Analysing COVID-19 Outbreaks Through Deterministic and Stochastic Agent-Based Models with Public Perception

Abstract This study compares the newly developed stochastic agent-based model (ABM) and deterministic system within the context of COVID-19 transmission dynamics, using the Susceptible-Exposed-Infectious-Removed-Perception (SEIRP) compartments. The primary aim is to analyse the similarities and differences between these modelling approaches, providing insights into the emergent behaviours of the epidemiological systems. We also investigate how social phenomena like public perception affect the two epidemiological models’ outcomes, focusing on scenarios with varying awareness levels and proportions of severe cases. Our results show that while deterministic model outputs align well with the ABM for large populations, discrepancies emerge for small populations in the ABM due to the impacts of stochastic extinction and discreteness of individuals. In scenarios with high proportions of severe cases and for large population sizes, the deterministic model exhibits oscillatory behaviour. In this situation, the averaged ABM densities initially capture the fluctuation dynamics when a substantial number of realisations is used in simulation; however, this stochastic system exhibits diminishing fluctuations across different realisations, contributing to a consistent average akin to an endemic steady state over a longer period. Interestingly, as the number of realisations is reduced, the agreement between stochastic and deterministic systems in depicting recurrent outbreaks is evident (i.e., realisations-dependent dynamical behaviour). In the next case study, the joint effect of recovery, latency period, disease severity and public perception is explored, highlighting how different factors can combine to influence the systems’ outcomes. Notably, our case study finds that even when the deterministic model demonstrates the persistence of sustained oscillations, the ABM can depict an extinction state; this stochastic ABM observation is caused by the realisations in this epidemiological system fluctuating to a very low population density and being excluded over a long run. Overall, our findings suggest the importance of considering both deterministic and stochastic models in infectious disease modelling, highlighting the need for comprehensive analyses to guide evidence-based decision-making in public health and epidemiology.