Epidemic Spreading in an Animal Trade Network - Comparison of Distance-Based and Network-Based Control Measures

Network analysis of live pig movements in North Macedonia: Pathways for disease spread

Globalization of trade, and the interconnectivity of animal production systems, continues to challenge efforts to control disease. A better understanding of trade networks supports development of more effective strategies for mitigation for transboundary diseases like African swine fever (ASF), classical swine fever (CSF), and foot-and-mouth disease (FMD). North Macedonia, bordered to the north and east by countries with ongoing ASF outbreaks, recently reported its first incursion of ASF. This study aimed to describe the distribution of pigs and pig farms in North Macedonia, and to characterize the live pig movement network. Network analyses on movement data from 2017 to 2019 were performed for each year separately, and consistently described weakly connected components with a few primary hubs that most nodes shipped to. In 2019, the network demonstrated a marked decrease in betweenness and increase in communities. Most shipments occurred within 50 km, with movements <6 km being the most common (22.5%). Nodes with the highest indegree and outdegree were consistent across years, despite a large turnover among smallholder farms. Movements to slaughterhouses predominated (85.6%), with movements between farms (5.4%) and movements to market (5.8%) playing a lesser role. This description of North Macedonia's live pig movement network should enable implementation of more efficient and cost-effective mitigation efforts strategies in country, and inform targeted educational outreach, and provide data for future disease modeling, in the region.

Modelling African swine fever virus spread in pigs using time-respective network data: Scientific support for decision-makers

African Swine Fever (ASF) represents the main threat to swine production, with heavy economic consequences for both farmers and the food industry. The spread of the virus that causes ASF through Europe raises the issues of identifying transmission routes and assessing their relative contributions in order to provide insights to stakeholders for adapted surveillance and control measures. A simulation model was developed to assess ASF spread over the commercial swine network in France. The model was designed from raw movement data and actual farm characteristics. A metapopulation approach was used, with transmission processes at the herd level potentially leading to external spread to epidemiologically connected herds. Three transmission routes were considered: local transmission (e.g. fomites, material exchange), movement of animals from infected to susceptible sites, and transit of trucks without physical animal exchange. Surveillance was represented by prevalence and mortality detection thresholds at herd level, which triggered control measures through movement ban for detected herds and epidemiologically related herds. The time from infection to detection varied between 8 and 21 days, depending on the detection criteria, but was also dependent on the types of herds in which the infection was introduced. Movement restrictions effectively reduced the transmission between herds, but local transmission was nevertheless observed in higher proportions highlighting the need of global awareness of all actors of the swine industry to mitigate the risk of local spread. Raw movement data were directly used to build a dynamic network on a realistic time-scale. This approach allows for a rapid update of input data without any pre-treatment, which could be important in terms of responsiveness, should an introduction occur. This article is protected by copyright. All rights reserved.

Complex network analysis to understand trading partnership in French swine production

The circulation of livestock pathogens in the pig industry is strongly related to animal movements. Epidemiological models developed to understand the circulation of pathogens within the industry should include the probability of transmission via between-farm contacts. The pig industry presents a structured network in time and space, whose composition changes over time. Therefore, to improve the predictive capabilities of epidemiological models, it is important to identify the drivers of farmers’ choices in terms of trade partnerships. Combining complex network analysis approaches and exponential random graph models, this study aims to analyze patterns of the swine industry network and identify key factors responsible for between-farm contacts at the French scale. The analysis confirms the topological stability of the network over time while highlighting the important roles of companies, types of farm, farm sizes, outdoor housing systems and batch-rearing systems. Both approaches revealed to be complementary and very effective to understand the drivers of the network. Results of this study are promising for future developments of epidemiological models for livestock diseases. This study is part of the One Health European Joint Programme: BIOPIGEE.

Spatio-temporal network analysis of pig trade to inform the design of risk-based disease surveillance

Network analysis is a powerful tool to describe, estimate, and predict the role of pig trade in the spread of pathogens and generate essential patterns that can be used to understand, prevent, and mitigate possible outbreaks. This study aimed to describe the network between premises such as production herds, slaughterhouses, and traders of pig movements and identify heterogeneities in the connectivity of premises in the state of Santa Catarina, Brazil, using social network analysis (SNA). We used static and temporal network approaches to describe pig trade in the state by quantifying network attributes using SNA parameters, such as causal fidelity, loyalty, the proportion of node-loyalty, resilience of outgoing contact chains, and communities. Two indexes were implemented, the first one is a normalized index based on SNA-farm level measures and other index-based SNA-farm level measures considering the swineherd population size from all premises, both indexes were summarized by a municipality to target and rank surveillance activities. Within Santa Catarina, the southwest region played a key role in that 80 % of trade was concentrated in this region, and thus acted as a hub in the network. Besides, nine communities were found. The results also showed that premises were highly connected in the static network, with the network exhibiting low levels of fragmentation and loyalty. Also, just 11 % of the paths in the static network existed in the temporal network which accounted for the order in which edges occurred. Therefore, the use of time-respecting-paths was essential to not overestimate potential transmission pathways and outbreak sizes. Compared to static networks, the application of temporal network approaches was more suitable to capture the dynamics of pig trade and should be used to inform the design of riskbased disease surveillance.

