Estimating sampling and laboratory capacity for a simulated African swine fever outbreak in the United States

The introduction of African swine fever virus (ASFV) into uninfected countries can impact1 economic and animal welfare. Rapid detection and control of the outbreak contribute to2 successful eradication and promote business continuity. We developed a model to determine3 thenumberofsamples, samplecollectors, laboratorycapacity, andprocessingtimesfollowing4 an ASFV introduction into the U.S. We simulated the spread of ASFV in one densely5 populated swine region, generating a median of 27 (range = 1-68) outbreaks in 150 days,6 resulting 616 (range = 1-15,011) sampling events with a total of 3,068 barns (range = 7-7 69,118)sampled. Wecalculatedthetotalsamplecollectorsneeded, consideringdailyworking8 hours, sampling and driving time, and laboratory capabilities with and without blood sample9 pooling. Samples included 31 blood samples and five oral fluid samples per barn, which equal10 84,830 (range = 52-2,066,831) and 14,195 (range = 10-345,590) blood and oral fluid samples,11 respectively. The median number of sample collectors needed to prevent sampling delay12 varied from 136 to 367 and, in the worst epidemic scenarios, from 833 to 3,115. Notably,13 excluding downtime–which prevented the sampler from visiting additional farms for 24 or14 72 hours–reduced the number of sample collectors needed between 28% and 75%, while15 switching from blood to oral fluid samples reduced this number between 47% and 75%. At a16 laboratory processing daily capacity of 1,000 samples, the median days for sample processing17 without pooling were 92 days, with a maximum of 5.7 years. We demonstrated a need18 to redistribute 10,062 (range = 2-67,940) unprocessed samples daily to other laboratories19 to prevent processing delays. Our study addresses the challenge of efficiently organizing20 resources for managing a potential ASFV outbreak, providing information about the number21 of sample collectors and laboratory capacity needed for one densely populated swine region22 in the U.S.23

Disease surveillance, sampler, pigs, response plan, outbreak management, foreign animal disease. Preprint submitted to Preventive Veterinary Medicine September 26, 2024 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint Introduction24 African swine fever virus (ASFV) is a highly contagious hemorrhagic viral disease af-25 fecting domestic swine and feral and wild pigs (Blome et al., 2020). It has disseminated26 worldwide, spreading widely through Europe (Cwynar et al., 2019), Asia (Mighell and Ward,27 2021) and Africa (Njau et al., 2021), and has more recently been identified in the island of28 Hispaniola in the Americas (Jean-Pierre et al., 2022; Schambow et al., 2023). The poten-29 tial introduction of ASFV into the U.S. swine industry presents a significant challenge that30 threatens not only animal health but also the national economic stability and food security31 (Sánchez-Cordón et al., 2018; Dixon et al., 2019). Consequently, robust surveillance and32 early detection mechanisms have become paramount to prevent catastrophic losses and en-33 sure business continuity of the swine industry (Sykes et al., 2023; Checkoff, 2023; Carriquiry34 et al., 2020). A critical component of a national response plan includes efficient collection,35 processing, and analysis of samples to promptly detect infected premises, and demonstrate36 ASFV effective control measures and disease elimination (Gallardo et al., 2019; Schambow37 et al., 2022).38 In the U.S., swine veterinarians oversee or directly handle the collection of samples to39 detect and control endemic pathogens. However, in the event of an ASFV epidemic, the40 number of swine industry veterinarians and animal health officials may be insufficient given41 the large number of samples needed to surveil for disease (Secure Pork Supply Plan, 2024).42 To address this challenge, a collaborative initiative involving the swine industry, government43 officials, and academic experts has developed a certified swine sample collector training44 program (Secure Pork Supply Plan, 2024). This program aims to train on-farm personnel,45 suchasanimalcaretakers, tocollect, package, andsubmitsamples. Byincreasingthenumber46 of qualified personnel who can submit high-quality samples, this program is expected to47 enhancetheefficiencyandeffectivenessoftheresponseplantocontrolASFVspread. Despite48 this unique sample collector program, the process of sample collection is intricate, requiring49 a well-coordinated effort that hinges on the availability of sufficient manpower to collect a50 large volume of samples (USDA, 2023; Alvarez et al., 2023). Insufficient staffing can lead to51 delays in sample collection and, consequently, delays in the detection and response (Sykes52 et al., 2023; Hayes et al., 2021).53 The volume and type of samples collected are critical factors when estimating the num-54 ber of sample collectors and time involved in ASFV response. The U.S. ASFV response55 plan includes blood swab samples from suspected farms for initial investigation (USDA,56 2023). Furthermore, these plans also have established that farms within infected, buffer,57 and surveillance zones, and those directly linked with ASFV-positive premises via recent58 animal shipments or vehicle movements need to be tested (USDA, 2023; Sykes et al., 2023).59 Although this national response