Network transmission inference: host behavior and parasite life cycle make social networks meaningful in disease ecology.

The process of disease transmission is determined by the interaction of host susceptibility and exposure to parasite infectious stages. Host behavior is an important determinant of the likelihood of exposure to infectious stages but is difficult to measure and often assumed to be homogenous in models of disease spread. We evaluated the importance of precisely defining host contact when using networks that estimate exposure and predict infection prevalence in a replicated, empirical system. In particular, we hypothesized that infection patterns would be predicted only by a contact network that is defined according to host behavior and parasite life cycle. Two competing host contact criteria were used to construct networks defined by parasite life cycle and social contacts. First, parasite-defined contacts were based on shared space with a time delay corresponding to the environmental development time of nematode parasites with a direct fecal-oral life cycle. Second, social contacts were defined by shared space in the same time period. To quantify the competing networks of exposure and infection, we sampled natural populations of the eastern chipmunk (Tamias striatus) and infection of their gastrointestinal helminth community using replicated longitudinal capture-mark-recapture techniques. We predicted that (1) infection with parasites with direct fecal-oral life cycles would be explained by the time delay contact network, but not the social contact network; (2) infection with parasites with trophic life cycles (via a mobile intermediate host; thus, spatially decoupling transmission from host contact) would not be explained by either contact network. The prevalence of fecal-oral life cycle nematode parasites was strongly correlated to the number and strength of network connections from the parasite-defined network (including the time delay), while the prevalence of trophic life cycle parasites was not correlated with any network metrics. We concluded that incorporating the parasite life cycle, relative to the way that exposure is measured, is key to inferring transmission and can be empirically quantified using network techniques. In addition, appropriately defining and measuring contacts according the life history of the parasite and relevant behaviors of the host is a crucial step in applying network analyses to empirical systems.