Accessing the temporal pig trade network in a Brazilian state to inform the design of risk-based disease surveillance
-
1
University of São Paulo, Brazil
-
2
University of Minnesota Twin Cities, United States
Introduction and background
In Brazil the swineherd is approximately 2 million animals, producing 3.75 million tons in 2017 of meat, this has made Brazil the fourth largest producer of pork in the world. Infectious diseases in farm animals cause great economic loss, close trade borders and compromise animal welfare, reduces productivity and induces big costs for disease control and eradication.
Thus, the control of infectious diseases is critical in order to maintain this leadership; however, the trade may also play an important role in the spreading of infectious diseases from a particular region to other geographically distant places.
Many works use static representations of the network considering all the possible connections over time within the network, but not always the contacts between different nodes of the network are active over time, and this representation increases the number of connections.
The use of temporal networks makes possible to consider the active and inactive connections according to a specific time interval. Therefore, the aim of this study is to describe the temporal data of the swine trade providing useful information for rank municipalities and properties in the state of Santa Catarina, Brazil, based on different parameters of social network analysis (SNA). This data helps to rank areas in order to improve the prevention, control, and eradication of communicable diseases by the contact network.
Material and methods
Data
Records of animal movements from Santa Catarina (SC) state in Brazil were available by the official veterinary service in this state. The database comprises livestock movements of pigs in SC State in Brazil since 2015-01-01 and 2017-12-31. For the Network analysis, the movements for the purpose of slaughter were excluded from the analysis, considering these as the end of the productive chain, records for which information on location, in which origin, destination, and/or date were not available, and movement to other states were removed prior to the analysis.
Static representation of the pig trade
We constructed contact networks, in which units were defined as nodes (vertices) and movements between units were considered edges. We described the static network of all period of study and by year using the parameters: Number of nodes, number of edges, degree, closeness centrality, betweenness. A spatial representation by municipalities was built using different node-level metrics.
Description of network analysis terminology and metrics are show in the supplementary table 1.
Temporal network
We considered a temporal network g = (V,E,T) as a set of nodes V and a set of edges E in a given observation period T. Each edge in E is given by a triple(i,j,T) and connects node i and node j at time T. To describe the temporal patterns over time we used a monthly time‐aggregate window, i.e. g= g_(1,)...,g_(T ), where T is the observation period and the increment is the temporal resolution.
Network vulnerability and resilience
The most common skills to contain an epidemic outbreak include: culling, isolation of holdings, etc., and therefore target holdings with specific centrality measures to apply this intervention measures can be used. For this purpose, nodes are first of all ranked according to their centrality. Then the impact in the giant strongly connected component (GSCC) and in the giant weekly connected component (GWCC) by the removal of nodes with the highest rank on the network can be measured. These components represent in theory the maximum epidemic outbreak size assuming 100% of rate infection through the networks. After each node removal, the GSCC and GWCC have to be computed again (Motta et al., 2017). The results are showed in the Supplementary Figure 1.
Causal fidelity
We use the “Causal Fidelity” (CF), which measures the fraction of the number of paths in a static network which can also be taken in the temporal counterpart. The causal fidelity ranges between 0 ≤ CF≤1(X), where large values indicate that the static aggregation of a temporal network gives a good approximation from a causal point of view. We calculate this due to some analysis being based on a static representation of networks (Lentz, Selhorst, & Sokolov, 2013).
Modeling outgoing temporal movements
The outgoing contact chain for each holding was calculated for a weekly aggregate time series network using the entire period of study following the chronological order of the movements. For each contact chain analysis calculated, the 10, 250, 500, 700 and 1000 nodes with the best rank based on the network metrics (all degree, clustering coefficient, Page rank, Cluster coefficient, and Closeness) were removed and the contact chain was recalculated.
Results
In this swine trade network, 11,606 premises and 6,0645,853 pigs were involved in 320,840 movements. The movements to slaughterhouses were excluded from the analysis, considering these as the end of the productive chain.
The static representation of the pig trade network is shown in the supplementary table 2.
We represent the temporal trends over time plotting the network metrics using a monthly time-window, the results are shown in the supplementary figure 1.
To assess the network’s vulnerability to the targeted removal of nodes, we measured the impact in the GWCC and GSCC after the removal of several nodes, in decreasing order of a given centrality measure. The results are shown in the supplementary figure 2.
Figure 2. Defragmentation of GSCC and GWCC, the y axis represents the size of the components and the x axis represent the number o of nodes removed by the best rank from network measures.
The study of the network vulnerability and resilience is a useful tool but, since it is based on the GWCC and GSCC components from a static network, this approach can create many paths that never exist if we consider the emergence of new edges over time. For that reason, we calculated the causal fidelity, and we obtained a result of 0.90 for the whole observation period. Thus, about 90 % of the time-respecting paths exist in both network representations.
So, our alternative to increase the accuracy of results was to calculate the outgoing contact chain considering time-respecting paths. In the Supplementary figure 3 we show the size of the contact chains over time taking into account the removal of specific nodes over time. The all degree and betweenness were the most effective metric to decrease the outgoing contac chain. This approach reveals the impact that the removal of an important node in the network has on a certain network metric.
Supplementary figure 3
We used the best network metrics based on the previously analysis and normalize them from zero to one, in order to integrate these metrics in a single measurement index and we represent it spatially in order to get the most important areas for a risk-based disease surveillance. The results are shown in the Supplementary figure 4.
Acknowledgements
This study was financed in part by CAPES (Finance Code 001)
References
Lentz, H. H. K., Selhorst, T., & Sokolov, I. M. (2013). Unfolding Accessibility Provides a Macroscopic Approach to Temporal Networks. Physical Review Letters, 110(11), 118701. https://doi.org/10.1103/PhysRevLett.110.118701
Motta, P., Porphyre, T., Handel, I., Hamman, S. M., Ngu Ngwa, V., Tanya, V., … Bronsvoort, B. M. deC. (2017).
Implications of the cattle trade network in Cameroon for regional disease prevention and control. Scientific Reports, 7(1), 43932. https://doi.org/10.1038/srep43932
Keywords:
Social network analyses (SNA),
swine trade,
Risk based surveillance,
Spatio temporal analysis,
GIS analyse
Conference:
GeoVet 2019. Novel spatio-temporal approaches in the era of Big Data, Davis, United States, 8 Oct - 10 Oct, 2019.
Presentation Type:
Student oral presentation
Topic:
Spatial-explicit or spatio-temporal network analysis
Citation:
Cespedes Cardenas
N,
Ardila Galvis
JO,
Lima
DM,
VanderWaal
K and
Grisi Filho
JH
(2019). Accessing the temporal pig trade network in a Brazilian state to inform the design of risk-based disease surveillance.
Front. Vet. Sci.
Conference Abstract:
GeoVet 2019. Novel spatio-temporal approaches in the era of Big Data.
doi: 10.3389/conf.fvets.2019.05.00066
Copyright:
The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers.
They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.
The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.
Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.
For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.
Received:
31 May 2019;
Published Online:
27 Sep 2019.
*
Correspondence:
Mx. Nicolas Cespedes Cardenas, University of São Paulo, São Paulo, Brazil, ncesped@ncsu.edu