Agent-Based Simulation of Covid-19 Vaccination Policies in CovidSIMVL

Abstract

ABSTRACTAn agent-based infectious disease modeling tool (CovidSIMVL) is employed in this paper to explore outcomes associated with MRNA two-dose vaccination regimens set out in Emergency Use Authorization (EUA) documents submitted by Pfizer and Moderna to the US Department of Health & Human Services. As well, the paper explores outcomes associated with a third “Hybrid” policy that reflects ranges of expected levels of protection according to Pfizer and Moderna EUA’s, but entails a 35 day separation between first and second dose, which exceeds the 21 days set out in Pfizer documentation or the 28 days in Moderna documentation.Four CovidSIMVL parameters are varied in the course of 75 simulated clinical trials. Two relate directly to the vaccines and their impacts (duration between doses; degree of expected protection conferred by different vaccines following first or second dose). Two relate to the simulation contexts to which the vaccines are applied (degree of infectivity; duration of infectivity). The simulated trials demonstrate expected effects for timing of second dose, and for degree of protection associated with first and second dose of Pfizer and Moderna vaccines, and the effects are consistent with an assumed value of 75% for degree of protection after first and second doses for the Hybrid vaccine. However, the simulated trials suggest a more complex interaction between expected level of protection following first dose, timing of second dose and degree of infectivity. These results suggest that policy options should not be considered independent of the transmission dynamics that are manifested in the contexts in which the policies could be applied.CovidSIMVL embodies stochasticity in the mechanisms that govern viral transmission, and it treats the basic reproduction number (R0)as an emergent characteristic of transmission dynamics, not as a pre-set value that determines those dynamics. As such, results reported in this paper reflect outcomes that could happen, but do not necessarily reflect what is more or less likely to happen, given different configurations of parameters. The discussion section goes into some measure of detail regarding next steps that could be pursued to enhance the potential for agent-based models such as CovidSIMVL to inform exploration of possible vaccination policies, and to project outcomes that are possible or likely in local contexts, where stochasticity and heterogeneity of transmission must be featured in models that are intended to reflect local realism.