Epidemic spreading in a weighted pig trade network

The inclusion of edge weights can add valuable insights in the spreading processes within trade networks and may identify factors influencing the final epidemic size. The aim of the study was to evaluate the effect of different network versions on the outcome of an epidemiological model. The weighted network versions included the number of trade contacts (A), the sum of delivered livestock (B) and the mean number of delivered livestock per trade contact (C). Furthermore, other factors, e.g. transmission probability and farm type of primary outbreak, were tested for their impact on the final epidemic size. From 2013-2014, data from a pig trade network in Northern Germany was recorded containing 678 farms connected by 1,018 directed edges. An epidemiological model was implemented considering a higher probability of disease spread for edges with a higher weight for each of the combinations between network version and transmission probability. Only transmission routes following the network structure were considered for disease transmission. The outcome of the epidemiological model (number of infected farms) was tested with a generalized linear mixed model including the fixed effects network version (unweighted, A, B, C), transmission probability and farm type of primary outbreak (breeding farm, farrowing farm, finishing farm, farrow-to-finishing farm, unknown) as well as all twofold interactions. The results revealed that all fixed effects as well as all twofold interactions were significant (p ≤ 0.05), i.e. in the following only the impact of the interactions on the number of infected farms can be interpreted. Network versions B and C showed in all combinations the highest number of infected farms independent of the underlying transmission probability. The unweighted network and network version A showed a significant increase of infected farms with increasing transmission probability. All interactions including the farm type of primary outbreak revealed a significant higher number of infected farms for farm types located at the beginning of the production chain, e.g. breeding farms. These farm types reached also more other farms in 1-4 steps compared to farm types located near to the end of the production chain. The inclusion of edge weights has a significant effect on the outcome of epidemiological models and dependent on the chosen edge weight the results need to be interpreted accordingly.

Illustration of Different Disease Transmission Routes in a Pig Trade Network by Monopartite and Bipartite Representation

Simple Summary

Besides direct animal movements between farms; indirect transmission routes of pathogens can have an immense impact on network structure and disease spread in animal trade networks. This study integrated these indirect transmission routes between farms via transport companies or feed supply as bipartite networks; which were compared to the monopartite animal movements network representing the direct transmission route. Both bipartite networks were projected on farm level to enable a comparison to the monopartite network. The number of edges increased immensely from the monopartite animal movements network to both projected networks. Thus, farms can be highly connected over indirect connections, although they are not directly trading animals. The ranking of the animals according to their centrality parameters, indicating their importance for the network, showed moderate correlations only between the animal movements and the transportation network. The epidemiological models based on the different network representations revealed significantly more infected farms for the networks including indirect transmission routes compared to the direct animal movements. Indirect transmission routes had an immense impact on the outcome of centrality parameters, as well as on the spreading process within the network. This knowledge is needed to understand disease spread and to establish reliable prevention and control measurements.

Abstract

Besides the direct transport of animals, also indirect transmission routes, e.g., contact via contaminated vehicles, have to be considered. In this study, the transmission routes of a German pig trade network were illustrated as a monopartite animal movements network and two bipartite networks including information of the transport company and the feed producer which were projected on farm level (n = 866) to enable a comparison. The networks were investigated with the help of network analysis and formed the basis for epidemiological models to evaluate the impact of different transmission routes on network structure as well as on potential epidemic sizes. The number of edges increased immensely from the monopartite animal movements network to both projected networks. The median centrality parameters revealed clear differences between the three representations. Furthermore, moderate correlation coefficients ranging from 0.55 to 0.68 between the centrality values of the animal movements network and the projected transportation network were obtained. The epidemiological models revealed significantly more infected farms for both projected networks (70% to 100%) compared to the animal movements network (1%). The inclusion of indirect transmission routes had an immense impact on the outcome of centrality parameters as well as on the results of the epidemiological models.

