Lorenzo Pellis: Inferring risk of coronavirus transmission from community household data

Abstract: Households are well known to be amplifiers of transmission of respiratory pathogens, due to the intimate and stable nature of contacts within them. Household-structured epidemiological data can be seen as a collection of repeated “transmission experiments” that naturally lends itself to parameter estimation. The Office for National Statistics COVID-19 Infection Survey (CIS) [1], a prospective longitudinal household survey in England and one of the largest studies of community SARS-CoV-2 infection in the world, offered an unprecedented opportunity for estimating how individual characteristics affect transmission risk, but also creates a range of challenges due to its unusual design and size. Existing methods that correctly account for the possibility of multiple introductions from outside the household are particularly suited for single-wave data followed by serology. However, the multiple pandemic waves and time-varying changes in transmission required adapting existing methods in real time to analyse data from the CIS [2]. I will discuss the numerous conclusions that we were able to draw from this work (e.g. Alpha variant 50% more transmissible than what seen before and 35% less than Delta within household; 25-300% higher risk for workers in patient-facing roles to bring the infection in the household before vaccination; children at increased risk of bringing the infection in the household when schools are open; etc.), as well as power and limitations of current methods. With the explosion of data collection caused by the pandemic, household studies are only likely to become more common in the future, making novel methodological developments urgently needed if we want to maximise the amount of information extracted from such expensive but potentially extremely informative studies.

Joint work with Heather Riley, Denys Z. A. Janes and Thomas House