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.

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.

Are false positives in suicide classification models a risk group? Evidence for "true alarms" in a population-representative longitudinal study of Norwegian adolescents.

Introduction: False positives in retrospective binary suicide attempt classification models are commonly attributed to sheer classification error. However, when machine learning suicide attempt classification models are trained with a multitude of psycho-socio-environmental factors and achieve high accuracy in suicide risk assessment, false positives may turn out to be at high risk of developing suicidal behavior or attempting suicide in the future. Thus, they may be better viewed as "true alarms," relevant for a suicide prevention program. In this study, using large population-based longitudinal dataset, we examine three hypotheses: (1) false positives, compared to the true negatives, are at higher risk of suicide attempt in future, (2) the suicide attempts risk for the false positives increase as a function of increase in specificity threshold; and (3) as specificity increases, the severity of risk factors between false positives and true positives becomes more similar. Methods: Utilizing the Gradient Boosting algorithm, we used a sample of 11,369 Norwegian adolescents, assessed at two timepoints (1992 and 1994), to classify suicide attempters at the first time point. We then assessed the relative risk of suicide attempt at the second time point for false positives in comparison to true negatives, and in relation to the level of specificity. Results: We found that false positives were at significantly higher risk of attempting suicide compared to true negatives. When selecting a higher classification risk threshold by gradually increasing the specificity cutoff from 60% to 97.5%, the relative suicide attempt risk of the false positive group increased, ranging from minimum of 2.96 to 7.22 times. As the risk threshold increased, the severity of various mental health indicators became significantly more comparable between false positives and true positives. Conclusion: We argue that the performance evaluation of machine learning suicide classification models should take the clinical relevance into account, rather than focusing solely on classification error metrics. As shown here, the so-called false positives represent a truly at-risk group that should be included in suicide prevention programs. Hence, these findings should be taken into consideration when interpreting machine learning suicide classification models as well as planning future suicide prevention interventions for adolescents.

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.

Mental Health, Well-Being, and Adolescent Extremism: A Machine Learning Study on Risk and Protective Factors.

We examined the relationship between adolescents' extremist attitudes with a multitude of mental health, well-being, psycho-social, environmental, and lifestyle variables, using state-of-the-art machine learning procedure and nationally representative survey dataset of Norwegian adolescents (N = 11,397). Three key research questions were addressed: 1) can adolescents with extremist attitudes be distinguished from those without, using psycho-socio-environmental survey items, 2) what are the most important predictors of adolescents' extremist attitudes, and 3) whether the identified predictors correspond to specific latent factorial structures? Of the total sample, 17.6% showed elevated levels of extremist attitudes. The prevalence was significantly higher among boys and younger adolescents than girls and older adolescents, respectively. The machine learning model reached an AUC of 76.7%, with an equal sensitivity and specificity of 70.5% in the test dataset, demonstrating a satisfactory performance for the model. Items reflecting on positive parenting, quality of relationships with parents and peers, externalizing behavior, and well-being emerged as significant predictors of extremism. Exploratory factor analysis partially supported the suggested latent clusters. Out of the 550 psycho-socio-environmental variables analyzed, behavioral problems, individual and social well-being, along with basic needs such as a secure family environment and interpersonal relationships with parents and peers emerged as significant factors contributing to susceptibility to extremism among adolescents.

Beat alignment ability is associated with formal musical training not current music playing.

The ability to perceive the beat in music is crucial for both music listeners and players with expert musicians being notably skilled at noticing fine deviations in the beat. However, it is unclear whether this beat perception ability is enhanced in trained musicians who continue to practice relative to musicians who no longer play. Thus, we investigated this by comparing active musicians', inactive musicians', and nonmusicians' beat alignment ability scores on the Computerized Adaptive Beat Alignment Test (CA-BAT). 97 adults with diverse musical experience participated in the study, reporting their years of formal musical training, number of instruments played, hours of weekly music playing, and hours of weekly music listening, in addition to their demographic information. While initial tests between groups indicated active musicians outperformed inactive musicians and nonmusicians on the CA-BAT, a generalized linear regression analysis showed that there was no significant difference once differences in musical training had been accounted for. To ensure that our results were not impacted by multicollinearity between music-related variables, nonparametric and nonlinear machine learning regressions were employed and confirmed that years of formal musical training was the only significant predictor of beat alignment ability. These results suggest that expertly perceiving fine differences in the beat is not a use-dependent ability that degrades without regular maintenance through practice or musical engagement. Instead, better beat alignment appears to be associated with more musical training regardless of continued use.

