Clustering of Synthetic Opioid Death Rates and Associated Factors in the U.S.

Presentation Type

Poster

Student

Yes

Track

Health Care Application

Abstract

Opioid overdose has become an epidemic as of 2013 according to the Center for Disease Control and Prevention in the United States, with around 1 million deaths from 1999-2020. Here we investigate the geographic patterns in age-adjusted death rates and some associated factors using model-based clustering. Clustering is accomplished through finite mixture models which are known for modeling heterogeneous data. More specifically, this is done through a \(L_1\) penalized mixture of regression models with parameter estimates calculated via the expectation-maximization algorithm. This allowed for variable selection as well as identifying US states with similar trends. The results suggest that increases in the rate of uninsured, lower-income populations, proportion of males, Asian populations relative to White populations, and the proportion of bachelor's degree or higher holders, correspond to a decrease in the expected death rate. Other factors had mixed responses depending on the cluster.

Start Date

2-6-2024 1:00 PM

End Date

2-6-2024 2:00 PM

This document is currently not available here.

Share

COinS
 
Feb 6th, 1:00 PM Feb 6th, 2:00 PM

Clustering of Synthetic Opioid Death Rates and Associated Factors in the U.S.

Volstorff A

Opioid overdose has become an epidemic as of 2013 according to the Center for Disease Control and Prevention in the United States, with around 1 million deaths from 1999-2020. Here we investigate the geographic patterns in age-adjusted death rates and some associated factors using model-based clustering. Clustering is accomplished through finite mixture models which are known for modeling heterogeneous data. More specifically, this is done through a \(L_1\) penalized mixture of regression models with parameter estimates calculated via the expectation-maximization algorithm. This allowed for variable selection as well as identifying US states with similar trends. The results suggest that increases in the rate of uninsured, lower-income populations, proportion of males, Asian populations relative to White populations, and the proportion of bachelor's degree or higher holders, correspond to a decrease in the expected death rate. Other factors had mixed responses depending on the cluster.