Burkholderia mallei: The dynamics of networks and disease transmission

Glanders is a highly infectious zoonotic disease caused by Burkholderia mallei. The transmission of B. mallei occurs mainly by direct contact, and horses are the natural reservoir. Therefore, the identification of infection sources within horse populations and animal movements is critical to enhance disease control. Here, we analysed the dynamics of horse movements from 2014 to 2016 using network analysis in order to understand the flow of animals in two hierarchical levels, municipalities and farms. The municipality-level network was used to investigate both community clustering and the balance between the municipality's trades and the farm-level network associations between B. mallei outbreaks and the network centrality measurements, analysed by spatio-temporal generalized additive model (GAM). Causal paths were established for the dispersion of B. mallei outbreaks through the network. Our approach captured and established a direct relationship between movement of infected equines and predicted B. mallei outbreaks. The GAM model revealed that the parameters in degree and closeness centrality out were positively associated with B. mallei. In addition, we also detected 10 communities with high commerce among municipalities. The role of each municipality within the network was detailed, and significant changes in the structures of the network were detected over the course of 3 years. The results suggested the necessity to focus on structural changes of the networks over time to better control glanders disease. The identification of farms with a putative risk of B. mallei infection using the horse movement network provided a direct opportunity for disease control through active surveillance, thus minimizing economic losses and risks for human cases of B. mallei.

Spatio-temporal modelling of verotoxigenic Escherichia coli O157 in cattle in Sweden: exploring options for control

A spatial data-driven stochastic model was developed to explore the spread of verotoxigenic Escherichia coli O157 (VTEC O157) by livestock movements and local transmission among neighbouring holdings in the complete Swedish cattle population. Livestock data were incorporated to model the time-varying contact network between holdings and population demographics. Furthermore, meteorological data with the average temperature at the geographical location of each holding was used to incorporate season. The model was fitted against observed data and extensive numerical experiments were conducted to investigate the model’s response to control strategies aimed at reducing shedding and susceptibility, as well as interventions informed by network measures. The results showed that including local spread and season improved agreement with prevalence studies. Also, control strategies aimed at reducing the average shedding rate were more efficient in reducing the VTEC O157 prevalence than strategies based on network measures. The methodology presented in this study could provide a basis for developing disease surveillance on regional and national scales, where observed data are combined with readily available high-resolution data in simulations to get an overview of potential disease spread in unobserved regions.

Using Machine Learning to Predict Swine Movements within a Regional Program to Improve Control of Infectious Diseases in the US

Between-farm animal movement is one of the most important factors influencing the spread of infectious diseases in food animals, including in the US swine industry. Understanding the structural network of contacts in a food animal industry is prerequisite to planning for efficient production strategies and for effective disease control measures. Unfortunately, data regarding between-farm animal movements in the US are not systematically collected and thus, such information is often unavailable. In this paper, we develop a procedure to replicate the structure of a network, making use of partial data available, and subsequently use the model developed to predict animal movements among sites in 34 Minnesota counties. First, we summarized two networks of swine producing facilities in Minnesota, then we used a machine learning technique referred to as random forest, an ensemble of independent classification trees, to estimate the probability of pig movements between farms and/or markets sites located in two counties in Minnesota. The model was calibrated and tested by comparing predicted data and observed data in those two counties for which data were available. Finally, the model was used to predict animal movements in sites located across 34 Minnesota counties. Variables that were important in predicting pig movements included between-site distance, ownership, and production type of the sending and receiving farms and/or markets. Using a weighted-kernel approach to describe spatial variation in the centrality measures of the predicted network, we showed that the south-central region of the study area exhibited high aggregation of predicted pig movements. Our results show an overlap with the distribution of outbreaks of porcine reproductive and respiratory syndrome, which is believed to be transmitted, at least in part, though animal movements. While the correspondence of movements and disease is not a causal test, it suggests that the predicted network may approximate actual movements. Accordingly, the predictions provided here might help to design and implement control strategies in the region. Additionally, the methodology here may be used to estimate contact networks for other livestock systems when only incomplete information regarding animal movements is available.

Data-driven network modelling of disease transmission using complete population movement data: spread of VTEC O157 in Swedish cattle

European Union legislation requires member states to keep national databases of all bovine animals. This allows for disease spread models that includes the time-varying contact network and population demographic. However, performing data-driven simulations with a high degree of detail are computationally challenging. We have developed an efficient and flexible discrete-event simulator SimInf for stochastic disease spread modelling that divides work among multiple processors to accelerate the computations. The model integrates disease dynamics as continuous-time Markov chains and livestock data as events. In this study, all Swedish livestock data (births, movements and slaughter) from July 1st 2005 to December 31st 2013 were included in the simulations. Verotoxigenic Escherichia coli O157:H7 (VTEC O157) are capable of causing serious illness in humans. Cattle are considered to be the main reservoir of the bacteria. A better understanding of the epidemiology in the cattle population is necessary to be able to design and deploy targeted measures to reduce the VTEC O157 prevalence and, subsequently, human exposure. To explore the spread of VTEC O157 in the entire Swedish cattle population during the period under study, a within- and between-herd disease spread model was used. Real livestock data was incorporated to model demographics of the population. Cattle were moved between herds according to real movement data. The results showed that the spatial pattern in prevalence may be due to regional differences in livestock movements. However, the movements, births and slaughter of cattle could not explain the temporal pattern of VTEC O157 prevalence in cattle, despite their inherently distinct seasonality.