Influenza pandemics pose a serious global health concern. When a new influenza virus emerges with pandemic potential, stochastic simulation models are used to assess the effectiveness of different intervention and containment strategies. Most simulation models assume "random mixing" within homes, schools, and workplaces: that is, contacts occur with equal probability within these mixing groups. In reality, social structure is more complex. In this dissertation, we model social contact networks within these mixing groups using surveys of contact behavior.

As households and schools are both primary mechanisms of influenza spread, much of this dissertation focuses on these two mixing groups. We model within-household contact networks using egocentric diaries of social contacts. Our model estimates the probability distribution of the within-household contact network for a household of arbitrary size and age composition. We find evidence for departure from the random mixing assumption within households, as well as evidence for a holiday effect and a household size effect on within- household contact behavior. Next, we use friendship network data in a high school as well as a survey on epidemics in high schools to build a detailed, dynamic within-school contact network model for a school with approximately 1100 students. We simulate influenza transmission over the contact network and find results to differ substantially from simulations over a random mixing scenario. We also simulate two interventions: a reactive grade closure strategy and a targeted antiviral prophylaxis strategy. We find estimates of intervention impact differ between the network model and the random mixing scenario. We also propose a method to model within-workplace contact networks from egocentric data. Finally, we discuss how to extend our models to include contacts occurring outside these mixing groups and model the network of contacts in an entire community. An additional chapter of this thesis uses social network methodology to analyze economic resource exchange in a Malawian village.

Our findings have important policy implications. We find evidence for departure from the standard random mixing assumption, and that a detailed, realistic, data-driven contact network model estimates different epidemic outcomes and different intervention impact than a model based on random mixing. We recommend further exploration of network structure on disease dynamics and further work to integrate network structure into current simulators.