plan established an ASFV sampling scheme, the number60 of samples, sample collectors, time, and resources necessary to collect and process these61 ∗These authors contributed equally to this work. ∗∗Corresponding Author. Email address: [email protected] (Gustavo Machado) 2 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint samples during an ASFV incursion could vary and have yet to be estimated. This gap in62 our understanding highlights opportunities to evaluate additional information to support63 planning to effectively control an ASFV outbreak.64 The ongoing outbreaks of ASFV in Europe after its reintroduction in 2007 have un-65 derscored the critical importance of timely detection to prevent a foreign animal disease66 from becoming endemic within new territories (Cwynar et al., 2019). To prepare the U.S.67 swine industry, a comprehensive understanding of the logistics involved in sample collection68 and processing is needed to optimize the strategic planning and implementation of ASFV69 response strategies (Cochran et al., 2023). This work aimed to evaluate the logistical con-70 siderations of sampling during a simulated ASFV outbreak in the U.S., calculating the total71 and daily number of blood samples, and an alternative scenario with oral fluids, the neces-72 sary personnel to collect these samples, the time from collecting to processing samples, and73 the laboratory capacity needed.74 Methodology75 Data76 This study used data from 1,898 farms from eight commercial swine companies in one77 U.S. state. The data comprises farm identification, production type (e.g., wean-to-finisher),78 number of pigs per farm, latitude, and longitude. For each farm, we also collected the lines79 of separations (LOS), which is part of enhanced on-farm Secure Pork Supply (SPS) biosecu-80 rity plans, to identify the number of barns within each farm (Center for Food Security and81 Public Health, 2017). These data were collected from the companies through the Rapid Ac-82 cess Biosecurity application (RABappTM) (https://machado-lab.github.io/rabapp/), a web83 application that serves as a platform for standardizing the approval of SPS biosecurity plans84 while storing and analyzing animal movement data (Machado et al., 2023). In this study,85 9.5% of farms were not in RABappTM, missing essential LOS data to identify the number86 of barns. We estimated the number of barns for these farms based on the average number87 of barns from the remaining 90.5% farms with LOS information. In addition, we collected88 addresses, latitudes, and longitudes from 27 sample supply offices of the participant com-89 panies and one certified laboratory from the National Animal Health Laboratory Network90 (NAHLN) to process ASFV samples from all swine commercial companies. This labora-91 tory reported the capacity to process 1,000 ASFV samples daily through polymerase chain92 reaction (PCR) tests, following the ASFV laboratory guidelines (USDA, 2023).93 The commercial swine companies also provided pig and vehicle movements from January94 1st, 2020, to December 31st, 2020. Animal movement data included the identification num-95 ber of farms sending and receiving pigs, the date, and the number of pigs moved. Vehicle96 movement was shared by two companies that represented 97% of the farms within the region,97 this data comprehended geographic coordinates every five seconds, speed (km/h), date, and98 time for each vehicle. The companies provided a list of 599 vehicles; these vehicles include99 (i) 224 trucks used to deliver feed to farms; (ii) 168 vehicles utilized in the transportation of100 live pigs between farms; (iii) 125 vehicles used in the transportation of pigs to markets (a.k.a.101 slaughterhouse, packing plants); and (iv) 82 vehicles without a defined role, which are used102 3 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint for multiple tasks such as delivering feed and pigs. Finally, we also gathered information103 on 14 company-owned cleaning stations where vehicles are regularly cleaned; more details104 about the vehicle data and contact network are presented elsewhere (Galvis and Machado,105 2024a).106 Figure 1: ASFV sampling flow. The diagram shows the order of events a sample collector follows to collect samples from farms and deliver them to the laboratory African swine fever outbreak simulation107 We simulated the ASFV epidemic through a stochastic, farm-level, compartmental trans-108 mission model with four health states: Susceptible (S), Exposed (E), Infected (I), and109 Detected (D), available in PigSpread-ASF (Sykes et al., 2023). This model included six110 transmission routes: i) movement of pigs between farms; ii) local transmission, reflect-111 ing transmission related to spatial proximity; iii) vehicles moving pigs between farms (pig112 trucks); iv) vehicles moving pigs from farms to slaughterhouses (market trucks); v) vehicles113 delivering feed to farms (feed trucks); and vi) vehicle movements between farms without a114 defined role (undefined trucks) (Galvis and Machado, 2024a). The pig movement network115 was reconstructed using the origin and destination farms of the pigs moved daily, and the ve-116 hicle contact network was reconstructed by a methodology that used the proximity between117 4 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint vehicles and farms to identify a vehicle contacting a farm while considering the effect of dis-118 infection when