018 - Identifying psychosocial and ecological determinants of enthusiasm in youth using machine learning

Introduction
Understanding the factors contributing to mental wellbeing in youth is a public health priority. Self-reported enthusiasm for the future may be a useful indicator of wellbeing and has been shown to forecast social and educational success. Typically, cross-domain measures of ecological and health-related factors with relevance to public policy are analyzed either in isolation or in targeted models assessing bivariate interactions. Here, we capitalize on a large provincial dataset and cutting-edge interpretable machine learning to analyze dozens of measures simultaneously in a non-linear, interactive framework.

Objective
The aim of this study is to identify the sociodemographic, experiential, behavioral, and other health-related factors most strongly associated with levels of subjective enthusiasm for the future in a large sample of elementary and secondary school students using machine learning.

Methods
We analyzed data from 14,142 participants in the 2019 Ontario Student Drug Use and Health Survey (OSDUHS) (grades 7 to 12 inclusive, age range: 11 to 20 years old). Our primary outcome of interest was self-reported enthusiasm, measured by level of agreement with the statement "I am very enthusiastic about my future" (4-point Likert Scale). We used 50 variables derived from the OSDUHS survey to model this outcome, including demographics, perception of school experience (including aspects of school connectedness and academic performance), physical activity and sleep patterns, substance use, and physical and mental health indicators, among others. Models were built using a non-linear decision-tree based machine learning algorithm, called eXtreme Gradient Boosting (XGBoost), to classify students as indicating either high or low levels of enthusiasm. Cross-validated hyperparameter optimization and model training were performed on 80% of the sample, and unbiased model performance was evaluated on a 20% withheld test set. SHapley Additive exPlanations (SHAP) values were used to interpret the model, providing a ranking of feature importance and identifying interactions between model features.

Results
An 81% classification accuracy was achieved with our top performing model; precision and recall were 82% and 81%, respectively. The top three contributors to higher self-rated enthusiasm for the future from this model include: higher self-rated physical health (SHAP=0.62), feeling able to discuss problems or feelings with their parents (0.49), and having a sense of belonging in their school (0.32). In addition, subjective social status at school was a top feature and showed non-linear effects, with benefits to predicted enthusiasm only present in the mid-high range of values (from 0-10). These physical health and school participation variables also interacted within the model, where improvements in one area may mitigate deficits in the other.

Conclusion
Using machine leaning, we identified key factors related to self-reported enthusiasm for the future in a large sample of young students: perceived physical health, school social status and connectedness, and quality of family relationships. These factors were more important in determining enthusiasm than many common policy targets including social media, drug, and alcohol use. Together, our findings suggest that a focus on physical health and school connectedness should be central to impactful policy aimed at improving the wellbeing of youth, particularly when it comes to improving enthusiasm for the future.

Youth Enthusiasm

Machine Learning

School Survey

Wellbeing

XGBoost

SHAP 

Mid-Level

Knowledge

International Conference on Health Policy Statistics 2023