TYPE Original Research PUBLISHED 14 July 2023 DOI 10.3389/fvets.2023.1201644 Disentangling transport OPEN ACCESS movement patterns of trucks either transporting pigs or while EDITED BY Ioannis Magouras, University of Bern, Switzerland REVIEWED BY Helen Roberts, empty within a swine production Department for Environment, Food and Rural Affairs, United Kingdom Shankar Yadav, system before and during the European Commission for the Control of Foot and Mouth Disease (EuFMD), Italy COVID-19 epidemic *CORRESPONDENCE Catalina Picasso-Risso picasso.2@osu.edu Catalina Picasso-Risso1,2,3*, Carles Vilalta1,4,5 , Cesar A. Corzo corzo@umn.edu Juan Manuel Sanhueza1,6 , Mariana Kikuti1 , Mark Schwartz1 and RECEIVED 06 April 2023 Cesar A. Corzo1* ACCEPTED 19 June 2023 1 PUBLISHED 14 July 2023 Department of Veterinary Population Medicine, University of Minnesota, Saint Paul, MN, United States, 2 Facultad de Veterinaria, Universidad de la Republica, Montevideo, Uruguay, 3 Department of Veterinary CITATION Preventive Medicine, College of Veterinary Medicine, The Ohio State University, Columbus, OH, Picasso-Risso C, Vilalta C, Sanhueza JM, United States, 4 Unitat mixta d’Investigació IRTA-UAB en Sanitat Animal, Centre de Recerca en Sanitat Kikuti M, Schwartz M and Corzo CA (2023) Animal, Campus de la Universitat Autònoma de Barcelona, Bellaterra, Spain, 5 IRTA, Programa de Sanitat Disentangling transport movement patterns of Animal, Centre de Recerca en Sanitat Animal, Campus de la Universitat Autònoma de Barcelona, trucks either transporting pigs or while empty Bellaterra, Spain, 6 Departamento de Ciencias Veterinarias y Salud Publica, Facultad de Recursos within a swine production system before and Naturales, Universidad Católica de Temuco, Temuco, Chile during the COVID-19 epidemic. Front. Vet. Sci. 10:1201644. doi: 10.3389/fvets.2023.1201644 Transport of pigs between sites occurs frequently as part of genetic improvement COPYRIGHT © 2023 Picasso-Risso, Vilalta, Sanhueza, Kikuti, and age segregation. However, a lack of transport biosecurity could have Schwartz and Corzo. This is an open-access catastrophic implications if not managed properly as disease spread would be article distributed under the terms of the imminent. However, there is a lack of a comprehensive study of vehicle movement Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other trends within swine systems in the Midwest. In this study, we aimed to describe forums is permitted, provided the original and characterize vehicle movement patterns within one large Midwest swine author(s) and the copyright owner(s) are system representative of modern pig production to understand movement trends credited and that the original publication in this journal is cited, in accordance with accepted and proxies for biosecurity compliance and identify potential risky behaviors that academic practice. No use, distribution or may result in a higher risk for infectious disease spread. Geolocation tracking reproduction is permitted which does not devices recorded vehicle movements of a subset of trucks and trailers from comply with these terms. a production system every 5 min and every time tracks entered a landmark between January 2019 and December 2020, before and during the COVID-19 pandemic. We described 6,213 transport records from 12 vehicles controlled by the company. In total, 114 predefined landmarks were included during the study period, representing 5 categories of farms and truck wash facilities. The results showed that trucks completed the majority (76.4%, 2,111/2,762) of the recorded movements. The seasonal distribution of incoming movements was similar across years (P > 0.05), while the 2019 winter and summer seasons showed higher incoming movements to sow farms than any other season, year, or production type (P < 0.05). More than half of the in-movements recorded occurred within the triad of sow farms, wean-to-market stage, and truck wash facilities. Overall, time spent at each landmark was 9.08% higher in 2020 than in 2019, without seasonal highlights, but with a notably higher time spent at truck wash facilities than any other type of landmark. Network analyses showed high connectivity among farms with identifiable clusters in the network. Furthermore, we observed a decrease in connectivity in 2020 compared with 2019, as indicated by the majority of network Frontiers in Veterinary Science 01 frontiersin.org Picasso-Risso et al. 10.3389/fvets.2023.1201644 parameter values. Further network analysis will be needed to understand its impact on disease spread and control. However, the description and quantification of movement trends reported in this study provide findings that might be the basis for targeting infectious disease surveillance and control. KEYWORDS transport, pork industry, network analysis, disease surveillance, Midwest, USA Introduction assumed zero risk for disease transmission when vehicles visit a site after stopping in a cleaning station (i.e., truck wash station) (4, 6). Ground transportation of pigs between sites within a Here, as a first step to understanding the role of trucks production system occurs frequently for age category and transporting pigs or empty-unwashed trucks in the spread of production phase segregation because it has been shown to increase diseases, we described and analyzed for the first time the vehicle production efficiency