vehicles are at a cleaning station (Galvis and Machado, 2024a). Ultimately,119 ASFV dissemination via pig and vehicle is driven by directed temporal networks. The lo-120 cal transmission was based on a kernel density where the probability of infection decreases121 with increased distance. Initial simulation conditions included ASFV infection seeded in122 a random farm for each simulation, and transmission was allowed from day two. 94,900123 simulations were run, which included seeding infection in each of the farms 50 times. The124 ASFV model simulated all control actions listed in the U.S. response plan (USDA, 2023),125 including i) depopulation of ASF-positive farms; ii) a 72-hour standstill of live pig move-126 ments; iii) contact tracing of farms connected to ASF positive cases by animal and vehicle127 movements; and iv) the implementation of control areas (a 3 km infected zone and a 2 km128 buffer zone) and surveillance zones (5 km). The model calculated the number of diagnostic129 tests required for both surveillance purposes and movements to/from farms in the control130 areas, hereafter referred to as pre-permit testing (Sykes et al., 2023). Given the lack of spe-131 cific USDA guidelines regarding the required number of individual blood swab samples for132 ASFV-suspect farms, as well as details on sampling frequency and duration (USDA, 2023),133 we adopted a strategy to collect 31 blood samples per barn based on the number required for134 pre-movement permits according to current guidelines (USDA, 2023). This approach also135 utilizes the sampling frequency and duration parameters from the USDA’s 2020 guidelines136 (USDA, 2020), as detailed in Supplementary Material Table S1. Additionally, we evaluated137 the use of oral fluid sampling. Even though oral fluids are a sample type not approved138 for ASFV testing by USDA policies, oral fluids could eventually be approved to serve as139 a complementary sample type alongside blood samples in future protocols (Goonewardene140 et al., 2021a). For this scenario, we assumed that oral fluid samples would be collected from141 five different pens per barn. Of note, we calculated logistics, sample collection, and testing142 for oral fluid based on discussions with the swine industry and state animal health officials143 (personal communication). Therefore, the results from the oral fluid should be considered as144 an initial exploration of its potential future adoption in some use cases. Our ASFV model145 calculated the samples required for all simulated outbreaks for up to 150 days.146 Sample collectors and laboratory capacity estimation147 Sampler collector route148 Sample collectors from each company are permanently assigned to pick up supplies at149 specific office locations, where they initiate their journey, which includes collecting the nec-150 essary sampling material (i.e., tube, needles) and driving to an assigned farm to perform the151 required sampling (Figure 1.) Once the sampler arrives at the farm, sampling is initiated152 immediately; the average time for each blood swab sample to be drawn was assumed to be153 five minutes. For our oral fluid alternative sample scenario, the collector used ropes hung at154 pens, and the simulated time for the five samples collection was 25 minutes (Goonewardene155 et al., 2021b; Mur et al., 2013). After the sampling, sample collectors transported the col-156 lected samples to the accredited laboratory. The sample collector’s journey path between157 supply office → farm → laboratory is calculated in our model; after calculating all possible158 pathways, the shortest driving route between these locations is used in the final simulation159 5 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint Figure 2: Farm sampling order for a group of eight farms. The first priority was given to the farm that received pigs from the ASFV-positive farm. The second priority was the farm that has been in contact with vehicles exposed to an ASFV-positive farm. The third priority was the farm requiring a pre-movement permit. Priorities four through eight involved farms within the control zones, where farm types are prioritized hierarchically. Specifically, the fourth priority was a farm within the infected zone, the fifth was the farm with the highest strategic value within the buffer zone, notably a sow farm, followed by a nursery farm, ranked sixth, and a finisher farm at seventh. The eighth priority was assigned to a farm within the surveillance zone. (Figure 1). We used Python’s package OSRM for route calculation (Luxen and Vetter,160 2011).161 We simulated the dispatch of sample collectors to farms that required sampling based162 on the ASFV national response plan sample requirements (USDA, 2023). We refer to those163 farms as "at-risk" farms. We also defined the order of sampling as follows: 1) farms that164 recorded animal visits from infected premises within the last 30 days; 2) vehicle visits from165 infected premises within the last 15 days; 3) Pre-movement permit testing from farms within166 infected and buffer zones three and one day before movement; 4) farms in infected zone(s); 5)167 farms in buffer zone(s); 6) farms in surveillance zone(s). In addition, farms with breeding age168 animals, a.k.a. sow farms, farrow-to-finisher, and boar studs, were a priority for sampling,169 followed by nursery, wean-to-finisher, and finisher farms (Figure 2.)170 6 