Regional COVID-19 spread despite expected declines: how mitigation is hindered by spatio-temporal variation in local control measures

Abstract

Successful public health regimes for COVID-19 push below unity long-term global R_t –the average number of secondary cases caused by an infectious individual. Most assessments use local information. Populations differ in R_t, amongst themselves and over time. We use a SIR model for two populations to make the conceptual point that even if each locality averages R_t < 1, the overall epidemic can still grow, provided these populations have asynchronous variation in transmission, and are coupled by movement of infectious individuals. This emergent effect in pandemic dynamics instantiates “Parrondo’s Paradox,” -- an entity comprised of distinct but interacting units can behave qualitatively differently than each part on its own. For effective COVID-19 disease mitigation strategies, it is critical that infectious individuals moving among locations be identified and quarantined. This does not warrant indiscriminate prevention of movement, but rather rational, targeted testing and national coordination.