Date of Award

5-17-2019

Degree Type

Thesis

Degree Name

Master of Public Health (MPH)

Department

Public Health

First Advisor

Lee Rivers Mobley, PhD

Second Advisor

Ruiyan Luo, PhD

Abstract

ABSTRACT

Spatial Analysis of County Level Drug Overdose Deaths and Associated Factors, Over Two Time Periods in the United States

By

SUNANDA SARKAR

April 18, 2019

INTRODUCTION: Recently, drug overdose is being considered as an important public health issue, the magnitude of which is yet to be adequately explored. The United States is experiencing a wide range of drug overdose problems over the past decades, where fatal overdoses have tripled from 1999 to 2016. Geographic approaches to drug overdose death research have emerged in recent years. Studies demonstrated that overdose mortalities are not equally distributed across different geographic areas. Therefore, it is important to consider geographic variations to inform effective prevention and treatment of drug overdoses and prevent premature deaths.

AIM: The aim is to explore spatial distribution of county level drug overdose death rates in the contiguous U.S. over two 5-year time intervals (2007-2011 and 2012-2016); identify and evaluate the extent to which the county level socio-economic and socio-demographic factors are associated with the spatial patterning and explain it.

METHODS: Exploratory spatial cluster analysis was performed to determine whether patterns of observed drug overdose mortality are spatially random or not over two time periods. Both traditional and Empirical Bayes standardization methods were used for spatial autocorrelation test. To determine any change over time, observations in the data are stacked based on time. Time stacked spatial regression analysis was performed to determine the associations between several county level socio-economic and socio-demographic factors and drug overdose death rates in the U.S.

RESULTS: Mean drug overdose death rate increased from early to late time period. Results indicates the presence of significant (at 5% significance level) spatial autocorrelation among the adjacent counties in the drug overdose death rates, and this spatial pattern differs in two time periods. Finally, spatial regression indicates that the effect of different contextual factors are heterogenous over time and across different population.

CONCLUSION: Findings may help inform efforts to prevent, diagnose or treat drug overdoses ahead of time, thus prevent premature deaths by understanding the geographic variations and identifying the areas with growing burdens. Studies focusing on similar associations across different age-groups and insured group may provide better insight.

Available for download on Monday, April 27, 2020

Share

COinS