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint Sample collectors allocation171 Our methodology utilized a Greedy Algorithm to generate solutions that satisfy the172 constraints of the situation and estimate the total number of sample collectors needed to173 finish the sampling process while keeping the sampling time reasonably low (Jungnickel,174 1999; Hernando et al., 2018). Figure 3 displays the overall flow of the sample collector175 allocation algorithm and a pseudo-code is available in Supplementary Material Appendices176 A.177 Figure 3: Flow chart of the sample collector allocation.The diagram describes the steps from the sample(s) required until the sample(s) is tested at the laboratory Our algorithm incorporates 17 conditions to estimate the number of sample collectors178 (Table 1.) It operates under the assumption that farm personnel will aid in the sampling179 process by restraining pigs. Additionally, several sample collectors were trained and certified180 and are available to sample any farm within the study region. However, once a sample181 collectorisassignedtoanofficefromaspecificcompany, thatsamplecollectorispermanently182 assigned to that location. We have implemented an allocation cap to prevent any single183 company from monopolizing all available sample collectors. This cap is proportional to184 the company’s farms within the region. For example, if company A owns 90% of farms,185 it can only claim up to 90% of the available number of sampler collectors. Furthermore,186 7 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint Table 1: Conditions to estimate the number of sample collectors and laboratory capacity Condition Description Study parameters T rained sample collectors (TSC) Sum of trained personnel to collect ASFV samples from all companies in the region 500, 1,000, 1,250, 1,500, 1,750, 2,000, and no limit Sample collectors per farm (SCPF) Upper limit on the number of collectors that can be assigned to a single farm 3 and 5 Regular working schedule (RWS) Regular number of hours a sample collector can work per day 8 hours and 12 hours Exceptional working schedule Maximum number of hours a sample collector can work per day 23 hours Collected sample type Sample methods at the farm Blood and oral fluids Collected sample size Number of samples collected from a barn 31 blood samples and five oral fluids samples Farm’s barns Number of barns within a farm. Estimated per farm through on-farm biosecurity plans Total farm samples Number of samples that need to be collected from an at-risk farm Collected sample size multiplied by Farm’s barns Time per sampling Number of minutes a sample collector needs to collect one sample at the farm Five minutes per blood sample and 25 minutes per barn with oral fluid samples Time to farm Number of minutes sample collector takes to go from the office to the farm Estimated by driving time Time to sample Number of minutes sample collector takes to collect the samples from the farm Time per sampling multiplied by Total farm samples Time to laboratory Number of minutes sample collector takes to go from the farm to the laboratory Estimated by driving time Total sampling time Number of minutes required for a sample collector to collect and deliver the samplers to the laboratory Sum of Time to farm, Time to sample and Time to laboratory Downtime Number of hours a sample collector is unavailable after finishing the sampling on a farm 0 hours, 24 hours, and 72 hours Sample pool size Number of samples combined from a single farm for processing at the laboratory 1, 5, and 10 Laboratory capacity Daily number of samples the laboratory can process 1,000, 2,500, and 5,000 Laboratory storage Total number of samples the laboratory can store Not limited the algorithm dictates that the number of samples required at a farm directly correlates187 with its number of barns (Figure 1), meaning larger farms with more barns will need more188 samples (USDA, 2023). If a single sample collector cannot complete the sampling required189 of the farm they have been dispatched to within one day due to their working schedule, the190 farm receives additional sample collectors. To prevent large farms from overwhelming the191 system, a cap on the maximum number of collectors per farm has been set (Table 1). In192 instances where no extra sample collectors can be allocated (e.g., all sample collectors are193 dispatched to farms, downtime restraints are placed on sample collectors), the algorithm194 permits existing collectors to extend their working hours (Table 1.)195 At the beginning of the simulated ASFV epidemic, the algorithm assigned one sample196 collector to each office (Figure 3), prioritizing the assignment of collectors to farms according197 to the relationship with the infected farms (e.g., zones) and farm type (Figure 2.) However,198 if the number of collectors proves insufficient to cover all farms, the algorithm deploys199 an additional collector to the office that can achieve the new sampling collection in the200 8 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint shortest time. In scenarios where downtime is minimal (e.g., no downtime), a collector201 might be capable of visiting several farms within a single day, depending on their working202 schedule (e.g., eight hours/day). However, if all trained sample collectors are either actively203 sampling or in their downtime period, the algorithm postpones further sampling activities.204 