Clustered populations have also been modeled in the analysis of the spread of TB. Schinazi (2002) explored epidemiological dynamics in a socially clustered population, and showed that the likelihood of an epidemic depends on the infection rate from outside the cluster as well as the cluster size being large enough. Belhadji and Lanchier (2006) used a similar model to determine that epidemics are highly unlikely in a “cluster recovery process” population, where there is a good tracking system in place as soon as an infection is identified. Other models cluster individuals into households (Ball and Lyne, 2002; Ball et al., 2004). A network-based model constructed by Keeling (2005) predicted an infection would persist in a clustered population much longer than in a mass-action model. Beyers et al. (1996), who studied the geographic distribution of TB infections in two Western Cape (South African) suburbs and found that infection persisted in certain houses, also showed empirical evidence for the importance of spatial effects. While human contact is more complex than either of these extremes, a robust model ought to include elements of each.
