Plenary Speakers

Uncertainty and error in SARS-CoV-2 epidemiological parameters inferred from population-level epidemic models

In spite of the complexity of SARS-CoV-2 epidemiology, simple epidemic models invoking a homogeneous population have been used to describe the epidemic at a national scale and infer, for example, the timing of shifts in viral transmission rates and the efficacy of non-pharmaceutical interventions. Whether inference based on these models is reliable depends on whether the models embody appropriate distributions for host infectiousness, and for the time from infection to outcomes such as symptom onset, hospitalisation and death. We introduce an SIR-type model with the infected population structured by ‘infected age’, i.e. the number of days since first being infected, a formulation that enables distributions to be incorporated which are consistent with clinical data. We show that inference based on simpler models without infected age, which implicitly misspecify these distributions, is inaccurate for quantities relevant to policy-making, such as the reproduction number and the impact of interventions. We consider uncertainty quantification via a fully Bayesian approach, implementing this for both synthetic and real data focusing on UK data in the period 15 Feb–14 Jul 2020, and emphasising circumstances where it is misleading to neglect uncertainty.

Hybrid Random Graph and ODE Models: A Space Mapping Approach to Disease Dynamics

Differential equation models are usually easy to implement and give simulation results in few seconds while agent-based and random-graph models need days of simulation times, even weeks when scaled on the whole population. This is the reason why in most applications differential equation-based models are more preferred than a random graph or agent-based models. The main drawback of differential equation models is however the lack of heterogeneity. In this talk, I want to show a technique that combines both models and leads to a hybrid model which is quicker to compute but still prevails the heterogeneity.