Investigating the impact of edge weight selection on the pig trade network topology

Abstract

AbstractTraceability of animal movements and robust surveillance are crucial for identifying and controlling animal diseases. Risk-based surveillance, i.e. network-based approaches, can identify higher-risk holdings or trades. However, node ranking, useful to identify ”influential” nodes (holdings) in the network, varies with the considered metrics.We use a dataset of pig movements in Upper Austria from 2021 to study the robustness of node ranking through three centrality metrics and compare them with epidemic model ranking. Incorporating edge weights may influence the network analysis, therefore, we simulate two representations using edge weights based on: i) the frequency of exchanges between holdings (”frequency-based”) and ii) the number of pigs exchanged (”volume-based”). We compare the impact of the edge weight on the network topology, community structure, and node ranking in a network with 5,766 nodes and 92,914 edges. Results revealed distinct edge weight distributions: frequency-based network exhibited a bimodal pattern, while volume-based was more uniform. Strength centrality exhibited the highest correlation with simulation-based rankings, particularly for the top 5% highest-ranked nodes (τb= 0.51 for frequency-based andτb= 0.5 for volume-based). These findings highlight that using strength centrality to identify critical nodes can significantly enhance surveillance strategies, making them more efficient and field-deployable, enhancing traditional methods without requiring extensive simulations.Author summaryEarly detection of infectious diseases through surveillance activities is important to prevent severe impacts on the livestock sector and avoid major economic losses that often result from delayed detection. Prioritizing surveillance efforts through a data-driven approach presents a strategic advantage compared to random sampling. Here, we propose two network representations of the Upper Austrian pig trade network 2021, which show livestock holdings as nodes and animal movements as links between these nodes, which can represent the volume of pigs exchanged or the frequency of these exchanges. Using network analysis methods, we identify ”influential” holdings in the network, i.e., those playing a key role in the network, either due to their numerous connections with other holdings, their position on many trade paths, or their crucial role in disease transmission dynamics. We show that using strength centrality can effectively identify key holdings for targeted surveillance. Adopting network-based surveillance can facilitate resource allocation for veterinary surveillance programs, offering a cost-effective strategy for disease management, enabling tailored, more effective, and timely interventions.