Quantifying the structure of transmission networks for future epidemics

The COVID-19 pandemic highlighted clear differences in the burden of disease between locations and communities, and sharp inequalities in infection and reinfection risk. Sociodemography, behaviour, employment, infrastructure, and public health interventions determined the opportunities individuals had to travel and interact, and thus were critical in determining individual-level infection risk.

An individual's infection risk can be amplified by the wider contact network they inhabit: infection risk is a product of individual behaviour and the behaviour of those around them. This 'higher-order' structure of contact networks ultimately determines variation in infection risk, and drives epidemic dynamics, yet is unquantified. A better understanding of the network structure of society will enable improved identification of communities at highest infection risk and settings with high-transmission, and more precise epidemic modelling.

This will enhance and inform the public health response to future outbreaks of respiratory pathogens.

The goal of this project is to identify the key network structures from the vast and underutilised contact tracing data collected in England during the pandemic, and to establish and assess methods to identify network structure when detailed contact tracing data is unavailable, through three aims:

Aim 1: Characterise the network structure of social interactions using data from the COVID-19 contact tracing program in England. The dataset contains interaction information between 27.9 million individuals. We will characterise how contact patterns differ by individual attributes (including age, ethnicity and occupational status), by geography, and over the course of the pandemic, from lock-downs to a fully open society.

We will use this network structure to identify where infection risk is greatest within geo-social space. We will also explore the role of specific contact settings - such as households, shops, personal services, visiting friends, workplaces - in transmission.

Aim 2: Critically assess the ability of population census data, coupled with pathogen genomic and telecom-mobility data, to recover the network structure observed from contact tracing. Large scale contact tracing data is unlikely to be readily available in future outbreaks. In this context, we will assess how different models applied to more readily available data sources can recover the network structure of contacts.

Additionally, we will identify if additional information about higher-order network structure can be inferred by combining genomic and tracing data.

Aim 3: Identify appropriate levels of population complexity required for general epidemic modelling and public health purposes. We will assess the required complexity of contact networks for them to be of use in policy decisions during future outbreaks and pandemics. We will identify the minimal set of characteristics (eg. age, ethnicity, vaccination status) required to accurately predict population-scale infection heterogeneities.

We will also assess the ability of telecom and census data sources to explain infection heterogeneity and transmission hotspots in the absence of tracing data.

This project will provide a comprehensive, high-resolution description of how England's population interacts, identify the structural elements of the network that generate observed variation in infection risk within the population, and inform data requirements for future epidemics. It will provide a critical assessment of the minimal set of data that can reconstruct the interaction structure of the population in situations when extensive contact tracing is unavailable.

This will be especially useful in revealing the contact network structure of populations in other countries, where tracing has not been performed, and to understand the likely course of future epidemics.