and animal health. Although trucks and movement patterns and network structure of one of the largest trailers are designed to load batches of animals with protocols multi-site pig production systems in the Midwestern US using that safeguard animals’ wellbeing, health, and safety, it is well- GPS trackers and including vehicles in the network, regardless established that animal movements are a source of transmission of the load status or origin/destination of the movement. We of infectious disease within the swine industry (1–6), and vehicles assessed the utility of GPS trackers as a reliable source to record which transported animals need to be correctly cleaned and truck movements, characterized vehicle network before and during disinfected before loading the next batch of pigs to prevent the risk the COVID-19 pandemic, fidelity (defined as the consistency of of pathogen spread (7, 8). vehicle movements over time), and time spent at each site, with the With the intent to mitigate the risk of potential disease ultimate goal of understanding movement trends that will help to spread among pig farms, biosecurity practices such as isolation identify potential risky behaviors and spatio-temporal variabilities of incoming breeding stock, testing before commingling animals, that can inform epidemic-preparedness and support decisions to shower-in/shower-out, disinfection, and drying rooms for supply improve biosecurity practices and compliance for reducing the entry are recommended among others (9). In addition, cleaning spread of diseases compromising swine health. and disinfection of trucks and trailers transporting pigs between loads is an area of focus especially when vehicles transport gilts, boars, recently weaned pigs, and culled animals. Although all these Materials and methods biosecurity measures are useful, they are not completely effective as these depend on a combination of standardized processes and their Population and data source consistent implementation (9). As an example, it has been described that washing and disinfecting vehicles without the appropriate Data were collected from a pig-producing company drying has a similar risk for disease transmission as unwashed participating in the University of Minnesota Morrison Swine vehicles (7, 8). Furthermore, a recent study indicated that PRRSv Health Monitoring Project (MSHMP) (17) that voluntarily agreed transmission associated with movements of trucks used for feed to participate in the study. The selected company is a typical and personnel transportation is not negligible (4), adding a layer multi-site system representative of modern pig production systems of complexity when understanding disease spread and designing in the US This company accounts for approximately 1% of the risk-mitigation (e.g., biosecurity) practices for the swine industry. swine production in the country. The system agreed to have a Social network analysis has proven to be a useful tool for selected subset of 12 out of 128 owned pig transportation vehicles understanding infectious disease dissemination among populations monitored through geolocation-tracking devices installed on seven by analyzing pig movement patterns among animal production trucks transporting only recently weaned pigs and five trailers that systems (10–14). Disease surveillance and control have been transport either gilts or culled sows. On the trailers, the Verizon traditionally assessed by tracing animal movements as a source of Networkfleet Asset Guard was attached through screws, whereas disease spread since infected animals are the most frequent way the Verizon Networkfleet 5000 Series Model 5500N4VL was to spread pathogens to susceptible populations. Furthermore, the connected to the onboard diagnostics (OBD) port on the tractor potential impact of pathogen dissemination through animals in cab. These devices (Figure 1) are designed to endure long periods the US has been assessed in silico using retrospective shipment of outdoor or other rugged environments while maintaining records (4) that can introduce bias in the analysis, when dealing communication and GPS signal during the journey and through with missing or incomplete data. Limited studies have focused vehicle washing. on understanding the potential roles that the vehicles used for Vehicle movement monitoring occurred between January feed transport and the personnel involved may have on pathogen 2019 and December 2020. A total of 6,213 records of vehicles dissemination (4, 15, 16), but in fact, the majority of studies have entering 114 georeferenced pig sites were initially considered overlooked the role that trucks may have on pathogen spread or for this study over the study period (3,659 and 2,554 in Frontiers in Veterinary Science 02 frontiersin.org Picasso-Risso et al. 10.3389/fvets.2023.1201644 FIGURE 1 Geolocation tracking devices installed in the vehicles (A, B). Full description of the devices is available online at: https://www.verizon.com/about/ news/verizon-expands-asset-tracking-portfolio-introduction-networkfleet-asset-guard-0. 