Risk Modeling and Association Pathways Integrating Psychosocial Factors for Drug Misuse in Adolescents and Young Adults: A Machine Learning Approach.

INTRODUCTION: Early identification for preventing drug misuse among adolescents and young adults (AYAs) is more cost-effective than drug treatment. However, there is a lack of scientific and comprehensive risk models for early identification. This study aimed to construct risk models and association pathways that integrate psychosocial factors influencing drug misuse in AYAs using a machine learning approach. METHODS: This cross-sectional study included 1,012 Chinese AYAs aged 14-35 years and was conducted from June to December 2023. Key psychosocial factors for drug misuse were identified using the least absolute shrinkage and selection operator and Boruta. A voting classifier combining logistic regression, K-Nearest Neighbors, support vector machine, and 5-fold cross-validation was used for model training and bootstrapping method wan applied for testing. Shapley additive explanation was used to determine the importance of each risk factor. A Bayesian network (BN) model was constructed to explore possible association pathways for drug misuse. RESULTS: A total of 3.85% of participants reported drug misuse. Significant risk factors included environmental exposure to drug misusers, working in nightlife venues, cognition of traditional drug types, perceived risks of drug misuse, permissive attitude towards drug misuse, refusal self-efficacy of drug misuse, externalizing problem behavior, sensation-seeking, and adverse childhood experiences (ACEs). The voting model showed strong performance (recall: 94.10%; area under the curve: 98.11%; accuracy: 92.63%; precision: 34.15%; F-score: 49.68%). The BN model revealed that ACEs played a central role in risk assessment of drug misuse. AYAs working in nightlife venues, with low refusal self-efficacy of drug misuse, and exposure to drug misusers, had exceeding 90% probability of drug misuse. CONCLUSION: ACEs serve as a primary risk factor, with a significant impact on social environmental influences. AYAs with ACEs and those working in nightlife venues require targeted and early interventions to prevent drug misuse.