Predicting fatal opioid-involved overdoses: A social-ecological framework matched to a linked-data warehouse

Abstract

Background

An estimated 60 million people used opioids non-medically worldwide in 2021. In 2019, opioid use disorder caused the loss of over 12.5 million healthy years of life due to disability and premature deaths, including those resulting from opioid-involved overdoses. Factors associated with opioid-involved overdoses are numerous, multi-layered, and interrelated. Using the social-ecological model as a foundation, we sought to comprehensively identify risk and preventive factors of fatal opioid-involved overdoses and operationalize them with quantifiable measures.

Methods

With our Community Advisory Board, investigators’ expertise, and an examination of the literature, we created an expansive, opioid-overdose specific social-ecological model structured as a matrix, with demographic, behavioral, environmental, and service domains and individual, interpersonal, community, and society/policy levels of influence. Factors contributed by the advisory board included those from two freelisting instruments. We used the resultant freelists to calculate a salience index of factors as a reference for prioritization. We organized the compiled factors in the social-ecological model matrix according to their theorized distal-proximal relationship with fatal opioid-involved overdoses. We operationalized the social-ecological model factors by matching them against measures in the Massachusetts Public Health Data Warehouse, which includes 26 individually-linkable datasets and 19 community-level datasets drawn from 85 data components.

Results

We identified 224 factors potentially associated with fatal opioid-involved overdoses and organized them in the social-ecological model. Of these, 53 had matches to measures in the Public Health Data Warehouse. Of those factors identified by freelisting, salience indexing further identified 10 as most related to the risk of fatal opioid-involved overdose, including housing stability, increased risk substances such as fentanyl, xylazine, and polysubstances, and using alone.

Conclusion

The opioid-overdose specific social-ecological model points to the need both for analysis that penetrates the complexities of the opioid crisis and for multi-faceted interventions. Further, the social-ecological model can provide a foundation for simulation models for prevention and intervention efforts. Our matrix-structured social-ecological model, salience index, and data matching table provide a holistic and relationship-oriented view of the factors associated with fatal opioid-involved overdose and will inform subsequent data analysis, model development, and opioid-involved overdose policy and prevention efforts.