Spatially explicit epidemic models explore population-level consequences of interactions between neighboring infectious and susceptible individuals. Most such models equate local and global host density, so that each individual interacts with the same number of neighbors. But many natural populations exhibit heterogeneity in local abundance. Therefore, we let host dispersion vary from uniform to clumped in a spatial epidemic with recovery. We analytically approximated the epidemic with a branching process to show how the probability of pathogen extinction could depend on the degree of host clumping. We then simulated the detailed model across a range of parameter combinations. Both approaches to the problem indicate that host spatial aggregation strongly increases the chance of pathogen extinction.