Spatial Disparities in Vaccination and the Risk of Infection in a Multi-Region Agent-Based Model of Epidemic Dynamics

We investigate the impact that disparities in regional vaccine coverage have on the risk of infection for an unvaccinated individual. To address this issue, we develop an agent-based computational model of epidemics with two features: 1) a population divided among multiple regions with heterogeneous vaccine coverage; 2) contact networks for individuals that allow for both intra-regional interactions and inter-regional interactions. The benchmark version of the model is specified using county-level flu vaccination claims rates from California. We isolate the effects of heterogeneity by holding overall vaccination levels constant, while changing the variance in the distribution of regional vaccine coverage. We find that an increase in spatial heterogeneity leads to larger epidemics on average. This effect is magnified when more connections that are inter-regional exist in the contact structure of the networks. The central result in the paper is that there is a non-monotonic relationship between the infection risk and the geographic resolution of vaccination rate measurement. Infection risk of an unvaccinated individual decreases in both the global rate of vaccinations and the rate of vaccination of the individual’s specific contacts. Surprisingly, we find that the vaccination rate in an individual’s home region does not have a significant impact on an individual’s infection risk in our model. This has significant implications for an individual’s vaccine choices. Global and local (network specific) vaccination rates are highly correlated with infection risk and thus should be prioritized as information sources for rational decision-making. Using the region-specific information, however, is likely to lead to non-optimal decisions.