Longitudinal social network analysis of avian mycobacteriosis incidence in a large population of zoo birds.

The goal of this study was to evaluate longitudinal patterns of avian mycobacteriosis spread through a social network. Specifically, we wanted to determine whether the patterns of connectivity over time can predict future infections, and whether this pattern can distinguish between different sources of infection. The study population included 13,409 individuals nested in a larger population of birds that were closely monitored in zoological facilities for over 22 years (1992-2014). A retrospective cohort study design and social network connectivity were used to estimate the association between exposure to an infected bird, and development of mycobacteriosis. Avian mycobacteriosis was diagnosed from histopathology and network connectivity was defined by enclosure histories over discrete time periods. Single-variable and multivariable longitudinal, mixed effects logistic regression models examined whether exposure to infected birds, both directly- and indirectly-connected, was associated with development of mycobacteriosis at the next time step. Our adjusted model showed an increased odds of developing mycobacteriosis (odds ratio = 2.15; 95 % CI: 1.48-3.12; p < 0.001) for birds that were directly exposed (i.e., housed in the same aviary) to another infected bird, compared to those with no exposure. Exposure to a positive, indirectly-connected bird at a previous time step was independently associated with an increased risk of mycobacteriosis (odds ratio = 1.56; 95 % CI: 1.11-2.19). This association persisted in adjusted models even when the indirect contacts were housed in distinctly different aviaries and never had contact with the subject of interest or its environment. Adjusted, risk-stratified models further characterized the type of exposure that increased the risk of avian mycobacteriosis. Birds that were exposed in small aviaries were more likely to develop mycobacteriosis than those exposed in larger aviaries and those with no exposure. The lesion distribution and species of the contact (same species versus different species) were also significant predictors of disease risk. Some findings were sensitive to model variation of time divisions and initiation time. Our study shows avian mycobacteriosis spread through the social network in quantifiable and discernable patterns. We provide empirical evidence that a contagious process drives some of the observed infection, but we also show low transmissibility based on sustained patterns of low incidence over time even when large groups of birds are exposed. Targeted risk mitigation efforts based on the characteristics of the exposure may be effective at reducing risk of avian mycobacteriosis while enhancing population sustainability.