A network model at both the population and individual levels, which simulates both between-patch and within-patch dynamics, is proposed. We investigated the effects of dispersal networks and distribution of local dynamics on the outcome of an epidemic at the population level. Numerical studies show that disease control on random networks may be easier than on small-world networks, depending on the initial distribution of the local dynamics. Spatially separating instead of gathering patches where disease locally persists is beneficial to global disease control if dispersal networks are a type of small-world networks. Dispersal networks with higher degree lead to a higher mean value of R0. Furthermore, irregularity of network and randomization are beneficial to disease stabilization and greatly affect the resulting global dynamics.