Addressing hospitalisations with non-error-free data by generalised SEIR modelling of COVID-19 pandemic

Abstract

Successive generalisations of the basic SEIR model have been proposed to accommodate the different needs of the organisations handling the SARS-CoV-2 epidemic. These generalisations have not been able until today to represent the potential of the epidemic to overwhelm hospital capacity until today. This work builds on previous generalisations, including a new compartment for hospital occupancy that allows accounting for the infected patients that need specialised medical attention. Consequently, a deeper understanding of the hospitalisations rate and probability as well as of the recovery rates for hospitalised and non-hospitalised individuals is achieved, offering new information and predictions of crucial importance for the planning of the health systems and global epidemic response. Additionally, a new methodology to calibrate epidemic flows between compartments is proposed. We conclude that the two-step calibration procedure is able to recalibrate non-error-free data and showed crucial to reconstruct the series in a specific situation characterised by significant errors over the official recovery cases. The performed modelling also allowed us to understand how effective the several interventions (lockdown or other mobility restriction measures) were, offering insight for helping public authorities to set the timing and intensity of the measures in order to avoid the implosion of the health systems.