2019 and 2020, respectively). Connections among sites (i.e., Network nomenclature and metrics vehicle movements) were recreated based on the time and date definition of entry to each of the geofences and the trajectory of the vehicles tracked. A yearly static directed non-weighted network was constructed for the complete period (2019–2020). Network nodes or vertices (i.e., elements or units of the network) (19) were polygons created within the extension of preexisting system site locations (i.e., Spatiotemporal characterization of the landmarks), which included the abovementioned five different movements categories of farms and truck wash facilities. The movement of vehicles was captured when tracking devices located in the vehicles We described the frequency of incoming movement patterns that entered or exited each predefined polygon, and thus, the time yearly, seasonally, and weekly over the study period by source farm spent within the polygon was recorded for each vehicle and node. (e.g., farm production type) and time spent by the vehicles at each Edges were created by linking nodes sequentially visited by site. Farm production type was categorized as either sow, nursery, each vehicle. We selected those nodes in which a stop was finisher, wean-to-market, or gilt development unit. Differences recorded for more than 5 min as these were considered an in the frequency of movements were assessed with univariable “operational stop”, such as loading/unloading animals or washing and multivariable negative binomial regression which allowed the vehicle. When vehicles showed a recorded stop in a node adjusting for over-dispersed data (e.g., frequency of incoming for <5 min, this was considered a drive-by or an incorrect movements mean = 18.71, variance = 274.47) (18) and adjusted for record (i.e., a result of rearranging vehicles for loading/unloading multiple comparisons. We described the proportion of movement animals or inaccurately captured by the geofences surrounding performed between pairs of sites grouped by production category the nodes-polygons when the vehicle was circulating close to (e.g., sow to nursery, sow to sow) and vehicle fidelity over time. the polygon boundaries). Furthermore, movements recorded by Incoming movement density was mapped by year to identify areas vehicles entering sequentially onto the same node-polygon for a of major movement activities. period of <15 min within an hour and without visiting any other Frontiers in Veterinary Science 03 frontiersin.org Picasso-Risso et al. 10.3389/fvets.2023.1201644 FIGURE 2 Weekly in-movement distribution by farm categories (proportion) and in total (%) for 2019 (left) and 2020 (right). node in between were merged and analyzed as one movement paths when all the shortest paths are traced between nodes in (i.e., loop to the same node). Additionally, those vehicles stopping the network. for more than 24 h at the nodes were considered as “parked” To identify cohesive groups in the network, we recognized fully and not performing an “operational activity” as they could be connected nodes in the network (i.e., cliques) and densely under maintenance or quarantine (e.g., downtime). The latter connected nodes (i.e., communities) across the networks situation was included in the network analysis but excluded from constructed. Communities were detected using the algorithm base our temporal analysis since they did not represent daily routine on propagating labels as an efficient and ease-of-implementation operation and the time spent at each node can be overestimated. algorithm (20). In brief, unique labels are assigned initially to each Network characteristics and features were described by the node, and after each iteration of the algorithm, new labels are calculation of (a) network parameters, including the number of adopted by the nodes, based on the most frequent labels of the nodes and edges, average path length, diameter, edge density, and neighbors until they converge into one label which represents the clustering coefficient, and (b) centrality metrics, such as degree (in community. In the case of ties among neighboring labels, one is and out), closeness (in and out), and betweenness for each node randomly selected before the next algorithm iteration. in the networks created. Definitions for each network parameter All statistical analyses were performed using R Studio and metrics have been reviewed and described in a previous study v4.1.0 software (21). Network creation, description, parameter by Martínez-López et al. (19). In brief, average path length is the calculations, and visualization were performed via the igraph average number of additional nodes contacted or visited in the package v1.2.6 (22). trajectory to go from nodei to nodej in the network, diameter is the length of the shortest path within the two most distant nodes in the Results network (i.e., nodes that need the largest number of intermediate nodes to be able to connect each other), edge density indicates the Site and movement descriptive results proportion of contacts that occur in the network over the potential ones happening, and clustering coefficient (i.e., transitivity) is the From the predefined 114 sites (i.e., nodes), 12 vehicles moved frequency of loops connecting nodes reciprocally between each among them within 2 years of the study period, comprising 12 other (5). The degree is the number of contacts that each node sow farms, 8 gilt development units (GDU), 17 nurseries, 71 has overall and, when directionality is applied, the number of wean-to-market (WTM), 4 finishers (i.e., sites/unique nodes), and connections originated (out-degree) or received (in-degree) by that 2 truck