Networks provide an intuitive framework to characterize the drivers of animal social behavior or predict its ecological and evolutionary consequences. The spread of infectious disease represents one important and well-studied consequence of network structure, and results from the theory and human disease literature have highlighted the potential importance of network properties in driving disease dynamics. However, there has been relatively little general quantification of the structure of wildlife contact networks, and it remains unclear to what extent the insights from theory apply to wildlife systems. To address this gap, we review 67 empirical networks from 45 studies and characterize their size (the number of individuals), density (the average number of contacts per individual), contact heterogeneity (the variance in contacts among individuals), and transitivity (the probability that an individual's contacts are also contacts). We then generate a large collection of random networks with properties encompassing those observed in the literature and use them to simulate the stochastic spread of an immunizing infection (i.e., and SIR model). We summarize each of these simulated outbreaks and evaluate the extent to which they are controlled by the properties of the contact network.
Results/Conclusions
Our review shows that most published wildlife contact networks are relatively small (tens to hundreds of individuals), dense, and have low variation among animals in numbers of contacts. Simulating disease spread on networks with 100 individuals, we find that the size of an outbreak grows as network density increases, while increasing contact heterogeneity and transitivity generally dampens disease spread, consistent with the predictions from theory. However, we also find that small size of wildlife networks imposes additional constraints on both the network properties that can be generated and on the outbreaks that can occur. In particular, increasing network density quickly washes out the effects of contact heterogeneity and transitivity, while high contact heterogeneity strongly limits the range of possible transitivities. Thus, in wildlife systems characterized by either dense or very heterogeneous contacts, detailed information about contact structure is not likely to be informative for predicting disease spread. In systems that have more sparse contacts, or that have moderate contact heterogeneity, the effect of transitivity on disease spread can be large and thus efforts to collect detailed contact data may be necessary.