(PS3-34) Examining Environmental Contributions to Adolescent Depression Risk: A GLMM Tree approach to Identifying Risk

Abstract Text: Adolescent mental health is a critical public health concern, with bullying, sleep health and suicidal ideation all being contributors to emotional distress. Previous research indicated that adolescents involved in bullying were more likely to experience sleep difficulties (Hunter et al., 2014), and further research highlighted the differences in bullying, either offline or online, between genders at different stages of adolescence (Smith et al., 2019). A meta-analytic study conducted by Bartel et al., (2015) identified several potential risk factors for adolescent depression, however, there was uncertainty across the literature studied regarding the severity of the impact of these risk factors leading to the development of adolescent depression. Given the complexity of these and other factors, identifying flexible yet comprehensive models is therefore essential for capturing nuanced relationships in treatment and guiding further research analyses. Using data from the Youth Risk Behavior Surveillance System (N = 17,441), this study uses Generalized Linear Mixed-Effects Model Trees (GLMM Trees) to examine the significant predictors of adolescent mental illness such as poor sleep health and suicidal ideation in predicting overall sadness in high school age adolescents. The GLMM tree model identified gender, presence of bullying, sleep patterns, and suicidal ideation as significant predictors of depression risk (experiences of sadness). Adolescents with the largest proportion of depression risk were identified by a history of online bullying and identifying as female. Male identifying adolescents were significantly associated with physical bullying risk. Across all subgroups, suicidal ideation was associated with increased odds of sadness. Healthy sleepers, those with over 8 hours of sleep per night, reported lower sadness levels, particularly in students without suicidal ideation. Adolescents with unhealthy sleep patterns, those with less than 7 hours of sleep per night, reported higher levels of sadness among those with suicidal thoughts. The GLMM tree model used in this study provides a flexible decision-tree method for examining multilevel and complex data from large study samples (Fokkema et al., 2021) and can be utilized to predict treatment outcomes. Future research should investigate the use of this model to inform the development of population specific treatments for vulnerable groups.