Date of Award

2017

Document Type

Open Access Dissertation

Department

Epidemiology and Biostatistics

Sub-Department

The Norman J. Arnold School of Public Health

First Advisor

Anwar Merchant

Abstract

Foodborne illness remains a serious public health problem in the United States in general as well as South Carolina in particular. Obtaining good food ingestion histories as well as possible risky environmental exposures is one of the earliest, most important tasks to complete in any foodborne outbreak investigations. Because time is of the essence in investigations, we have evaluated a rarely used biostatistical method, Random Forests, to data obtained from DHEC. Random Forests has the potential to facilitate more rapid identification of foods or environmental exposures that may be associated with outbreaks. We also examined previous cases of salmonellosis using two different definitions (state and FDA) of what constitutes a foodborne outbreak using logistic regression with a Poisson distribution. Dietary patterns were similarly evaluated, as they are associated with mortality from all causes. We aimed to characterize the nutrition and dietary intake of South Carolina residents and see what foods eaten may be associated with foodborne outbreaks. In summary, we have used Random Forests to analyze data that are routinely collected during foodborne outbreak investigations. This new application of Random Forests can make identification of foods responsible outbreaks more efficient. This information will address the challenges of a rural southern state with a high obesity rate by using a representative sample that contains geographic and socio-demographic diversity and using said information to help affect change in the programs available. The results of this study can potentially improve foodborne disease outbreak investigations in South Carolina.

Rights

© 2017, Alecia T. Alianell

Included in

Epidemiology Commons

Share

COinS