wash facilities. The nodes were located in Minnesota, node. Closeness is an estimate of how closely one node is connected South Dakota, Nebraska, and Iowa. After excluding 26% of the to every other node in the network based on all incoming (in- movements recorded as described earlier, 4,579 movements were closeness) or outgoing (out-closeness) connections, and betweenness analyzed, with 60% (2,749/4,579) of them occurring in 2019 and represents the number of times certain node lies on the shortest the remaining 40% (1,830/4,579) in 2020. Frontiers in Veterinary Science 04 frontiersin.org Picasso-Risso et al. 10.3389/fvets.2023.1201644 TABLE 1 Frequency of in-movements by farm-type (%). Graded scale colors represent a higher (red) and lower (white) percentage of movements across dyad connections in the 2 years of study (2019–2020). Origin/ destination GDU NURSERY SOW TRUCK WASH WTM FINISHER FINISHER 0.07 0.02 0.04 0.07 0.00 0.00 GDU 1.62 0.07 1.81 0.07 0.02 0.00 NURSERY 0.20 1.62 1.18 3.49 0.28 0.00 SOW 1.27 4.02 5.48 5.87 19.76 0.00 TRUCK WASH 0.50 0.63 20.92 2.42 2.07 0.00 WTM 0.02 0.44 7.29 14.48 1.05 0.00 GDU, Gilt development unit, WTM, Wean to Market. FIGURE 3 Density of incoming movements within the study area. Higher (dark) and lower (light) density areas (movements per county) are represented for 2019 and 2020, respectively. Spatiotemporal results Supplementary Table 2). The spatial distribution of movements was consistently higher in the central area of the Midwest states The monthly and seasonal in-movement counts were similar analyzed for both years mapped (Figure 3). across 2019 and 2020 (negative binomial regression P > 0.05), with Overall, time spent in the nodes was higher (RR: 1.1, P < 0.03) 10–15% of the sites performing at least one movement weekly. The in 2020 than in 2019, with higher time spent at truck wash sites in lowest frequency of movements was observed in weeks 70–78 of 2020 but without any significant variation observed when analyzed the study (i.e., May and June 2020, Figure 2). Movements toward by season (Table 2). However, time spent at truck wash facilities sow farms in the winter and summer of 2019 were significantly was significantly higher (P < 0.001) than any other site categories, higher (P < 0.05) than movements toward any other farm category with the occurrence of this category being 3.32, 5.83, 14.86, 16.00, or season-year with mean risk ratios ranging from 1.10 to 2257.34. and 21.19 times higher, on average, than sow farms, GDU, finishers, Furthermore, there were fewer movements to finishers in spring nursery, and WTM sites, respectively. 2020 (P < 0.01) than any other category or season-year, with risk ratios ranging from 0.0018 to 0.5 (Supplementary Table 1). The seven trucks performed the majority of the movements in Network characterization both years studied, with 76.8% (2,111/2,749) in 2019 and 66.99% (1,226/1,830) in 2020, while the five trailers did the rest of the Descriptive statistics for network parameters and centrality movements, 638 in 2019 and 604 in 2020. metrics for the complete network and the network constructed In-movements classified by production type dyad indicated for 2019 and 2020, respectively, are shown in Table 3. For the that 55.43% of the in-movements were performed by three dyads, global network, on average, every site can be reached typically sow farm to WTM sites (19.76%/4,579), WTM to truck wash in three steps (average path distance), with less than a one-step (14.48%), and truck wash to sow farm (20.92%). The remaining reduction in 2019 and an increase in 2020, respectively. Although in-movements were performed by multiple dyads (Table 1 and less than approximately 35% of the potential connections that Frontiers in Veterinary Science 05 frontiersin.org TABLE 2 Median (max-min) minutes spent in each node (i.e., sites) overall by production type and by season during the study period. Frontiers in Veterinary Science Picasso-Risso et al. Production type FINISHER GDU NURSERY SOW TRUCK WTM Overall WASH Stop time Median n Median n Median n Median N Median n Median n Median n (Min, Max) (Min, Max) (Min, Max) (Min, Max) (Min, Max) (Min, Max) (Min, Max) Winter19a 44.0 25 36.0 66 28.0 215 162 173 26.0 150 36.0 629 (6.00, 602) (14.0, 76.0) (6.00, 1,220) (11.0, 1,360) (9.00, 75.0) (6.00, 1,360) Spring19 50.0 1 64.0 25 36.0 32 26.0 246 149 195 22.0 167 32.0 666 (50.0, 50.0) (6.00, 478) (9.00, 60.0) (8.00, 1,290) (9.00, 1,410) (6.00, 146) (6.00, 1,410) Summer19 NA 60.0 15 29.0 62 26.0 267 149 210 20.0 171 32.0 725 (18.0, 160) (10.0, 46.0) (6.00, 1,280) (8.00, 1,440) (6.00, 101) (6.00, 1,440) Fall19 20.0 1 44.0 34 30.0 43 27.5 244 152 187 20.0 167 32.0 676 (20.0, 20.0) (10.0, 148) (8.00, 76.0) (6.00, 1,260) (6.00, 1,390) (10.0, 64.0) (6.00, 1,390) 2019 35.0 2 54.0 99 32.0 210 27.0 992 150+ 779 22.0 668 34.0 2,750 (20.0, 50.0) (6.00, 602) (8.00, 76.0) (6.00, 1,290) (6.00, 1,440) (6.00, 146) (6.00, 1,440) Winter20b 70.0 25 26.0 39 28.0 173 870 137 24.0 116 34.0 490 (8.00, 167) (6.00, 60.0) (6.00, 1,270) (11.0, 1,420) (10.0, 179) (6.00, 1,420) Spring20 NA 98.0 13 22.0 12 31.0 112 127 73 20.0 43 42.0 253 (26.0, 229) (16.0, 128) (6.00, 1,280) (10.0, 1,350) (10.0, 66.0) (6.00, 1,350) Summer20 NA 110 15 31.0 24 43.5 218 733 125 24.5 120 45.0 502 06 (38.0, 155) (14.0, 88.0) (6.00, 1,240) (6.00, 1,430) (6.00, 94.0) (6.00, 1,430) Fall20 32.0 7 77.0 16 31.0 