Jacob Keller

Evidence-based practices (EBPs) can support the delivery of high-quality, evidence-informed care for children and adolescents, yet are rarely implemented in German Child and Youth Welfare Service (CYWS) facilities. Although attitudes toward EBPs have been associated with EBP use, implementation is also shaped by organizational and individual factors. However, little is known about these relationships among child welfare workers (CWWs) in German CYWS facilities. This study therefore examined attitudes toward EBPs and their associations with perceived implementation climate and professional quality of life among CWWs. Baseline data of 161 child welfare workers in 37 German CYWS facilities from multi-center cluster-randomized controlled trial BETTER CARE were analyzed. Associations between the implementation climate (Implementation Climate Scale (ICS)) and the Professional Quality of Life (ProQOL) on attitude toward EBPs (Evidence-Based Practice Attitude Scale (EBPAS)) were examined using LASSO-based predictor selection followed by multiple linear regression analyses with cluster-robust standard errors. The ICS subscales recognition ( β = 0.22, p = .002), rewards ( β = -0.21, p = .004), and selection for openness ( β = 0.26, p = .002), as well as the ProQOL subscales burnout ( β = -0.60, p < .001) and compassion satisfaction ( β = 0.32, p = .022) were significant predictors of the attitudes toward EBPs among CWWs. Among CWWs in German CYWS facilities, attitudes toward EBPs were associated with selected dimensions of implementation climate and professional quality of life. The findings suggest that attitudes toward EBPs should be considered within their organizational and individual context.

Background: Suicidality is a major public health concern worldwide. Evidence on the prevalence and risk factors of suicidality amongst unaccompanied young refugees (UYRs), a population already at risk for mental health disorders, is scarce. Methods: Given the complexity of individual risk factor constellations influencing suicidality, machine learning (ML) methods offer a statistical approach that can detect complex relations within the data. Four ML classifiers, (logistic regression (LR), random forest (RF), support vector machines (SVM), and extreme gradient boosting (XGB)) were trained on a dataset of n = 623 UYRs (M _age =16.77, SD = 1.34, range: 12–21), retrieved from the large-scale randomized controlled trial Better Care to predict suicidal ideation. Features used in the classifiers were age, gender, asylum status, having contact with the family, and whether parents are alive as well as clinically elevated post-traumatic stress symptoms (PTSS), depressive symptoms and past suicide attempts. The classifiers were then tested on the independent dataset of n = 94 UYRs (M _age =16.31, SD = 2.03, range: 5–21) retrieved from the screening tool porta project to examine their predictive performance. Results: The prevalence of past-week suicidal ideation in the combined sample of N = 717 was 18.13%. All classifiers yielded good predictive performance (accuracy 0.734–0.840, sensitivity 0.857, AUC 0.853–0.880). The most relevant features were past suicide attempts, PTSS and depressive symptoms as risk factors, and having a living mother as protective factor. Conclusions: Suicidal ideation is prevalent amongst UYRs, and using ML approaches, the classifiers were able to classify roughly 85% of the cases with suicidal ideation in the past week correctly as suicidal. Building on the findings of this study, screening for suicidality could be further improved by implementing ML classifiers in the assessment to highlight potential at risk cases early, and suitable interventions be developed.

