Forecasting COVID-19: Using SEIR-D quantitative modelling for healthcare demand and capacity

Campillo-Funollet, E.; Van Yperen, J.; Allman, P.; Bell, M.; Beresford, W.; Clay, J.; Evans, G.; Dorey, M.; Gilchrist, K.; Gurprit, P.; Walkley, R.; Watson, M.; Madzvamuse, A.

Rapid evidence-based decision-making and public policy based on quantitative modelling and forecasting by local and regional National Health Service (NHS-UK) managers and planners in response to the deadly severe acute respiratory syndrome coronavirus 2 (SARS-Cov-2), a virus causing COVID-19, has largely been missing. In this pilot study, we present a data-driven epidemiological modelling framework that allows to integrate quantitative modelling, validation and forecasting based on current available local and regional datasets to investigate and mitigate the impact of COVID-19 on local NHS hospitals in terms of healthcare demand and capacity as well as allowing for a systematic evaluation of the predictive accuracy of the modelling framework for long-term forecasting. We present an epidemiological model tailored and designed to meet the needs of the local health authorities, formulated to be fitted naturally to datasets which incorporate regional and local demographics. The model yields quantitative information on the healthcare demand and capacity required to manage and mitigate the COVID pandemic at the regional level. Furthermore, the model is rigorously validated using partial historical datasets, which is then used to demonstrate the forecasting power of the model and also to quantify the risk associated with the decision taken by healthcare managers and planners. Model parameters are fully justified, these are derived purely based on the time series data available at the regional level, with minimal assumptions. Using these inferred parameters, the model is able to make predictions under which secondary waves and re-infection scenarios could occur. Hence, our modelling approach addresses one of the major criticisms associated with the lack of transparency and precision of current COVID-19 models. Our approach offers a robust quantitative modelling framework where the probability of the model giving a wrong or correct prediction can be quantified.