Center for Modeling Complex Interactions

Modeling efforts for COVID-19 within the US have focused primarily on helping urban centers cope with theconsequent health care crisis. The impact of the pandemic on rural communities is still emerging, and theseareas have not received the same degree of modeling attention. At the same time, rural communities aredifferent from urban centers in ways that affect the disease and its dynamics: they have lower densities, aremore isolated, have smaller social networks, tend to be poorer and older, and have scant health careinfrastructure. Rural communities are also the primary source of food production and natural resourceextraction in this country. As the pandemic unfolds across the coming months, rural communities will be facedwith highly variable circumstances: some will have no infections and be focused on early detection; some willhave active cases and be attempting to stop their spread; some will have eliminated active cases and beattempting to reopen economic and community activities while guarding against resurgence. Treating allcommunities as the same would be foolish. At the local level, decision makers need tools tailored to realcommunities: tools that emulate the way people come and go and interact there, tools to consider the mostrelevant interventions, and tools that account for real variation in how able and willing people will be to complywith possible interventions. At the larger health-district and state level, officials need forecasts of how localdecisions, health care infrastructure, and the virus itself will interact to drive the epidemic. The purpose of thecurrent proposal is to provide these tools by building a model of COVID-19 for largely rural states that links thedynamics within communities together into a statewide network. This will be achieved in three specific aims. InAim 1, we develop a predictive epidemiological model of COVID-19 spread and intensity for rural states. Thiswill be done with a spatial, age-structured metapopulation model that relies on differential equations and theirstochastic extensions. In Aim 2, we evaluate how potential interventions in individual communities affectoutbreak risk, transmission, access to health care, and intervention efficacy and adoption. Here we combinesurveys-of both rural and urban communities in Idaho and several broader regions of the US-to estimatepatterns of compliance and the motivations behind them. Using these results, we will then use agent-basedmodels of synthetic communities to simulate interventions. Net effects will be relayed up to the statewidemodel. In Aim 3, we provide support for decision making to state public health officials and local policy makersin rural communities. This will be done by developing two online graphical interfaces for visualizing forecastsand exploring interventions-one high-level application for non-specialists and a second, more sophisticatedversion, for public health professionals. Education, empowerment, and appreciation of uncertainty will beemphasized. Finally, the models and tools we develop here will be implemented in Idaho, but will be designedfor easy export to states with significant rural populations.