Adolescents' psychosocial well-being one year after the outbreak of the COVID-19 pandemic in Norway.

The COVID-19 pandemic has dramatically restricted adolescents' lives. We used nationwide Norwegian survey data from 2014-2021 (N = 227,258; ages 13-18) to examine psychosocial outcomes in adolescents before and during the pandemic. Multilevel models revealed higher depressive symptoms and less optimistic future life expectations during the pandemic, even when accounting for the measures' time trends. Moreover, alcohol and cannabis use decreased, and screen time increased. However, the effect sizes of all observed changes during the pandemic were small. Overall, conduct problems and satisfaction with social relationships remained stable. Girls, younger adolescents and adolescents from low socio-economic backgrounds showed more adverse changes during the pandemic. Estimated changes in psychosocial outcomes varied little with municipality infection rates and restrictions. These findings can inform means and interventions to reduce negative psychological outcomes associated with the pandemic and identify groups that need particular attention during and after the pandemic.

Estimating measurement error in child language assessments administered by daycare educators in large scale intervention studies.

Measurement error is a ubiquitous element of social science studies. In large-scale effectiveness intervention studies on child language, administration of the assessment of language and preliteracy outcomes by speech and language pathologists is costly in money and human resources. Alternatively, daycare educators can administer the assessment, which preserves considerable resources but may increase the measurement error. Using data from two nationwide child language intervention studies in Denmark, this article evaluates daycare educators' measurement error when administering a test of language and preliteracy skills of 3 to 5 year old children that in part is used in a national screening program. Since children were randomly assigned to educators, hierarchical linear models can estimate the amount of additional measurement error caused by educators' language assessments. The result shows that the amount of additional measurement error varied between different language subscales, ranging from 4% to 19%, which can be compensated for by increasing the sample size by the latter percentage. The benefits and risks of having daycare educators administer language assessments are discussed.

Measurement invariance of assessments of depression (PHQ-9) and anxiety (GAD-7) across sex, strata and linguistic backgrounds in a European-wide sample of patients after Traumatic Brain Injury.

BACKGROUND: The Patient Health Questionnaire-9 (PHQ-9) and the Generalized Anxiety Disorder (GAD-7) are two widely used instruments to screen patients for depression and anxiety. Comparable psychometric properties across different demographic and linguistic groups are necessary for multiple group comparison and international research on depression and anxiety. OBJECTIVES AND METHOD: We examine measurement invariance for the PHQ-9 and GAD-7 by: (a) the sex of the participants, (b) recruitment stratum, and (c) linguistic background. This study is based on non-randomized observational data six months after Traumatic Brain Injury (TBI) that were collected in 18 countries. We used multiple methods to detect Differential Item Functioning (DIF) including Item Response Theory, logistic regression, and the Mantel-Haenszel method. RESULTS: At the 6-month post-injury, 2137 (738 [34.5%] women) participants completed the PHQ-9 and GAD-7 questionnaires: 885 [41.4%] patients were primarily admitted to the Intensive Care Unit (ICU), 805 [37.7%] were admitted to hospital ward, and 447 [20.9%] were evaluated in the Emergency Room and discharged. Results supported the invariance of PHQ-9 and GAD-7 across sex, patient strata and linguistic background. For different strata three PHQ-9 items and one GAD-7 item and for different linguistic groups only two GAD-7 items were flagged as showing differences in two out of four DIF tests. However, the magnitude of the DIF effect was negligible. LIMITATION: Despite high number of participants from ICU, patients have mostly mild TBI. CONCLUSION: The findings demonstrate adequate psychometric properties for PHQ-9 and GAD-7, allowing direct multigroup comparison across sex, strata, and linguistic background.

The Effectiveness of a Large-Scale Language and Preliteracy Intervention: The SPELL Randomized Controlled Trial in Denmark.

The present article reports results of a real-world effectiveness trial conducted in Denmark with six thousand four hundred eighty-three 3- to 6-year-olds designed to improve children's language and preliteracy skills. Children in 144 child cares were assigned to a control condition or one of three planned variations of a 20-week storybook-based intervention: a base intervention and two enhanced versions featuring extended professional development for educators or a home-based program for parents. Pre- to posttest comparisons revealed a significant impact of all three interventions for preliteracy skills (= .21-.27) but not language skills (d = .04-.16), with little differentiation among the three variations. Fidelity, indexed by number of lessons delivered, was a significant predictor of most outcomes. Implications for real-world research and practice are considered.