METHODS FOR MODELING THE SPREAD OF INFECTIOUS DISEASE

Abstract

Mathematical and statistical models are widely used in studying infectious disease. They provide important insights – including mechanisms of the spread of infectious disease, forecast epidemic size and duration, and effects of intervention strategies – which are useful in studying and combating infectious disease. Over the last couple of decades, modeling techniques have advanced tremendously due to improvements in computational power, data availability, and data accessibility; this enables researchers to use various modeling approaches to capture more realistic aspects of infectious disease epidemics. Despite having flexible modeling techniques, these approaches use different modeling assumptions to incorporate information and propagate uncertainty, often arriving at inconsistent conclusions. My work focuses on exploring and improving methods for modeling the spread of infectious disease; in particular, exploring the state of the art techniques for disease modeling in real epidemic outbreaks and simulation settings. Motivated by a synthetic forecasting challenge inspired by the 2014 West African Ebola outbreak, I compared simple Markov chain Monte Carlo approaches to simulated epidemics (Chapter 2). Using high-resolution data from an ongoing rabies contact- tracing study, I apply robust techniques to reassess global historical risk estimates of canine rabies (Chapter 3), and show that disease trait correlations bias generation time estimates, with implications for conclusions about control (Chapter 4). In Chapter 5, I developed a method to improve modeling trait relationships while incorporating phylogenetic relationships by reformulating phylogenetic mixed models to improve flexibility and speed.