Belonging, discrimination, and risk in displacement: Bayesian variable selection in predicting mental health and resilience among refugee adolescents in Greece

Background: Understanding the mental health challenges of refugee adolescents is critical for informing targeted educational policies and interventions. Yet few studies identify which school-based factors most strongly relate to their resilience and well-being. This study draws on a multi-informant dataset from 170 refugee students attending school in Greece, applying Bayesian variable selection to identify key predictors of mental health outcomes. Methods: We collected student ( n = 170 ), teacher ( n = 165 ) and parent/guardian ( n = 124 ) data about each student (Strengths and Difficulties Questionnaire, SDQ; Child and Youth Resilience Measure-Revised, CYRM-R); sense of belonging, school attendance. We ran two models, one based on student self-reports, the other based on teacher, parent, and guardian data. Results: The self-reported sense of belonging was the most consistent protective factor of SDQ and CYRM-R outcomes, while bullying, discrimination, and disrupted schooling were associated with greater vulnerability. In triangulated analysis, student self-reports contributed the largest share of predictive information, followed by adult reports Discussion: This study highlights the importance of centering youth perspectives in mental health assessment and intervention and the critical role of inclusive school environments and educational continuity in supporting refugee integration. The Bayesian approach offers a robust framework for guiding evidence-based interventions for supporting the integration of refugee students in host-country educational settings. • We analyzed multi-informant data on refugee adolescents’ mental health difficulties and resilience. • Sense of belonging emerged as the strongest predictor of school-based mental health and resilience. • Bullying, discrimination, and disrupted schooling were associated with increased vulnerability. • Students’ self-reports showed higher predictive validity, while teacher reports aligned more closely with outcomes than caregiver reports. • Findings underscore the importance of inclusive, student-centered school environments.