Evaluating the use of social contact data to produce age-specific short-term forecasts of SARS-CoV-2 incidence in England

Mathematical and statistical models can be used to make predictions of how epidemics may progress in the near future and form a central part of outbreak mitigation and control. Renewal equation based models allow inference of epidemiological parameters from historical data and forecast future epidemic dynamics without requiring complex mechanistic assumptions. However, these models typically ignore interaction between age groups, partly due to challenges in parameterising a time varying interaction matrix. Social contact data collected regularly during the COVID-19 epidemic provide a means to inform interaction between age groups in real-time. We developed an age-specific forecasting framework and applied it to two age-stratified time-series: incidence of SARS-CoV-2 infection, estimated from a national infection and antibody prevalence survey; and, reported cases according to the UK national COVID-19 dashboard. Jointly fitting our model to social contact data from the CoMix study, we inferred a time-varying next generation matrix which we used to project infections and cases in the four weeks following each of 29 forecast dates between October 2020 and November 2021. We evaluated the forecasts using proper scoring rules and compared performance with three other models with alternative data and specifications alongside two naive baseline models. Overall, incorporating age interaction improved forecasts of infections and the CoMix-data-informed model was the best performing model at time horizons between two and four weeks. However, this was not true when forecasting cases. We found that age group interaction was most important for predicting cases in children and older adults. The contact-data-informed models performed best during the winter months of 2020–2021, but performed comparatively poorly in other periods. We highlight challenges regarding the incorporation of contact data in forecasting and offer proposals as to how to extend and adapt our approach, which may lead to more successful forecasts in future.Author summary: Short term epidemic forecasts help policy makers to plan and implement response activities. It can be useful to have such forecasts separately for different age groups, in order to reflect potential differences in transmission and incidence of infection between age groups as well as differences in risk of severe disease or death. A key challenge in developing age-specific models is understanding how different age-groups interact. We used data collected during a large-scale weekly survey of social contacts in the UK to inform this interaction in a model for short-term forecasts of COVID-19. To assess whether allowing interaction between age-groups and the use of contact data improved forecasts, we compared our forecasts to those from a set of models that either didn’t use current contact data or treated each age group as a separate population. We found that including timely contact data improved predictions when forecasting two to four weeks into the future, but this improvement was not consistent throughout the epidemic. The best improvement was measured during a long national "lockdown". We also found that inclusion of age-group interaction and use of contact data were most important when forecasting infections in older adults and young children.