40 51.0 233 459 136 26.0 144 47.0 576 (9.00, 58.0) (44.0, 167) (6.00, 77.0) (7.00, 1,330) (6.00, 1,440) (8.00, 80.0) (6.00, 1,440) Winter21c 110 1 25.0 2 44.5 30 185 13 28.0 17 44.0 63 (110, 110) (24.0, 26.0) (18.0, 1,260) (12.0, 1,350) (14.0, 59.0) (12.0, 1,350) 2020 32.0 7 81.0 70 26.0 110 40.0 746 345+ 470 24.0 427 42.0 1,830 (9.00, 58.0) (8.00, 229) (6.00, 128) (6.00, 1,330) (6.00, 1,440) (6.00, 179) (6.00, 1,440) Overall 32.0 9 60.0 169 30.0 320 32.0 1,738 176∗ 1,249 22.0 1,095 36.0 4,580 (9.00, 58.0) (6.00, 602) (6.00, 128) (6.00, 1,330) (6.00, 1,440) (6.00, 179) (6.00, 1,440) n, number of stops per season and production type. Multivariate negative binomial is significantly different (P < 0.05), (∗ ) from other production types overall or between estimates from 2019 and 2020 (+). GDU, Gilt development unit, WTM, Wean to Market. a, Winter 19 includes January and February 2019; b, Winter 20 includes December 2019, January and February 2020; c, Winter 21 includes December 2020. 10.3389/fvets.2023.1201644 frontiersin.org Picasso-Risso et al. 10.3389/fvets.2023.1201644 could occur really occurred (edge density) in the complete network when assessing it on a per-year basis of the study period, the proportion doubled in 2019 compared with 2020 (43.5 vs. 25.6%). Furthermore, degree, in-degree, and out-degree were higher for 2019 than 2020 with important higher outliers observed, corresponding to truck wash facilities and sow farms. A similar tendency was 0.0027 (1.4e-04–0.0036) 0.0027 (1.4e-04–0.004) Mean (min–max) 0.003 (1.4e-04–0.004) observed for centrality closeness. However, betweenness showed 203.72 (0–3017.81) 43.06 (2–653) 21.53 (1–325) 21.53(1–325) higher values for 2020 than 2019, and the nodes with the highest 2020 0.2563 3.4837 0.2573 1,830 85 values were the two truck wash facilities (Figure 4B, red and 7 orange circles). Complete network cluster analysis recognized three communities, and 85.3% (2,808/3,292) of the cliques observed were fully connected in pairs (i.e., cliques’ size of two). When yearly analyses were performed, larger cliques with up to seven nodes fully TABLE 3 Network parameters and centrality metrics at the node level calculated for the complete vehicle network and the 2019 and 2020 networks, respectively. interconnected were identified each year (Table 4). Furthermore, two and four communities were observed in 2019 and 2020, respectively, after excluding those with two or fewer members. Discussion 0.0036 (1.5e-04–0.0052) 0.0035 (1.5e-04–0.0048) 0.004 (1.6e-04–0.0058) Mean (min–max) 122.94 (0–3539.24) 68.72 (2–1,546) 34.36 (1–770) 34.36 (1–776) This study, to the best of our knowledge, was the first 2019 2,749 0.435 2.596 0.205 description of vehicle movements using GPS data of a typical 80 7 pig production system in the Midwestern US. Moreover, we have provided the first comparison between networks from a typical year of production (2019), and 2020, a year in which the swine industry was strongly affected due to the temporary decline in slaughter rates as a result of the COVID-19 pandemic in the USA (23). A total of 74% of the records were accurate and used in these analyses. Data generated by vehicle-tracking devices have shown to be a robust tool to understand connectivity among farms, which can give a more in-depth picture of the vehicle trajectory, sites visited, and time spent at each stop. This data can serve as a proxy for compliance regarding times Complete network 0.0022 (7.628e-05–0.0033) 0.0024 (7.63e-05–0.0034) 0.0022 (7.68e-05–0.0028) Mean (min–Max) dedicated to animal loading/unloading, vehicle flow and vehicle 234.40 (0–5694.07) 79.63 (2–2,199) 39.82 (1–1,095) 39.82 (1–1,104) cleaning, and disinfection and downtime procedures (7). The 0.3493 3.0925 0.2218 4,579 114 majority of the movements were performed in 2019, with a 20% 7 reduction during the following year and strong fluctuations in the proportion of farms with weekly movements in 2020 (5–20% of the sites performing weekly movements). As expected, trucks that performed the majority of movements were those used for transporting weaned piglets that, according to production management practices, were weaned from sow farms more than a week ago. The highest reduction of vehicle movements was during spring 2020, with a follow-up increase in mid-2020, mostly due to sow and WTM movements. One potential explanation could be the limitations associated with the high rates of COVID-19 illness among workers at slaughter plants, limiting the capacity for slaughter in abattoirs during the spring of 2020 (24) and, in other cases, the temporary shutdown of major slaughter plants in the region during April and May 2020 (23). Although, movements Global parameters to slaughter were not included in this study, limitations on pig- Node parameters slaughter is expected to affect the pig flow within the systems in Transitivity (global) Average path length the initial levels of production, which will recover later in that year. Edge density Betweenness Another potential explanation could be an increase in sites covered Closeness Diameter by the trucks in response to the initial movement reduction, Degree Nodes Edges Out Out despite only 5 extra nodes being observed in 2020 compared to In In 2019 (85/80). Frontiers in Veterinary Science 07 frontiersin.org Picasso-Risso et al. 10.3389/fvets.2023.1201644 FIGURE 4 Network representation of the complete period analyzed (2019–2020). Heat scale colors represent (A) closeness and (B) betweenness relative values for each node, with the highest (red) and lowest (white) node values for each centrality metric, respectively. The five different categories of nodes and their connections are shown in different colors as reference (C). TABLE 4 Cliques and communities identified by the year of study (2019 and 2020). 