For instance, if 500 sample collectors exist, all of whom have already finished their working205 schedule for the current day, then the sampling of the remaining farms will resume the next206 day. The algorithm then seeks the next available collector who can most efficiently fulfill the207 pending sampling requests. Indeed, our approach ensures flexibility by allowing the assigned208 group of sample collectors to initiate their tasks at different times at a designated farm,209 thereby minimizing delays in the sampling process. The algorithm continues to reallocate210 sample collectors until sampling is completed across all farms. After the sampling collection211 at a farm is finished, each collector is responsible for delivering the samples to the laboratory212 before returning to their respective offices for a required downtime period (Figure 1).213 Sample processing214 Upon arrival at the laboratory, samples are stored and labeled based on the testing215 priority order of the originating farms (Figure 2.) This sample order urgency dictates their216 processing, ensuring that samples from farms with higher urgency are prioritized over those217 arriving earlier from less critical locations. The laboratory’s daily processing capacity is218 limited, and once this limit is reached, we evaluated two possible scenarios: 1) any remain-219 ing samples are deferred to the next day or 2) redistributed to other certified laboratories220 from NAHLN (USDA, 2024). In the first scenario, the backlog of unprocessed samples ac-221 cumulates, adding to the new samples received each day up to a pre-established storage222 capacity limit. In the second scenario, we assume that the surplus samples are redistributed223 on the same day they are received and that the destination laboratories can accommodate224 and process all redirected samples. Moreover, to enhance efficiency, our algorithm permits225 the laboratory to process samples individually or in pools, with a predetermined number of226 samples from the same farm combined into a single test (Figure 1). The maximum size of227 these sample pools was set at 10, based on discussions with animal health officials involved228 in ASFV surveillance in the U.S. (personal communication.)229 Outputs230 In this study, we estimated the sample collectors and laboratory processing capacities231 required for a simulated ASFV epidemic by employing a factorial combination of various232 conditions detailed in Table 1. Our model outputs are derived from the number of ASFV233 simulated outbreaks, which provided the number of at-risk farms and barns that needed to234 be sampled. Therefore, our results include 1) the total and daily number of sample collectors,235 2) the number of days sampling was delayed, 3) the number of samples waiting to undergo236 processing at the laboratory, and 4) the number of samples that needed to be redistributed237 to other laboratories to prevent processing delays.238 9 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint Figure 4: ASFV outbreak simulations. A) Number of farms sampled per day; and B) number of barns sampled per day. The blue solid line represents the median value and the orange dashed line indicates the maximum values. Results239 African swine fever simulated epidemics240 Our model simulated the spread of ASFV for 150 days. The summary of model simu-241 lations shows a median of 0 (range = 0-28) secondary infection detected daily, culminating242 in 27 outbreaks (range = 1-68) throughout the simulation period. In 26% of simulations,243 infection continued until the last day of the simulation, indicating that the ASFV epidemic244 was not completely eliminated. The median sampling duration in simulations, defined as245 the time between the first sampling event and the last sampling event, was 54 (range =246 1-135) days. This duration represents the sampling under optimal conditions, where farms247 were sampled without any delays, and it is used as a baseline to compare our results. The248 daily number of farms sampled peaked at 99 (range = 1-291) on day 39 (range = 37-78),249 while over the 150-day timeframe, a median of 23 (range = 1-516) farms were sampled per250 day (Figure 4A). By day 54 (median sampling duration) of the sampling, the cumulative251 number of sampling events at the farm level was 420 (range = 1-635) which increased to252 616 (range = 1-15,011) by day 150. At the barn level (Figure 4B), a median of 107 (range253 = 1-2,312) barns were sampled per day, resulting in 2,173 (range = 7-2,975) cumulative254 sampling events at day 54, and 3,068 (range = 7-69,118) at day 150.255 Estimated number of sample collectors256 Thenumberofsamplecollectorsanddaysrequiredforsamplingthroughoutthesimulated257 ASFV epidemics were significantly impacted by the number of sampler collectors available,258 downtime, and the sample type (blood swabs vs oral fluids) (Tables 2 and 3.) Scenarios259 collecting blood samples involving 72 hours of downtime, and unlimited trained sample260 collectors proved to be the best, completing all sampling within a median of 54 days and 135261 days in the worst epidemic scenario. However, this efficiency required a significantly higher262 median number of sample collectors, ranging between 247 and 367, and in the worst scenario263 between 2,095 and 3,115 (Table 2.) Scenarios with the same downtime but limited to sample264 collectors available also completed sampling within a