Background: Suicide represents a critical public health concern, and machine learning (ML) models offer the potential for identifying at-risk individuals. Recent studies using benchmark datasets and real-world social media data have demonstrated the capability of pretrained large language models in predicting suicidal ideation and behaviors (SIB) in speech and text. Objective: This study aimed to (1) develop and implement ML methods for predicting SIBs in a real-world crisis helpline dataset, using transformer-based pretrained models as a foundation; (2) evaluate, cross-validate, and benchmark the model against traditional text classification approaches; and (3) train an explainable model to highlight relevant risk-associated features. Methods: We analyzed chat protocols from adolescents and young adults (aged 14-25 years) seeking assistance from a German crisis helpline. An ML model was developed using a transformer-based language model architecture with pretrained weights and long short-term memory layers. The model predicted suicidal ideation (SI) and advanced suicidal engagement (ASE), as indicated by composite Columbia-Suicide Severity Rating Scale scores. We compared model performance against a classical word-vector-based ML model. We subsequently computed discrimination, calibration, clinical utility, and explainability information using a Shapley Additive Explanations value-based post hoc estimation model. Results: The dataset comprised 1348 help-seeking encounters (1011 for training and 337 for testing). The transformer-based classifier achieved a macroaveraged area under the curve (AUC) receiver operating characteristic (ROC) of 0.89 (95% CI 0.81-0.91) and an overall accuracy of 0.79 (95% CI 0.73-0.99). This performance surpassed the word-vector-based baseline model (AUC-ROC=0.77, 95% CI 0.64-0.90; accuracy=0.61, 95% CI 0.61-0.80). The transformer model demonstrated excellent prediction for nonsuicidal sessions (AUC-ROC=0.96, 95% CI 0.96-0.99) and good prediction for SI and ASE, with AUC-ROCs of 0.85 (95% CI 0.97-0.86) and 0.87 (95% CI 0.81-0.88), respectively. The Brier Skill Score indicated a 44% improvement in classification performance over the baseline model. The Shapley Additive Explanations model identified language features predictive of SIBs, including self-reference, negation, expressions of low self-esteem, and absolutist language. Conclusions: Neural networks using large language model–based transfer learning can accurately identify SI and ASE. The post hoc explainer model revealed language features associated with SI and ASE. Such models may potentially support clinical decision-making in suicide prevention services. Future research should explore multimodal input features and temporal aspects of suicide risk.

BACKGROUND Suicide represents a critical public health concern, and machine learning (ML) models offer the potential for identifying at-risk individuals. Recent studies using benchmark datasets and real-world social media data have demonstrated the capability of pretrained large language models in predicting suicidal ideation and behaviors (SIB) in speech and text. OBJECTIVE This study aimed to (1) develop and implement ML methods for predicting SIBs in a real-world crisis helpline dataset, using transformer-based pretrained models as a foundation; (2) evaluate, cross-validate, and benchmark the model against traditional text classification approaches; and (3) train an explainable model to highlight relevant risk-associated features. METHODS We analyzed chat protocols from adolescents and young adults (aged 14-25 years) seeking assistance from a German crisis helpline. An ML model was developed using a transformer-based language model architecture with pretrained weights and long short-term memory layers. The model predicted suicidal ideation (SI) and advanced suicidal engagement (ASE), as indicated by composite Columbia-Suicide Severity Rating Scale scores. We compared model performance against a classical word-vector-based ML model. We subsequently computed discrimination, calibration, clinical utility, and explainability information using a Shapley Additive Explanations value-based post hoc estimation model. RESULTS The dataset comprised 1348 help-seeking encounters (1011 for training and 337 for testing). The transformer-based classifier achieved a macroaveraged area under the curve (AUC) receiver operating characteristic (ROC) of 0.89 (95% CI 0.81-0.91) and an overall accuracy of 0.79 (95% CI 0.73-0.99). This performance surpassed the word-vector-based baseline model (AUC-ROC=0.77, 95% CI 0.64-0.90; accuracy=0.61, 95% CI 0.61-0.80). The transformer model demonstrated excellent prediction for nonsuicidal sessions (AUC-ROC=0.96, 95% CI 0.96-0.99) and good prediction for SI and ASE, with AUC-ROCs of 0.85 (95% CI 0.97-0.86) and 0.87 (95% CI 0.81-0.88), respectively. The Brier Skill Score indicated a 44% improvement in classification performance over the baseline model. The Shapley Additive Explanations model identified language features predictive of SIBs, including self-reference, negation, expressions of low self-esteem, and absolutist language. CONCLUSIONS Neural networks using large language model–based transfer learning can accurately identify SI and ASE. The post hoc explainer model revealed language features associated with SI and ASE. Such models may potentially support clinical decision-making in suicide prevention services. Future research should explore multimodal input features and temporal aspects of suicide risk.