The statistics underpinning COVID-19 research and informing policy-making decision worldwide.
On 11 March 2020, the World Health Organization declared a pandemic: SARS-CoV-2, the virus that causes COVID-19, had spread far and wide around the world. After the initial outbreak in China, the epicentre shifted to Europe, where countries were considering drastic strategies to control the epidemic.
Understanding COVID-19 infections in real-time
During this first wave of COVID-19, and due to limited testing and asymptomatic spread, reports of case numbers were not a reliable measure of the spread of SARS-CoV-2. This meant that existing approaches to quantifying R0, the basic reproduction number, and R(t), the time-varying reproduction number, were inadequate to answer the pressing questions facing public health authorities worldwide: How far had SARS-CoV-2 spread? Were control measures effective and, in particular, did they bring R below 1?
A group of Imperial academics with expertise in epidemiology, statistics, and machine learning, set out to answer these questions. Led by Seth Flaxman and Axel Gandy (Department of Mathematics) and Swapnil Mishra, Samir Bhatt and Juliette Unwin (School of Public Health), they started work on a new semi-mechanistic statistical modelling framework to understand the extent of the spread of SARS-CoV-2 infections in real-time, and to infer the effect of government interventions.
Research with global impact
The framework allowed governments to quantify the effectiveness of non-pharmaceutical interventions in combatting the pandemic, especially lockdowns, contributing to worldwide understanding of the effect of such measures and providing scientific backing for communicating the importance of social distancing measures.
Its since been used to inform the decisions of policy makers worldwide, for example, in the UK and the US, affecting the lives of millions of people. During subsequent waves the modelling framework was taken up by health authorities to guide ongoing policy decision-making and was a key component in determining the speed of spread of the Alpha variant first detected in England. It also provided the evidential basis for the introduction of Tier 4 measures in England in December 2020.
How the research evolved
The first results showed that lockdowns were beginning to be effective across Europe, in terms of bringing the reproduction number below 1 at a time when death counts were still rising. It also estimated how many lives had already been saved through the non-pharmaceutical policy interventions.
From these methods and approaches, further research activity evolved at Imperial, including specific studies of the epidemic in Italy, Brazil and the USA, each including key novelties in the statistical models.
COVID-19 news stories
Italy, May 2020
The Italy report was a subnational analysis incorporating data on human mobility.
“Using a model based on mobility patterns across Italy, we consider various scenarios for the post-lockdown period. Without effective community surveillance, even a partial return to pre-lockdown levels of mobility could lead to a resurgence in the epidemic...”
Dr Seth Flaxman
Department of Mathematics, May 2020
Brazil, May-July 2020
Brazil provided the first subnational analysis of the reproduction number in Brazil, finding that R(t)>1 in all states analysed, meaning that the epidemic was not under control in Brazil in May.
USA, May-July 2020
The US report correctly warned about the precarious position of many states in the US at a critical period after the first wave:
Developing tools to inform decision-making
We hope this will be a useful tool for local and national governments trying to bring hotspots under control."
Professor Axel Gandy, Department of Mathematics
September 2020
Based on the methodology developed in these studies, Imperial researchers developed an R package called “epidemia”.
In addition, the team also developed a local area model for the UK, made public on 3 September 2020, which provides projections of COVID-19 cases per 100,000 individuals at the local-authority (LTLA) level across all four nations of the UK. Daily updates to these projections are released on a public website. Estimates of R(t) from this model were essential to the rapid analysis of the increased transmissibility of the new B.1.1.7 variant of the virus in December 2020.
View other impact case studies
Find out more about some of the research happening in the Department of Mathematics and how our academics are having real-world impact.
