Live fish movement is considered as having an important role in the transmission of infectious diseases. For that reason, interventions for cost-effective disease prevention and control rely on a sound understanding of the patterns of live fish movements in a region or country. Here, we characterize the network of live fish movements in the Irish salmonid farming industry during 2013, using social network analysis and spatial epidemiology methods, and identify interventions to limit the risk of disease introduction and spread. In the network there were 62 sites sending and/or receiving fish, with a total of 130 shipments (84 arcs) comprising approx. 17.2 million fish during the year. Atlantic salmon shipments covered longer distances than trout shipments, with some traversing the entire country. The average shipment of Atlantic salmon was 146,186 (SD 194,344) fish, compared to 77,928 (127,009) for trout, however, variability was high. There were 3 periods where shipments peaked (February-April, June-September, and November), which were related to specific stages of fish. The network was disconnected and had two major weak components, the first one with 39 nodes (mostly Atlantic salmon sites), and the second one with 10 nodes (exclusively trout sites). Correlation between in and out-degree at each site and assortativity coefficient were slightly low and non-significant: -0.08 (95% CI: -0.22, 0.06) and -0.13 (95% CI: -0.36, 0.08), respectively, indicating random mixing with regard to node degree. Although competing models also produced a good fit to degree distribution, it is likely that the network possesses both small-world and scale-free topology. This would facilitate the spread and persistence of infection in the salmon production system, but would also facilitate the design of risk-based surveillance strategies by targeting hubs, bridges or cut-points. Using Infomap community detection algorithms, 2 major communities were identified within the giant weak component, which were linked by only 4 nodes. Communities found had no correspondence with geographical zones within the country, which could potentially hinder the implementation of zoning strategies for disease control and eradication. Three significant spatial clusters of node centrality measures were detected, two in county Donegal (betweenness and outcloseness) and one in county Galway (incloseness), highlighting the importance of these locations as hot spots of highly central sites with a higher potential for both introduction and spread of infection. These results will assist in the design and implementation of measures to reduce the sanitary risks emerging from live fish trade within Ireland.
Keywords: Fish diseases; Infectious disease epidemiology; Ireland; Risk-based surveillance; Social network analysis; Spatial scan statistic.
Copyright © 2015 Elsevier B.V. All rights reserved.