Understanding the impact of modelling assumptions and population heterogeneity on the robustness of outputs of different epidemiological models in the context of the COVID-19 pandemic and beyond

Thesis

The rapid emergence of the SARS-CoV-2 virus in late 2019 and its high virulence generated extensive public interest from both specialists and non-specialists alike and has highlighted the importance of mathematical modelling in epidemiology.

In this DPhil project, we explore how differing modelling approaches used in epidemiology can impact the recommendations made by scientists and scientific advisory groups to policy makers in Government, with a particular focus on the decisions made in the UK during the COVID-19 pandemic. The first major focus of this work revolves around a class of models popularly used in epidemiology, known as renewal models. These models require fewer assumptions to be made compared to alternative modelling techniques, yet still output quantities of relevance to policy. Classic renewal modelling approaches treat a population as a homogeneous group and ignore differences in contact patterns that may vary systematically across groups (e.g. that teenagers typically have more contacts per day than those aged 65+). We develop de novo renewal models to incorporate the effects of population heterogeneity for better prediction of public health outcomes, and derive straightforward and entirely new mathematical results to assess the long-term behaviour of a pathogen within a multi-layered population.

Finally, we compare three of the key compartmental models that were used by the UK government and the industrial research sector to model the COVID-19 UK epidemic. We reimplement these models in a unified framework using industrial-strength software engineering practices by creating an open-source package called ‘epimodels’ [Bouros, 2021]. This work highlights the pitfalls of relying on individual models to inform policy responses for future epidemics and pandemics, as well as the need for a more in-depth study of the impact of modelling assumptions on the quality of model outputs.

Authors: Bouros, I

• DOI: 10.5287/ora-dojpx5kkv
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: renewal equation models; epidemiology; Bayesian inference; compartmental ODE models; model comparison
• Subjects: Infectious disease modelling