CovidSIMVL – Agent-Based Modeling of Localized Transmission within a Heterogeneous Array of Locations – Motivation, Configuration and Calibration

CovidSIMVL is an agent-based infectious disease modeling tool that is designed specifically to simulate localized spread of infectious disease. It is intended to support tactical decision-making around localized/staged re-institution of pre-pandemic levels and patterns of social/economic/health service delivery activity, following an initial stage of pan-societal closures of social/economic institutions and broad-based reductions in services.

By design, CovidSIMVL supports the generation of dynamic models that reflect heterogeneity within and between a network of interacting localized contexts. This heterogeneity is embodied in a hierarchically organized set of rules. Primary rules reflect the pathophysiology of transmission. Secondary rules (“HazardRadius” and “Mingle Factor” in CovidSIMVL) relate transmission to proximity and movement within physically demarcated and relatively contained spaces (“Universes”). Tertiary rules (“Schedules”) relate probabilities of transmission to movement of people between a network of localized contexts (a CovidSIMVL “Multiverse”).

This report focuses mainly on calibration of secondary rules. To calibrate the HazardRadius and MingleFactor parameters, growth curves were generated with CovidSIMVL by setting different configurations of values on those two proximal determinants of viral transmission. These were compared to the characteristic shapes of curves generated by equation-based compartmental models (e.g., SEIR models) that fit different real-world datasets embodying different reproduction numbers (R ₀ ).

By operating with parameter values in CovidSIMVL that generate “real-world” growth curves, the tool can be used to produce plausible simulations of localized chains of transmission. These include transmission among different groups of persons (e.g., staff, patients) who are co-located within a single setting such as a long-term care facility. The Multiverse version of CovidSIMVL can be used to simulate localized cross-over transmission among arrays consisting of both unaffected and impacted contexts and associated sub-populations, via agents who interact within and across arrays of contexts such as schools, multigenerational families, recreational facilities, places of work, emergency shelters for homeless persons, or other settings in which people are in close physical proximity.