Researchers from the School of Mathematics propose new approach that more accurately predicts the evolution of epidemics

Neha Bansal, Dr Katerina Kaouri and Dr Thomas E. Woolley’s newly published study could inform public health policy decisions.

The study, published in the Journal of Theoretical Biology, proposes a new way to improve sampling the population data required for public health policy making.

This new approach would improve the reliability of disease modelling for policymakers by selecting the most appropriate sampling algorithm on a case-by-case basis, looking at factors such as network structure and epidemic features.

The widely used Random Walk (RW) sampling method, which is applied in many epidemic studies, disproportionately selects highly connected individuals - an error known as ‘size bias’. As a result, this can lead to an inaccurate prediction of how an epidemic spreads.

The team evaluated a modified algorithm, the Metropolis‑Hastings Random Walk (MHRW) algorithm, which can more accurately sample data from the population, reducing size bias.

They concluded that RW is appropriate for fast-spreading, high-mortality epidemics, such as foot-and-mouth, or smallpox, or for use in social networks where everyone has similar connections. MHRW, on the other hand, would be better suited for slower and low-severity epidemics, for example, bovine viral diarrhoea, seasonal influenza, or for use in social networks that are highly diverse, like real social networks.

These findings have implications for responses to future epidemics, offering guidance for choosing the right sampling method according to the type of disease and the patternof interactions between people .

Dr Thomas E. Woolley, Reader in Applied Mathematics, said: “Although mathematicians worked hard to predict how Covid-19 would spread, our research shows that the real-world data collection process can heavily influence these predictions.”

Dr Katerina Kaouri, Reader in Applied Mathematics, said: “Our work contributes to a better understanding of epidemics and helps inform policy.”

This work is supported by the Natural Environment, Biotechnology and Biological Sciences and Medical Research councils (as part of the ‘OneZoo’ Centre for Doctoral Training which funds Bansal’s PhD studies), and by an EPSRC Impact Acceleration Account grant awarded to Dr Kaouri and Dr Woolley

You can read the paper, ‘Reducing size bias in epidemic network modelling’, in the Journal of Theoretical Biology.

This study is part of a multi-year, interdisciplinary research programme at the School of Mathematics on epidemic preparedness that started with Dr Kaouri and Dr Woolley advising policymakers during the Covid-19 pandemic.