2019 2020 Cliques (N = 805) Community Nodes Cliques (N = 594) Community Nodes member member (N = 80) (N = 85) Size n n (n/N) Size n n (n/N) 2 250 1 66 (0.825) 2 220 1 36 (0.424) 3 297 2 10 (0.125) 3 205 2 10 (0.118) 4 178 4 109 3 12 (0.141) 5 62 5 44 4 24 (0.282) 6 16 6 14 7 2 7 2 Communities with one or two members (i.e., nodes) are not included in this table. As expected, a level of “fidelity” was observed within sites, with wash facilities to sow farms and from sow farms to WTM farms. vehicles connecting sow, WTM, and truck wash facilities in the However, approximately 14.5% of movements were connecting direction indicated by the production system. We observed that WTM to truck wash instead of the approximately 20% expected, approximately 20% of the movements were performed from truck which seemed to be balanced with movements from WTM to other Frontiers in Veterinary Science 08 frontiersin.org Picasso-Risso et al. 10.3389/fvets.2023.1201644 sites, such as sow farms and WTM. Although WTM to WTM patterns of different states or regions of the USA. Another movements are common when pigs need to be finished in different limitation is related to the number of trucks and trailers monitored, sites, the movements of WTM to sow farms at the commercial level which perhaps underestimated connectivity trends. Although a are rare as their terminal genetic makeup of the commercial market low (9.3%) proportion of company vehicles were enrolled in the pig populations is not suitable if these animals were to be used study, these vehicles were those used the most and continuously as replacement breeding stock. Additionally, commercial growing during 2019–2020, resulting in a robust representation of typical pigs housed in WTM facilities are usually located in highly dense movement patterns before and during the COVID-19 pandemic. regions where the prevalence of endemic pathogens is high, which In addition, the network built in this study was composed of will represent a risk to the sow farm if these were to be introduced. breeding farms, growing pig farms, and truck wash facilities which WTM to sow farm movement exists but these are related to moving leave out the connectivity with slaughter plants and market-hog high-health gilts from the genetic multiplication herds that tend transport cleaning and disinfection facilities. Furthermore, the to have better health than commercial sow farms. Overall time unusual aspects observed during the COVID-19 global pandemic spent in truck wash facilities (∼3 h) denotes a sufficient time for highlight a unique situation within the swine industry. However, the washing and drying process recommended for animal transport this analysis builds on the understanding of the complex network within the industry (25), with a noticeable median increase among interactions of vehicles within the Midwest swine industry and seasons in 2020, although maximum times remain constant. This gives insights into how unique events can affect the flow of the pork can indicate changes in disinfection procedures due to COVID-19 supply production, informing preparedness for future infectious awareness which resulted in longer times at the truck wash facilities. disease epidemics and events. It is important to mention that trucks could spend more time at the truck wash after being cleaned as they would be parked until needed again. This could bias the parameter, overestimating the Conclusion mean time spent at the truck wash facilities overall. This metric needs to be further studied and analyzed but is the first insight as a In this study of vehicle movements within a pig production proxy for biosecurity compliance. Furthermore, there were more system, we observed a highly connected and structured organized paths connecting truck washes in 2020, which together with the network with high fidelity among the site triad of sow, WTM, increase in the median time at truck wash observed can indicate and truck wash. We also described the impact on movements and that these changes in cleaning and disinfection procedures could network connectivity in 2020, when the peak of the COVID-19 potentially be a result of an increase in biosecurity concerns as a epidemic was observed within the USA. The directionality whole and affected the connectivity among the network that year. of movements among sites and time spent at truck wash The observed larger number of communities in 2020 could be facilities represented a proxy for good biosecurity practices within reflecting management practices applied to segregate the network the system. Monitoring and understanding spatiotemporal and and avoid disease transmission between sites. While we lack data to network patterns of movements beforehand can prove useful in identify the reason for observing these nodes largely interconnected identifying high-risk areas and preparedness in the face of a new (i.e., clusters, cliques, and communities), it exposes the potential disease