Date of Award
Fall 2023
Document Type
Open Access Dissertation
Department
Epidemiology and Biostatistics
First Advisor
Melissa Nolan
Abstract
Statistical models have been utilized to understand and forecast infectious diseases across the globe. However, these models have historically been difficult to access and interpret for non-specialist audiences. This dissertation focuses on one computationally efficient model fitting method, the integrated nested LaPlace approximation approach, to forecast three arboviruses in hopes of encouraging accessibility to statistical modelling. Such accessibility could provide areas who simultaneously are most susceptible to infection, and have the least resources, an invaluable tool for informing policy and prevention decisions. The resulting models included biological, land cover, and environmental factors and produced accurate forecasts of neuroinvasive West Nile virus, dengue virus, and LaCrosse encephalitis virus in three at-risk areas. Analysis showed Bayesian spatio-temporal models performed best in small to mid-sized populations; they consistently underestimated counts and rates in largely populous areas in USA counties and Puerto Rico municipalities, yet very accurately predicted in smaller and mid-sized counties such as in Appalachia. Infectious disease forecasting has immense utility in informing governments and institutions’ prevention initiatives, and modelling techniques such as the integrated nested LaPlace approximation approach offer fast, efficient, and inexpensive ways to model such forecasts.
Rights
© 2024, Maggie Suzanne Jennave McCarter
Recommended Citation
McCarter, M. S.(2023). The Utility of Bayesian Predictive Models to Forecast Arboviruses in the United States of America and Territories. (Doctoral dissertation). Retrieved from https://scholarcommons.sc.edu/etd/7615