Long-term coexistence and regional heterogeneity of antibiotic-resistant infections reproduced by a simple spatial model

Antibiotic-resistant infections are a growing threat to human health. Mathematical models are a common tool to compare potential intervention strategies, but often struggle to reproduce a ubiquitous pattern seen in data: the long-term coexistence of both drug-susceptible and drug-resistant strains. Here we show that simple models of infection in structured or spatially-heterogeneous host populations lead to persistent coexistence where well-mixed models fail. This coexistence is robust over a wide range of treatment coverage, drug efficacies, costs of resistance, and mixing patterns. Perhaps more importantly, this model can explain other puzzling spatiotemporal features of drug resistance epidemiology that have received less attention, such as large differences in the prevalence of resistance between geographical regions that consume similar amounts of antibiotics or that neighbor one another. Our analysis identifies key features of host population structure that can be used to assess drug resistance risk and highlights the need to include spatial or demographic heterogeneity in models to guide resistance management.