Abstract
A model is presented that brings together two apparently conflicting observations: There were transitions from peaks of activity to prolonged periods of fluctuation in relatively narrow bands, indicating strong constraints for possible transmission paths; yet, mutant strains could replace wild-types rapidly, indicating that there were enough transmission paths from the mutant index cases to the nodes in the contact chains from active wild-type cases to block the transmission. It is shown that this is explainable if the spread was driven and modulated by spread in open reservoirs, such as hospitals, where prevalence could persist because there was continuous admission of susceptible individuals while the duration of stay was sufficient to support transmission chains, and where strains competed for the resource of susceptible admissions. Furthermore, strong segmentation of the general community allowed multiplication of the nosocomial cases without supporting long transmission chains outside hospitals which would have kept the wild type in the population.