Data-driven analytics of disease spread under close contact for optimal testing and mitigation

Abstract

Models emulating the spread of infectious diseases in close-contact environments present a set of unique challenges. This field of research has exploded over the past three years due in part to the SARS-CoV-2 pandemic. In this thesis, we present and explore five different models of infection in close-contact environments which aim to fulfill five different needs. The first model (Chapter 2) is used to study the ability to estimate the outbreak size, i.e. the total number of individuals in a group who have been infected following an exposure event, based on the number of observed symptomatic individuals by a certain time after the event. Of the three quantities we investigated, the proportion of individuals who have been observed to be symptomatic, the outbreak size and the group size, the first one is shown to have the greatest influence on the entropy of the resulting predicted distribution for the outbreak size. The second model (Chapter 3) explores the effect of rota patterns on workplace infection rates by calculating a central estimate for the length of time an individual is at work whilst infectious. We first explore this model numerically and then approach an analytical solution by representing rota patterns as a Fourier series. In both cases, we find that longer shifts reduce in-work infectiousness and, for a parameter set emulating SARS-CoV-2, a rota length of approximately 10–11 days is optimal. Nosoco (introduced in Chapter 4) is a tool we have generated for approximating the total number of in-hospital infections based on the timing of positive swabs. We explore how efficient it is as a tool compared to declaring as nosocomial infections all individuals diagnosed after a fixed number of days since their admission, explore how the proportions of total cases attributed to infection within and outside the hospital changes over time and estimate each individual’s daily rate of infection. The fourth model (Chapter 5) is used to estimate the incidence of Hepatitis C from results of a cross-sectional survey when assuming a constant rate of infection. We use an example study of a cross-sectional survey of Scottish prisons to show no evidence, among people who inject drugs, of a lower incidence in prison compared to the external incidence . The final model (Chapter 6) incorporates a distance element into transmission trees as we investigate an outbreak aboard a cruise-liner, using cabin location as a proxy for infector location.

Date of Award	31 Dec 2023
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Lorenzo Pellis (Supervisor), Ian Hall (Supervisor), Andrew Ustianowski (Supervisor) & Thomas House (Supervisor)