Date of Award

Summer 8-11-2023

Embargo Period

7-31-2028

Document Type

Dissertation - MUSC Only

Degree Name

Doctor of Philosophy (PhD)

Department

Public Health Sciences

College

College of Graduate Studies

First Advisor

Brian Neelon

Second Advisor

Hong Li

Third Advisor

John Pearce

Fourth Advisor

Sara Benjamin-Neelon

Fifth Advisor

Noel Mueller

Abstract

COVID-19 has created a global health crisis since its emergence in late 2019. According to the Centers for Disease Control and Prevention (CDC), there have been over 104 million cases and 1.13 million deaths reported in the United States (US) as of June 2023. COVID-19 data are complex and present a number of statistical challenges that must be addressed to ensure valid inferences. These challenges include overdispersion of case and death counts, modeling of complex health effects, and zero-inflation. This dissertation proposes three aims to address these challenges: In Aim 1, we develop a Bayesian negative binomial regression model with spatially varying dispersion. This aim extends existing methods to address the issue of spatial heterogeneity and varying degrees of overdispersion in COVID-19 incidence data across counties. Using a simulation study, we demonstrate that ignoring heterogeneity in dispersion can lead to biased and inefficient estimation. For illustration, we apply the model to study the effect of social vulnerability index (SVI) on COVID-19 incidence from March 15 to December 31, 2020 in the state of Georgia. In Aim 2, we extend the current methods for continuous outcomes in environmental health mixture studies to count settings and develop a negative binomial Bayesian kernel machine regression (BKMR) method to model complex exposure-response associations involving count outcomes. Using a simulation study, we evaluate the performance of the proposed method in estimating the exposure-response function and identifying the most relevant mixture components. We apply the proposed method in modeling the joint effect of the social vulnerability index (SVI) variables on COVID-19 deaths from January 1 to December 31, 2021 in South Carolina. In Aim 3, we extend marginalized zero-inflated models to spatial setting and develop a marginalized zero-inflated negative binomial model for spatial data. In addition to capturing zero-inflation, the proposed method allows for direct modeling of the marginal mean, which is often the target of interest in public health and disease mapping studies involving zero-inflated data. We conduct simulation studies to investigate the features of the model and use the model to examine predictors of COVID-19 deaths in the US state of Georgia for the 2021 calendar year.

Rights

Copyright is held by the author. All rights reserved.

Download

Available for download on Monday, July 31, 2028

Share

COinS