Identifying High dose Opioid Prescription Risks using Machine Learning: A Focus on Socio-Demographic Characteristics

This study addresses the critical public health issue of high-dose opioid prescriptions in the United States, exploring the influence of socio-demographic factors on opioid prescribing patterns using machine learning techniques. Utilizing a comprehensive dataset that integrates opioid prescription claims data from Missouri (2017-2021) with socioeconomic context from the American Community Survey and broadband data, the research focuses on identifying patients at risk of receiving high-dose opioid prescriptions (defined as ≥120 MME/day). The study emphasizes the interpretability of machine learning models, particularly employing Shapley Additive exPlanations (SHAP) for transparent and understandable findings. The analysis reveals significant sociodemographic disparities in opioid dispensing, highlighting the need for nuanced healthcare policies and interventions to address these inequalities effectively. This research contributes to the growing body of knowledge on opioid prescription risks and demonstrates the potential of machine learning in providing deeper insights into public health issues.