epidemic. The results also showed the grouping of farms usefulness of these datasets when applying control or surveillance within the system when considering the vehicles, indicating that strategies targeting groups of sites at high/low risk of infection many farms were connected by these indirect means. Knowledge through vehicle movements (16, 26). obtained from this movement characterization in the Midwest Network metrics displayed a high interconnection among may assist with targeting interventions, prevention, and control sites through vehicles, in which regardless of the geographical strategies for infectious diseases within the swine industry, with an distance among farms, they were only three steps away from each emphasis on monitoring and reducing the prevalence in different other (i.e., mean distance). Areas with a high density of vehicle communities connected by vehicles. movements were detected. This finding shows that movements are not randomly distributed among all sites integrating the system, highlighting potential target areas for disease active surveillance Data availability statement when resources are limited as described in other production The datasets presented in this article are not readily because systems globally (5, 27, 28). The validity of this finding in other available are confidential. Requests to access the datasets should be states is yet to be evaluated. Comparison of these network metrics directed to corzo@umn.edu. with others can be challenging since systems, data collection, movements, inclusion criteria, and time frame can vary across studies (5, 29, 30). However, the need to improve within-system Author contributions movement records has been reported, as it represents an important factor in disease spread (6), and our study provides information on CP-R, CV, and CC: conceptualization. CV, JS, and CC: one of the denser swine production regions within the USA, and fieldwork and data collection. CC and MS: resources. CP-R: insights are representative of the Midwestern USA swine industry. methodology and data analysis and writing the original draft. CP-R, A limitation of this study is that the network characterization CC, MK, CV, and JS: writing—reviewing and editing. CC: project was performed in one production system representative of current administration. CC and CV: funding acquisition. All authors have pig production practices in the US and might not fully represent read and agreed to the published version of the manuscript. Frontiers in Veterinary Science 09 frontiersin.org Picasso-Risso et al. 10.3389/fvets.2023.1201644 Funding that could be construed as a potential conflict of interest. The MSHMP is a Swine Health Information Center (SHIC) funded project (Project # 19-235 SHIC). Publisher’s note All claims expressed in this article are solely those of the Acknowledgments authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the The authors thank the MSHMP participating system reviewers. Any product that may be evaluated in this article, or for the willingness to collaborate in this study and claim that may be made by its manufacturer, is not guaranteed or the time committed to solving questions regarding the endorsed by the publisher. data collected. Supplementary material Conflict of interest The Supplementary Material for this article can be found The authors declare that the research was conducted online at: https://www.frontiersin.org/articles/10.3389/fvets.2023. in the absence of any commercial or financial relationships 1201644/full#supplementary-material References 1. Neumann E, Hall W, Dahl J, Hamilton D, Kurian A. Is transportation a risk 14. VanderWaal K, Picasso C, Enns EA, Craft ME, Alvarez J. Network factor for African swine fever transmission in Australia: a review. Aust Vet J. (2021) analysis of cattle movements in Uruguay: quantifying heterogeneity 99:459–68. doi: 10.1111/avj.13106 for risk-based disease surveillance and control. Prev Vet Med. (2016) 123:12–22. doi: 10.1016/j.prevetmed.2015.12.003 2. Smith RP, Cook AJC, Christley RM. Descriptive and social network analysis of pig transport data recorded by quality assured pig farms in the UK. Prev Vet Med. (2013) 15. Melmer DJ, O’Sullivan TL, Greer AL, Poljak Z. An investigation of transportation 108:167–77. doi: 10.1016/j.prevetmed.2012.08.011 practices in an Ontario swine system using descriptive network analysis. PLoS ONE. (2020) 15:e0226813. doi: 10.1371/journal.pone.0226813 3. VanderWaal K, Perez A, Torremorrell M, Morrison RM, Craft M. Role of animal movement and indirect contact among farms in transmission of porcine epidemic 16. Sterchi M, Faverjon C, Sarasua C, Vargas ME, Berezowski J, Bernstein A, et al. diarrhea virus. Epidemics. (2018) 24:67–75. doi: 10.1016/j.epidem.2018.04.001 The pig transport network in Switzerland: structure, patterns, and implications for the transmission of infectious diseases between animal holdings. PLoS ONE. (2019) 4. Galvis JA, Corzo CA, Machado G. Modelling and assessing 14:e0217974. doi: 10.1371/journal.pone.0217974 additional transmission routes for porcine reproductive and respiratory syndrome virus: vehicle movements and feed ingredients. 