median of 54 days; however, the worst265 epidemic scenarios required from 139 to 292 days. Thus, in comparison to the maximum266 10 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint expected sampling duration of 135 days, the worst-case scenarios required an additional 4267 to 157 days to sample all at-risk farms. Increasing the regular working hours allowed per268 collector from eight to 12 hours decreased the median number of sample collectors needed269 by 33%. Similarly, reducing the number of sampler collectors sent per farm from five to270 three led to a 4% reduction in the median number of collectors required.271 In scenarios with 24 hours or no downtime and at least 1,000 trained sample collectors,272 sampling was completed within a median of 54 days and a maximum of 135 days in the worst273 epidemic scenarios (Table 2). However, with only 500 sample collectors, the sampling time274 exceeded 135 days in the worst epidemic scenarios, requiring an additional 23 days for the275 24-hour downtime scenario and one extra day for the no-downtime scenario. For scenarios276 with 24 hours of downtime, the median number of required sample collectors ranged from277 181 to 264, with the maximum ranging from 500 to 1,818. Scenarios with no downtime278 required a median of 136 to 228 sample collectors and a maximum of 500 to 1,433 in the279 worst epidemic scenarios (Table 2).280 In our alternative scenarios using oral fluids, sampling was completed within a median281 of 54 days and a maximum of 135 days in all tested scenarios (Table 3.) Additionally,282 the number of sample collectors needed for oral fluids compared with blood samples was283 notably lower (Tables 2 and 3.) Specifically, in scenarios with 72 hours of downtime, oral284 fluid sampling required a median between 51% and 65% fewer sample collectors, while in285 the worst epidemic scenario required between 0% and 66% fewer sample collectors (Table286 3.) In scenarios with 24 hours of downtime, the reduction in the median number of sample287 collectors ranged between 43% and 61%, and between 1% and 66% in the worst epidemic288 scenario. Finally, in scenarios without any downtime, the median reduction was between289 74% and 79%, and 66% and 79% in the worst epidemic scenario (Table 3.)290 Table 2: Number of sample collectors–median and min-max range–needed to collect 31 blood samples from each barn of at-risk farms in a simulated ASFV epidemic. Conditions 72 hours downtime 24 hours downtime 0 hours downtime SCPF TSC R WS Sample collectors Last sample day Sample collectors Last sample day Sample collectors Last sample day 5 ∞ 12 247 (1 - 2,095) 54 (1 - 135) 181 (1 - 1,206) 54 (1 - 135) 136 (1 - 833) 54 (1 - 135) 3 ∞ 12 238 (1 - 1,992) 54 (1 - 135) 176 (1 - 1,167) 54 (1 - 135) 137 (1 - 841) 54 (1 - 135) 5 ∞ 8 367 (1 - 3,115) 54 (1 - 135) 264 (1 - 1,818) 54 (1 - 135) 228 (1 - 1,433) 54 (1 - 135) 5 2,000 12 247 (1 - 1,959) 54 (1 - 139) 181 (1 - 1,206) 54 (1 - 135) 136 (1 - 833) 54 (1 - 135) 5 1,750 12 247 (1 - 1,750) 54 (1 - 146) 181 (1 - 1,206) 54 (1 - 135) 136 (1 - 833) 54 (1 - 135) 5 1,500 12 247 (1 - 1,500) 54 (1 - 154) 181 (1 - 1,192) 54 (1 - 135) 136 (1 - 833) 54 (1 - 135) 5 1,250 12 247 (1 - 1,250) 54 (1 - 163) 181 (1 - 1,138) 54 (1 - 135) 136 (1 - 832) 54 (1 - 135) 5 1,000 12 247 (1 - 1,000) 54 (1 - 180) 181 (1 - 1,000) 54 (1 - 135) 136 (1 - 830) 54 (1 - 135) 5 500 12 245 (1 - 500) 54 (1 - 292) 181 (1 - 500) 54 (1 - 158) 136 (1 - 500) 54 (1 - 136) The median and maximum demand for sample collectors gradually increased over the291 first week after the sampling started (Figure 5). The median daily growth rate of sample292 collectors for blood samples ranged between 5.1% and 34.4%, but in the worst epidemic293 scenario, it varied between 2.5% and 5.9%. For oral fluid samples, the median daily growth294 rate was between 8.1% and 182%, while in the worst epidemic scenario, it fluctuated between295 2.5% and 6.0%.296 An optimal ASFV response requires that at-risk farms be sampled on the same day297 they are identified to avoid delays. Among the scenarios evaluated, collecting blood samples298 11 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint Figure 5: Cumulative number of sample collectors needed during the simulated ASFV epidemic considering 31 blood or five oral fluid samples from each barn in at-risk farms. The x-axis terminates at the median sampling duration (54 days) where the total number of samplers plateaus. A) blood samples with 72 hours downtime, B) oral fluid samples with 72 hours downtime, C) blood samples with 24 hours downtime, D) oral fluid samples with 24 hours downtime, E) blood samples with 0 hours downtime, F) oral fluid samples with 0 hours downtime. SCPF = Sample collectors per farm, TSC = Trained sample collectors, RWS = Regular working schedule. 