CONCLUSION

Early-life adversities represent a potentially important driver in explaining lifetime STB among incoming college students. Comprehensive intervention strategies that prevent or reduce the negative effects of CAs may reduce subsequent onset of STB.

Validating a predictive algorithm for suicide risk with Alaska Native populations

INTRODUCTION

The American Indian/Alaska Native (AI/AN) suicide rate in Alaska is twice the state rate and four times the U.S. rate. Healthcare systems need innovative methods of suicide risk detection. The Mental Health Research Network (MHRN) developed suicide risk prediction algorithms in a general U.S.

PATIENT POPULATION

METHODS

We applied MHRN predictors and regression coefficients to electronic health records of AI/AN patients aged ≥13 years with behavioral health diagnoses and primary care visits between October 1, 2016, and March 30, 2018. Logistic regression assessed model accuracy for predicting and stratifying risk for suicide attempt within 90 days after a visit. We compared expected to observed risk and assessed model performance characteristics.

RESULTS

10,864 patients made 47,413 primary care visits. Suicide attempt occurred after 589 (1.2%) visits. Visits in the top 5% of predicted risk accounted for 40% of actual attempts. Among visits in the top 0.5% of predicted risk, 25.1% were followed by suicide attempt. The best fitting model had an AUC of 0.826 (95% CI: 0.809-0.843).

CONCLUSIONS

The MHRN model accurately predicted suicide attempts among AI/AN patients. Future work should develop clinical and operational guidance for effective implementation of the model with this population.

The First Year After Military Service: Predictors of U.S. Veterans' Suicidal Ideation

INTRODUCTION

Little is known about predictors of military veterans' suicidal ideation as they transition from service to civilian life, a potentially high-risk period that represents a critical time for intervention. This study examined factors associated with veterans' suicidal ideation in the first year after military separation.

METHODS

A national sample of U.S. veterans (N=7,383) from The Veterans Metrics Initiative Study reported on their mental health, psychosocial well-being, and demographic/military characteristics in an online survey at 3 and 9 months after separation. Cross-validated random forest models and mean decrease in accuracy values were used to identify key predictors of suicidal ideation. Bivariate ORs were calculated to examine the magnitude and direction of main effects associations between predictors and suicidal ideation. Data were collected in 2016/2017 and analyzed in 2021.

RESULTS

In the first year after separation, 15.1% of veterans reported suicidal ideation. Endorsing depression symptoms and, to a lesser extent, identifying oneself as experiencing depression, were most predictive of suicidal ideation. Other psychopathology predictors included higher anxiety and posttraumatic stress disorder symptoms. Psychosocial well-being predictors included higher health satisfaction and functioning, community satisfaction and functioning, and psychological resilience. Logistic models performed similarly to random forest models, suggesting that relationships between predictors and suicidal ideation were better represented as main effects than interactions.

CONCLUSIONS

Results highlight the potential value of bolstering key aspects of military veterans' mental health and psychosocial well-being to reduce their risk for suicidal ideation in the first year after separation. Findings can inform interventions aimed at helping veterans acclimate to civilian life.

Pain avoidance and functional connectivity between insula and amygdala identifies suicidal attempters in patients with major depressive disorder using machine learning

Pain avoidance can effectively classify suicide attempters from non-attempters among patients with major depressive disorder (MDD). However, the neural circuits underlying pain processing in suicide attempters have not been described comprehensively. In Study 1, we recruited MDD patients with a history of suicide attempts (MDD-SA), and those without (MDD-NSA) to examine the patterns of psychological pain using the latent profile analysis. Further, in Study 2, participants including the MDD-SA, MDD-NSA, and healthy controls underwent resting-state functional magnetic resonance imaging. We used machine learning that included features of gray matter volume (GMV), the functional connectivity (FC) brain patterns of the region of interest, and behavioral data to identify suicide attempters. The results identified three latent classes of psychological pain in MDD patients: the low pain class (18.9%), the painful feeling class (37.2%), and the pain avoidance class (43.9%). Furthermore, the proportion of suicide attempters with high pain avoidance was the highest. The accuracy of multimodality classifiers (63%-92%) was significantly higher than that of brain-only classifiers (56%-85%) and behavior-only classifiers (64%-73%). Pain avoidance ranked first in the optimal feature set of the suicide attempt classification model. The crucial brain imaging features were FC between the left amygdala and right insula, right orbitofrontal and left thalamus, left anterior cingulate cortex and left insula, right orbitofrontal, amygdala, and the GMV of right thalamus. Additionally, the optimal feature set, including pain avoidance and crucial brain patterns of psychological pain neural circuits, was provided for the identification of suicide attempters.

Newtonian science, complexity science and suicide-critically analysing the philosophical basis for suicide research: A discussion paper

AIM

A critical discussion comparing Newtonian science and complexity science as the philosophical basis for suicide research and its impact on suicide knowledge development and clinical practice.