17. Perez AM, Linhares DCL, Arruda AG, VanderWaal K, Machado G, Transbound Emerg Dis. (2022) 69:e1549–60. doi: 10.1111/tbe Vilalta C, et al. Individual or common good? Voluntary data sharing to d.14488 inform disease surveillance systems in food animals. Front Vet Sci. (2019) 6:194. doi: 10.3389/fvets.2019.00194 5. Lee K, Polson D, Lowe E, Main R, Holtkamp D, Martínez-López B. Unraveling the contact patterns and network structure of pig shipments in the United States and 18. Byers AL, Allore H, Gill TM, Peduzzi PN. Application of negative its association with porcine reproductive and respiratory syndrome virus (PRRSV) binomial modeling for discrete outcomes: a case study in aging research. outbreaks. Prev Vet Med. (2017) 138:113–23. doi: 10.1016/j.prevetmed.2017.02.001 J Clin Epidemiol. (2003) 56:559–64. doi: 10.1016/S0895-4356(03)0 0028-3 6. Machado G, Vilalta C, Recamonde-Mendoza M, Corzo C, Torremorell M, Perez A, et al. Identifying outbreaks of Porcine Epidemic Diarrhea 19. Martínez-López B, Perez AM, Sánchez-Vizcaíno JM. Social network virus through animal movements and spatial neighborhoods. Sci Rep. (2019) analysis. Review of general concepts and use in preventive veterinary medicine. 9:457. doi: 10.1038/s41598-018-36934-8 Transbound Emerg Dis. (2009) 56:109–20. doi: 10.1111/j.1865-1682.2009.0 1073.x 7. Dee SA, Deen J, Otake S, Pijoan C. An experimental model to evaluate the role of transport vehicles as a source of transmission of porcine reproductive and respiratory 20. Raghavan UN, Albert R, Kumara S. Near linear time algorithm to syndrome virus to susceptible pigs. Can J Vet Res. (2004) 68:128–33. detect community structures in large-scale networks. Phys Rev E. (2007) 76:036106. doi: 10.1103/PhysRevE.76.036106 8. Thakur KK, Revie CW, Hurnik D, Sanchez J. Modelling contamination of trucks used in the shipment of pigs infected with porcine reproductive and respiratory 21. R Core Team. R: A Language And Environment For Statistical Computing. syndrome virus. J Swine Health Product. (2017) 25:183–93. Vienna: R Foundation for Statistical Computing (2017). Available online at: https:// www.R-project.org/ 9. Alarcón LV, Allepuz A, Mateu E. Biosecurity in pig farms: a review. Porcine Health Manage. (2021) 7:5. doi: 10.1186/s40813-020-00181-z 22. Csárdi G, Nepusz T. The igraph software package for complex network research. InterJ Complex Syst. (2006) 1695. 10. Chaters GL, Johnson PCD, Cleaveland S, Crispell J, de Glanville WA, Doherty T, et al. Analysing livestock network data for infectious disease control: an argument for 23. Padilla S, Schulz LL, Vaiknoras K, MacLachlan M. COVID-19 Affected Hog routine data collection in emerging economies. Philos Trans R Soc Lond Ser B Biol Sci. Slaughter Volumes in 2020, but Impacts and Trajectory of Recovery Differed Across (2019) 374:20180264. doi: 10.1098/rstb.2018.0264 Regions. USDA Economic Research Service (2021). Available online at: https://www. ers.usda.gov/amber-waves/2022/february/covid-19-affected-hog-slaughter-volumes- 11. Lentz HHK, Koher A, Hövel P, Gethmann J, Sauter-Louis C, in-2020-but-impacts-and-trajectory-of-recovery-differed-across-regions/ Selhorst T, et al. Disease spread through animal movements: a static and temporal network analysis of pig trade in Germany. PLoS ONE. (2016) 24. Hayes DJ, Schulz LL, Hart CE, Jacobs KL. A descriptive analysis of the COVID- 11:e0155196. doi: 10.1371/journal.pone.0155196 19 impacts on U.S. pork, turkey, and egg markets. Agribusiness. (2021) 37:122– 41. doi: 10.1002/agr.21674 12. Marquetoux N, Stevenson MA, Wilson P, Ridler A, Heuer C. Using social network analysis to inform disease control interventions. Prev Vet Med. (2016) 126:94– 25. Connor J, Lower A, Lowe J. Truck Wash Biosecurity Critical. National Hog 104. doi: 10.1016/j.prevetmed.2016.01.022 Farmer (2014). Available online at: https://www.nationalhogfarmer.com/news/isu- economist-lee-schulz-receives-applied-research-award 13. Silk MJ, Croft DP, Delahay RJ, Hodgson DJ, Boots M, Weber N, et al. Using social network measures in wildlife disease ecology, epidemiology, and management. 26. Dorjee S, Revie CW, Poljak Z, McNab WB, Sanchez J. Network analysis of swine Bioscience. (2017) 67:245–57. doi: 10.1093/biosci/biw175 shipments in Ontario, Canada, to support disease spread modelling and risk-based Frontiers in Veterinary Science 10 frontiersin.org Picasso-Risso et al. 10.3389/fvets.2023.1201644 disease management. Prev Vet Med. (2013) 112:118–27. doi: 10.1016/j.prevetmed.2013. 29. Arruda AG, Friendship R, Carpenter J, Hand K, Poljak Z. Network, 06.008 cluster and risk factor analyses for porcine reproductive and respiratory syndrome using data from swine sites participating in a disease control 27. Martin V, Zhou X, Marshall E, Jia B, Fusheng G, FrancoDixon MA, et al. program. Prev Vet Med. (2016) 128:41–50. doi: 10.1016/j.prevetmed.201 Risk-based surveillance for avian influenza control along poultry market chains 6.03.010 in South China: the value of social network analysis. Prev Vet Med. (2011) 196– 205. doi: 10.1016/j.prevetmed.2011.07.007 30. Büttner K, Krieter J, Traulsen A, Traulsen I. Static network analysis of a 28. VanderWaal K, Enns EA, Picasso C, Alvarez J, Perez A, Fernandez F, et al. pork supply chain in Northern Germany-Characterisation of the potential spread Optimal surveillance strategies for bovine tuberculosis in a low-prevalence country. of infectious diseases via animal movements. Prev Vet Med. (2013) 110:418– Sci Rep. (2017) 7:4140. doi: 10.1038/s41598-017-04466-2 28. doi: 10.1016/j.prevetmed.2013.01.008 Frontiers in Veterinary Science 11 frontiersin.org