Fig 1.
Diagram of potential transmission relationships and connectivity of birds in the network.
The figure represents three example birds, assessed for the potential for each to have acquired infection from the other. Each bird, or “subject”, was defined to have an incubation period, initially set to the period occurring six to 24 months before the bird’s final date in the study. Any other bird that shared an enclosure with the subject during its incubation period was defined as a “friend” if the two birds shared the space during the second bird’s infectious period. A friend’s infectious period was initially set to the period occurring two years prior to its final date in the study. Thus, the figure shows the incubation and infectious periods for each bird in the larger bars while the smaller bars show the overlapping period when the other two birds would be defined as its infectious”friends”. The network edges were created from identifying the spatial and temporal overlap of potential incubation and infectious periods of subjects and friends in the study population.
Fig 2.
Illustration of network assembly and evaluated relationships.
Each circle represents a different bird, showing a subject (black), their friends (gray), and their friends of friends (white). Colors represent pathways along which the different evaluated relationships were formed. Lines represent the network edges which connect birds that shared enclosures. (A) Illustration of network construction. Friends of friends differed by the timing of contact (those that could vs. could not transmit infection to the subject) and location of contact (same enclosure vs. different enclosure) with respect to their subject. (B) The subject’s assembled network of friends, friends of friends, and evaluated relationships. Friends: Clustering of disease associated with direct contacts. Friends-of-Friends: All influential. Clustering of disease associated with all indirect contacts that could influence the subject’s disease status (based on timing of contact). Friends-of-Friends: Same environment. Clustering of disease associated with influential indirect contacts that were exposed to the same enclosure/environment. Friends-of-Friends: Contagion. Clustering of disease associated with influential indirect contacts that were never exposed to the same environment. This evaluation is key for removing the confounding effects of the environment and testing for a contagious process. Friends-of-friends: Homophily. Clustering of disease associated with friends of friends that were never exposed to the same environment and could not have transmitted disease to the subject based on the timing of the connection. This reverse-time placebo test evaluates our data for the presence of homophily, or whether disease clustering can be explained by similarities among connected birds.
Fig 3.
Social network graph of a subset the San Diego Zoo Global bird network, 1992–2014.
The subset of the network illustrates all positive study birds (“subjects”) and their direct contacts (“friends”). Each node represents one bird in the data set and connections between birds were defined by enclosure sharing. There are 3417 birds represented in this subset with 6066 unique connections between them. The color of the circle indicates each bird’s disease status: red denotes a bird with mycobacteriosis and light blue denotes a bird that did not have disease. Statistical tests for clustering of disease on the network showed significant increases in disease risk for subjects directly or indirectly connected to an infected friend, compared to an uninfected friend.
Fig 4.
Estimates of disease clustering between friends in the bird transmission network, San Diego Zoo Global (n = 16,430).
The estimated relative risk (RR) for each of five different relationships between subjects and friends that were directly and indirectly connected. Evaluated relationships are described in the Methods and Fig 2B. Significance of the estimate was determined by comparing conditional probability of mycobacteriosis in the observed network with 1000 permutations of an identical network (with the topology and incidence of mycobacteriosis preserved) in which the same number of infected birds were randomly distributed. Error bars show the null 95% confidence intervals generated from the random permutations. RRs that were outside of the null and significant are indicated with *.
Table 1.
Results of Friends-of-friends network analysis and sensitivity analyses for birds housed at San Diego Zoo Global, 1992–2014 (n = 16,430).