12 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint Table 3: Number of sample collectors–median and min-max range–needed to collect five oral fluid samples from each barn in at-risk farms in a simulated ASFV epidemic. Conditions 72 hours downtime 24 hours downtime 0 hours downtime SCPF TSC R WS Sample collectors Last sample day Sample collectors Last sample day Sample collectors Last sample day 5 ∞ 12 116 (1 - 882) 54 (1 - 135) 99 (1 - 573) 54 (1 - 135) 28 (1 - 172) 54 (1 - 135) 3 ∞ 12 116 (1 - 882) 54 (1 - 135) 99 (1 - 573) 54 (1 - 135) 35 (1 - 211) 54 (1 - 135) 5 ∞ 8 127 (1 - 1,047) 54 (1 - 135) 103 (1 - 619) 54 (1 - 135) 52 (1 - 339) 54 (1 - 135) 5 2,000 12 116 (1 - 882) 54 (1 - 135) 99 (1 - 573) 54 (1 - 135) 28 (1 - 172) 54 (1 - 135) 5 1,750 12 116 (1 - 882) 54 (1 - 135) 99 (1 - 573) 54 (1 - 135) 28 (1 - 172) 54 (1 - 135) 5 1,500 12 116 (1 - 882) 54 (1 - 135) 99 (1 - 573) 54 (1 - 135) 28 (1 - 172) 54 (1 - 135) 5 1,250 12 116 (1 - 882) 54 (1 - 135) 99 (1 - 573) 54 (1 - 135) 28 (1 - 172) 54 (1 - 135) 5 1,000 12 116 (1 - 866) 54 (1 - 135) 99 (1 - 573) 54 (1 - 135) 28 (1 - 172) 54 (1 - 135) 5 500 12 116 (1 - 500) 54 (1 - 155) 99 (1 - 496) 54 (1 - 135) 28 (1 - 172) 54 (1 - 135) with a 72-hour downtime resulted in significant sampling delays (Figure 6.) Specifically,299 the scenario exhibited an average delay of 23 days and a maximum delay of 247 days with300 500 trained sample collectors, marking it as the least effective. Conversely, scenarios with301 an unlimited number of trained sample collectors, a downtime of 24 or zero hours, or the302 collection of oral fluid samples demonstrated considerably lower sampling delays, with a303 median delay that ranged between zero and six days (Figure 6.)304 Estimating laboratory capacity requirements305 Among the different scenarios evaluated, the median number of blood samples received306 at the laboratory was 84,830 (range = 52 - 2,066,831) (Table 4), and a median of 1,530307 (range = 0-68,939) per day (Supplementary Material Figure S1.) Regardless of downtime,308 unprocessed individual blood samples began accumulating rapidly after the sampling was309 initiated, reaching a peak of 40,618 (range = 52 - 1,955,126) unprocessed samples (Supple-310 mentary Material Figure S2-S37.) Over time, the unprocessed samples gradually reduced,311 yet processing all samples required a median of 92 days and, in the worst-case scenario,312 more than 5.7 years. Implementing sample pooling by five and 10 significantly reduced the313 processing time. For pools of five samples, the processing period was shortened to a median314 of 56 days and 1.2 years in the worst scenario (Table 4), a reduction of 39% and 79% in time315 compared to individual sample processing, respectively. For pools of 10 samples, the me-316 dian time required was 55 days, reducing the processing time by 40% compared to scenarios317 without pooling. In the worst scenario, the median time required for pools of 10 samples318 ranged from 234 days to 293 days (Table 4), reducing the processing time between 85% and319 88% compared to scenarios without pooling.320 The alternative scenario of collecting oral fluid samples showed that the laboratory re-321 ceived a median of 14,195 (range = 10 - 345,590) samples (Table 4), with a median of 255322 samples per day (Supplementary Material Figure S1.) Independent of the conditions eval-323 uated, the time required to process these samples was 55 days and one year in the worst324 scenario, reducing the processing time between 40% and 82% compared to individual blood325 samples, respectively.326 The median number of unprocessed single blood samples that the laboratory needed to327 redistribute was 10,062, given a daily capacity of 1,000 samples 7. This number decreases328 to 8,562 and 6,062 for increased capacities of 2,500 and 5,000 samples per day, respectively.329 13 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint Figure 6: Maximum number of days sampling is delayed during epidemic; SCPF = Sample collectors per farm, TSC = Trained sample collectors, RWS = Regular working schedule. A) blood samples with 72 hours downtime, B) oral fluid samples with 72 hours downtime, C) blood samples with 24 hours downtime, D) oral fluid samples with 24 hours downtime, E) blood samples with 0 hours downtime, F) oral fluid samples with 0 hours downtime. 14 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint Table 4: Number of samples–median and min-max range–arrived at the laboratory and needed to be pro- cessed in a simulated ASFV epidemic. Conditions 72 hours downtime 24 hours downtime 0 hours downtime Sample type and number SCPF TSC R WS Last processing day Last processing day Last processing day P ool of 1 blood sample Total number of processed samples: 84,830(52 - 2,066,831) 5 ∞ 12 92(2 - 2,070) 92(2 - 2,070) 92(2 - 2,070) 3 ∞ 12 92(2 - 2,070) 92(2 - 2,070) 92(2 - 2,070) 5 ∞ 8 92(2 - 2,070) 92(2 - 2,070) 92(2 - 2,070) 5 2,000 12 92(2 - 2,070) 92(2 - 2,070) 92(2 - 2,070) 5 1,750 12 92(2 - 2,070) 92(2 - 2,070) 92(2 - 2,070) 5 1,500 12 92(2 - 2,070) 92(2 - 2,070) 92(2 - 2,070) 5 1,250 12 92(2 - 2,070) 92(2 - 2,070) 92(2 - 2,070) 5 1,000 12 92(2 - 2,070) 92(2 - 2,070) 92(2 - 2,070) 5 500 12 92(2 - 2,070) 92(2 - 2,070) 92(2 - 2,070) P ool of 5 blood samples Total number of processed samples: 16,966(10 - 413,366) 5 ∞ 12 56(2 - 432) 56(2 - 432) 56(2 - 432) 3 ∞ 12 56(2 - 432) 56(2 - 432) 56(2 - 432) 5 ∞ 8 56(2 - 432) 56(2 - 432) 56(2 - 432) 5 2,000 12 56(2 - 432) 56(2 - 432) 56(2 - 432) 5 1,750 12 56(2 - 432) 56(2 - 432) 56(2 - 432) 5 1,500 12 56(2 - 432) 56(2 - 432) 56(2 - 432) 5 1,250 12 56(2 - 432) 56(2 - 432) 56(2 - 432) 5 1,000 12 56(2 - 432) 56(2 - 432) 56(2 - 432) 5 500 12 56(2 - 432) 56(2 - 432) 56(2 - 432) P ool of 10 blood samples Total number of processed samples: 8,483(5 - 206,683) 5 ∞ 12 55(2 - 234) 55(2 - 234) 55(2 - 234) 3 ∞ 12 55(2 - 234) 55(2 - 234) 55(2 - 234) 5 ∞ 8 55(2 - 234) 55(2 - 234) 55(2 - 234) 5 2,000 12 55(2 - 234) 55(2 - 234) 55(2 - 234) 5 1,750 12 55(2 - 234) 55(2 - 234) 55(2 - 234) 5 1,500 12 55(2 - 234) 55(2 - 234) 55(2 - 