Improving sampling design and surveillance strategies for inferring the spatiotemporal dynamics of emerging infectious diseases

Thesis

Emerging infectious diseases represent a continuing threat to global public health. Our ability to respond effectively requires accurate inference of where and when pathogens first emerged, and how they subsequently spread through populations. Despite recent progress, there remains substantial gaps in disease surveillance systems, resulting in incomplete or biased sampling and delayed situational awareness. The primary goal of this thesis is to better understand the implications of these limitations for both downstream inferences and outbreak response, and to develop more effective sampling design and surveillance strategies.

In Chapter 2, I apply phylogeographic methods to the introduction and subsequent local spread of SARS-CoV-2 Omicron BA.1 in the UK, demonstrating how human mobility shaped its dissemination across multiple spatial scales. I also show that travel restrictions implemented at the time were largely ineffective, partly due to the delayed detection of local transmission in international travel hubs. This finding motivates Chapter 3, where I investigate how limited testing resources can be optimally allocated across a mobility network for more accurate inference of the underlying disease distribution during an epidemic. Chapter 4 examines how undersampling of local infections leads to underestimation of viral importation – a limitation in phylogeographic inference highlighted in Chapter 2 – by developing a theoretical framework that characterises the underlying sampling process. Building on this, Chapter 5 explores the broader impacts of heterogeneous sampling on phylogeographic inference and how different sampling strategies can mitigate them, using a simulation-based evaluation framework developed in this work.

Together, these studies provide insights into how limitations in disease surveillance affect our ability to infer the spatiotemporal dynamics of pathogen spread, while offering practical approaches for mitigating these biases. I conclude by discussing how methodologies developed in this thesis can be generalised to other questions in epidemiology and public health, particularly considering recent advances in artificial intelligence.

Authors: Tsui, L

• DOI: 10.5287/ora-qaqoegxb6
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: disease surveillance; machine learning; epidemiology; phylodynamics
• Subjects: Epidemiology