Date on Master's Thesis/Doctoral Dissertation

5-2014

Document Type

Doctoral Dissertation

Degree Name

Ph. D.

Department

Industrial Engineering

Degree Program

Industrial Engineering, PhD

Committee Chair

Alexander, Suraj Mammen

Committee Co-Chair (if applicable)

Evans, Gerald

Committee Member

Evans, Gerald

Committee Member

Biles, William

Committee Member

Zeng, Weibin

Subject

Public health surveillance; Social networks--Health aspects; Information storage and retrieval systems--Public health

Abstract

Over the centuries, human beings have been inflicted with a variety of contagious diseases, resulting in tens of millions of respiratory illnesses and deaths worldwide. Early detection of disease spread facilitates timely responses that can greatly reduce its impact on a population. Therefore, this early information is a major public health objective and is crucial for policy makers and public health officials responsible for protecting the public from the spread of contagious diseases. Current indicators of the spread of contagious outbreaks lag behind its actual spread, leaving no time for a planned response. The studies of Christakis et al. in 2010 have shown that social networks can provide more timely information for prediction. However, the reported social network methods used to monitor disease spread do not consider contact patterns of individuals over space and time, such as during their movement from place to place. In this dissertation we propose a more effective way to chart the spread of contagious outbreaks, in a spatio-temporal sense, using “contact networks”. This enables more effective control of the spread of contagious outbreaks in their early stages so as to “nip a potential pandemic in the bud.” In order to enhance the prediction model developed we introduce factors to consider the intensity of exposure to the disease, and the susceptibility of the individual. This would involve the consideration of both space and time factors, since diseases caused by either viruses or bacteria involve some type of contact, either direct (e.g. shaking hands) or through the atmosphere (e.g. coughing or sneezing) between the susceptible and infected individuals. In this dissertation, we apply data mining methodologies and predictive modeling technologies, such as logistic regression, decision trees and neural networks to estimate the infection risk based on an individual’s demographic information and health status. The information used in the models can be obtained from a wide variety of data sources, including historical medical records from hospitals and clinics. Early information on the presence of a potential disease outbreak can be obtained from "sensors", such as, First Watch and EARS (Early Aberration Response Systems) and "central" individuals in “contact” networks.

Share

COinS