Abstract

The United States is currently grappling with a severe and urgent overdose crisis, driven by synthetic fentanyl and the increasing involvement of multiple substances. This evolving pattern has made it increasingly challenging to detect, predict, and respond to overdose events in a timely manner. While public health surveillance systems play a crucial role in providing information about emerging trends, traditional data sources often struggle to keep pace with the rapid changes in the drug supply. However, the advent of machine learning presents a ray of hope, offering opportunities to identify individuals and communities at elevated risk and to forecast overdose patterns before they manifest in official statistics.
This review synthesizes studies that used public health surveillance, machine learning, or both to predict opioid or polysubstance overdose in the United States. A structured search of major academic databases identified research that employed modeling approaches, such as random forests, gradient boosting methods, neural networks, and spatiotemporal surveillance strategies, to estimate overdose risk. Studies were included if they were conducted in the United States, used a predictive or surveillance-based method, and reported fatal or nonfatal overdose outcomes.
The findings demonstrate growing interest in models that integrate multiple data sources, including electronic health records, Medicaid claims, emergency medical services data, criminal justice information, and statewide surveillance systems. Across studies, predictors that consistently contributed to overdose risk included prior overdose events, mental health conditions, patterns of substance use, and community-level factors. While model performance varied, many studies achieved intense discrimination.
This review highlights opportunities to strengthen prediction systems by combining advanced modelling strategies with more integrated and timely surveillance frameworks. Such approaches support earlier detection and more targeted public health responses to the ongoing overdose crisis in the United States.