DESIGN

Discussion paper.

DATA SOURCES

A review of literature on suicide research and complexity science ranging from 2000 to 2022.

IMPLICATIONS FOR NURSING

Suicide research based on a Newtonian worldview can have negative consequences for suicide knowledge development and can permeate nursing practice in ways that take away from addressing the complex needs of patients, their families and healthcare teams.

CONCLUSION

A Newtonian worldview as a philosophical basis for research is insufficient for the study of a phenomenon as complex as suicide. A complexity science approach is better suited to the study of suicide given the multiple, interrelated, emerging factors that can contribute to a person's decision to end their own life.

IMPACT

Suggestions are provided as to how a complexity science approach to the research of suicide can inform useful knowledge development that better meets the needs of individuals facing suicidality and their families. Researchers, healthcare administrators and nurses providing care to those struggling with suicidality can benefit from adopting a complexity science worldview in addressing this multifaceted phenomenon.

Predicting Children with ADHD Using Behavioral Activity: A Machine Learning Analysis

Attention deficit hyperactivity disorder (ADHD) is one of childhood’s most frequent neurobehavioral disorders. The purpose of this study is to: (i) extract the most prominent risk factors for children with ADHD; and (ii) propose a machine learning (ML)-based approach to classify children as either having ADHD or healthy. We extracted the data of 45,779 children aged 3–17 years from the 2018–2019 National Survey of Children’s Health (NSCH, 2018–2019). About 5218 (11.4%) of children were ADHD, and the rest of the children were healthy. Since the class label is highly imbalanced, we adopted a combination of oversampling and undersampling approaches to make a balanced class label. We adopted logistic regression (LR) to extract the significant factors for children with ADHD based on p-values (<0.05). Eight ML-based classifiers such as random forest (RF), Naïve Bayes (NB), decision tree (DT), XGBoost, k-nearest neighborhood (KNN), multilayer perceptron (MLP), support vector machine (SVM), and 1-dimensional convolution neural network (1D CNN) were adopted for the prediction of children with ADHD. The average age of the children with ADHD was 12.4 ± 3.4 years. Our findings showed that RF-based classifier provided the highest classification accuracy of 85.5%, sensitivity of 84.4%, specificity of 86.4%, and an AUC of 0.94. This study illustrated that LR with RF-based system could provide excellent accuracy for classifying and predicting children with ADHD. This system will be helpful for early detection and diagnosis of ADHD.

Evaluating the heterogeneous effect of a modifiable risk factor on suicide: The case of vitamin D deficiency

OBJECTIVES

To illustrate the use of machine learning methods to search for heterogeneous effects of a target modifiable risk factor on suicide in observational studies. The illustration focuses on secondary analysis of a matched case-control study of vitamin D deficiency predicting subsequent suicide.

METHODS

We describe a variety of machine learning methods to search for prescriptive predictors; that is, predictors of significant variation in the association between a target risk factor and subsequent suicide. In each case, the purpose is to evaluate the potential value of selective intervention on the target risk factor to prevent the outcome based on the provisional assumption that the target risk factor is causal. The approaches illustrated include risk modeling based on the super learner ensemble machine learning method, Least Absolute Shrinkage and Selection Operator (Lasso) penalized regression, and the causal forest algorithm.

RESULTS

The logic of estimating heterogeneous intervention effects is exposited along with the illustration of some widely used methods for implementing this logic.

CONCLUSIONS

In addition to describing best practices in using the machine learning methods considered here, we close with a discussion of broader design and analysis issues in planning an observational study to investigate heterogeneous effects of a modifiable risk factor.

Predictors of suicide attempt in patients with obsessive-compulsive disorder: an exploratory study with machine learning analysis

BACKGROUND

Patients with obsessive-compulsive disorder (OCD) are at increased risk for suicide attempt (SA) compared to the general population. However, the significant risk factors for SA in this population remains unclear - whether these factors are associated with the disorder itself or related to extrinsic factors, such as comorbidities and sociodemographic variables. This study aimed to identify predictors of SA in OCD patients using a machine learning algorithm.

METHODS

A total of 959 outpatients with OCD were included. An elastic net model was performed to recognize the predictors of SA among OCD patients, using clinical and sociodemographic variables.

RESULTS

The prevalence of SA in our sample was 10.8%. Relevant predictors of SA founded by the elastic net algorithm were the following: previous suicide planning, previous suicide thoughts, lifetime depressive episode, and intermittent explosive disorder. Our elastic net model had a good performance and found an area under the curve of 0.95.

CONCLUSIONS

This is the first study to evaluate risk factors for SA among OCD patients using machine learning algorithms. Our results demonstrate an accurate risk algorithm can be created using clinical and sociodemographic variables. All aspects of suicidal phenomena need to be carefully investigated by clinicians in every evaluation of OCD patients. Particular attention should be given to comorbidity with depressive symptoms.