A Gaussian-Mixture Model Analysis of Polysubstance Drug Use in Opioid Overdose Deaths

Background: In the midst of the opioid crisis, it is imperative to identify risk factors for various subgroups of polysubstance drug users to reduce the risk of mortality. Methods: A Gaussian-Mixture Model analysis utilizing age group, White or African American race, sex, and dichotomized presence of illicit opioids (e.g., fentanyl), stimulants (e.g., cocaine), and benzodiazepines (e.g., Xanax) was conducted to develop advanced characterizations of subgroups and polysubstance use. 3,318 accidental overdose deaths (ICD10 X40-X44) from the Allegheny County Office of the Medical Examiner from years 2008-2019 were included in the analysis. Results: Nine demographic and substance use subgroups were identified. Of those, three may have particular implications for tailoring interventions for polysubstance use: (1) White females, ages 35-44, with presence of benzodiazepines and opioids, (2) older African American males, ages 55-64, with presence of illicit opioids and stimulants, and (3) White males, ages 35-44, who are utilizing heroin and/or prescription opioids. Conclusion: The heterogeneity of the polysubstance use in Allegheny County makes it necessary to develop further advanced characteristics of the subgroups being impacted by this epidemic. Statistical learning and GMM provided an optimal tool to generate such inferences. Public health significance: This analysis, identifying clusters of subgroups in accidental opioid overdose deaths, can be utilized to inform public health professionals and policy experts as to how to further tailor and improve interventions for the opioid epidemic in Allegheny County