Download files
Abstract
Avian Influenza (AI) is a viral disease that affects birds and humans. Several AI lineages infect humans and there is potential for lineages to cause human pandemics. New developments in AI have occurred in recent years: several novel zoonotic strains (e.g. H7N9, H5N6, H10N8) emerged in China, and North America experienced mass outbreaks of novel lineages in commercial poultry. This presented an opportunity to study emerging strains at a time when there was greatest uncertainty about the nature of these viruses and how they would behave. This PhD aimed to use and compare novel epidemiologic, modelling, geospatial and computational methods to uncover epidemiological insights, particularly to evaluate risk and patterns of spread of novel viruses. In chapters 2 and 4, publically available information was used to describe trends and patterns of recently emerged zoonotic AIs in comparison to historical occurrences of zoonotic AI, and to quantitatively characterise the prevalence of two potentially pandemic strains (H5N1 and H7N9) in avian populations through a meta-analysis. In chapters 3, 5 and 6, spatial data was used to identify locations of highest disease incidence, predict areas at risk of disease incursions, and identify factors driving disease spread. Chapter 3 used a spatial overlaying approach to show spatial dynamics of H5 clade 2.3.4.4 viruses in North America appeared were more consistent with domestic poultry distribution compared to migratory flyways. Chapter 5 combined a machine learning algorithm with a spatial risk framework to identify specific areas at risk of H5N1 and H7N9 transmission. In Chapter 6, H5N1 and H7N9 spread was modelled using a Bayesian phylogeography approach, and factors contributing to disease spread were identified. There is a vastly growing amount of digital data sources and computational tools now available to the desktop epidemiologist. This thesis provides examples of how these can be used; the final chapter discusses their benefits and pitfalls, and outlines how they could be improved. Whilst spatial risk models identified specific areas of high AI risk, they did not account for movement. Phylogeography methods were able to uncover space-time dynamics, albeit at a much lower spatial resolution. Future efforts could aim to connect disparate data across multiple spatial and temporal scales, and to incorporate mechanistic models of disease transmission to predict disease trajectories over time and space.