234) 5 1,250 12 55(2 - 234) 55(2 - 234) 55(2 - 234) 5 1,000 12 55(2 - 234) 55(2 - 234) 55(2 - 234) 5 500 12 55(2 - 293) 55(2 - 234) 55(2 - 234) P ool of 1 oral fluid sample Total number of processed samples: 14,195(10 - 345,590) 5 ∞ 12 55(2 - 366) 55(2 - 366) 55(2 - 366) 3 ∞ 12 55(2 - 366) 55(2 - 366) 55(2 - 366) 5 ∞ 8 55(2 - 366) 55(2 - 366) 55(2 - 366) 5 2,000 12 55(2 - 366) 55(2 - 366) 55(2 - 366) 5 1,750 12 55(2 - 366) 55(2 - 366) 55(2 - 366) 5 1,500 12 55(2 - 366) 55(2 - 366) 55(2 - 366) 5 1,250 12 55(2 - 366) 55(2 - 366) 55(2 - 366) 5 1,000 12 55(2 - 366) 55(2 - 366) 55(2 - 366) 5 500 12 55(2 - 366) 55(2 - 366) 55(2 - 366) On the days requiring the most redistribution—top 25%—the range of samples varied from330 19,434 to 67,940 with a capacity of 1,000 samples per day, from 17,934 to 66,440 with331 a capacity of 2,500 samples per day, and from 15,434 to 63,490 with a capacity of 5,000332 samples per day. Pooling samples remarkably reduces the need for redistribution. With the333 three different laboratory capacities considered, pooling five samples decreased the median334 daily redistributed samples by 88% to 100% and the maximum by 83% to 91%. Similarly,335 a pool of 10 samples reduced the median daily redistributed samples by 99% to 100% and336 the maximum by 91% to 97%. With a laboratory capacity of 1,000 samples per day, the337 median number of total samples that needed redistribution ranged from 118 to 57,826,338 while the maximum ranged from 115,783 to 1,958,888 (Supplementary Material Figure S38).339 Increasing the laboratory capacity to 2,500 decreased the median values by 43% and 100%,340 15 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint and further increasing the capacity to 5,000 reduced the median values by 75% and 100%.341 Figure 7: Daily number of samples required to be distributed to other laboratories in a sampling scenario using 72 hours downtime with 3 SCPF, 12 RWS, and no limit on TSC. SCPF = Sample collectors per farm, TSC = Trained sample collectors, RWS = Regular working schedule. Discussion342 Our results reveal that the number of sample collectors required in an ASFV outbreak343 was mostly impacted by the epidemic size, downtime requirements, and sample type blood344 versus oral fluids. For the study area, having 238 (range = 1-1,992) sample collectors was345 estimated to be sufficient to complete sampling without delays. Particularly, scenarios with346 no downtime or 24-hour downtime between sampling exhibited the lowest testing delays,347 highlighting the benefit of continuous operational capacity in surveillance systems but also348 introducing the risk of further dissemination as multiple farms were visited by the same349 collector. Additionally, oral fluid sampling required fewer collectors than blood samples.350 Oral fluid samples not only required fewer resources in terms of sample collectors but also351 contributed to shorter laboratory processing time (Goonewardene et al., 2021b; Mur et al.,352 2013). This suggests that integrating less conventional sampling techniques could enhance353 surveillance efficiency, especially during large-scale outbreaks where traditional resources354 are stretched thin. While it is unrealistic to expect that oral fluids could be used alone355 to address testing needs, they could become important tools in some use cases or could356 be used as complementary surveillance tools to conventional sampling techniques. The357 16 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint significant backlog of unprocessed samples in the laboratory, even with increased pooling and358 processing capacity, further illustrates the need for improvements in laboratory processing359 capabilities and or complement with additional strategies, such as oral fluid samples or360 sample redistribution to other NAHLN laboratories (Goonewardene et al., 2021b; USDA,361 2024). Therefore, these findings advocate for an integrated approach to disease surveillance362 that enhances both field sample collection and laboratory processing capabilities, ensuring363 rapid response to outbreaks and effective disease management for future ASFV outbreaks364 or similar animal health emergencies in the U.S.365 A large number of farms were identified as at-risk during the simulated ASFV epidemic,366 primarily due to the high density of farms in the studied regions—a median of 48 (range =367 0-128) farms within a 10 km radius of one another (Supplementary Material Figure S39). As368 a result, many farms were within the control zones (infected, buffer, and surveillance areas)369 (USDA, 2023). Furthermore, the frequent movement of vehicles, especially those transport-370 ing feed and previously in contact with infected farms, further increased the number of371 at-risk farms (Galvis and Machado, 2024a). As a result, a large number of samples need to372 be collected, which impacts the demand for sample collection, necessitating a larger work-373 force of trained sample collectors and extending laboratory processing times. Alternatives374 to reduce the number of samples could include, for example, the use of oral fluids, which375 would reduce the number of farm visits, which is a concern for breeding farms with high376 biosecurity levels (Sykes et al., 2022; Harlow et al., 2024; Campler et al., 2024; Alarcón et al.,377 2021). Reducing the number of vehicle visits would reduce the number of samples; thus,378 implementing stringent policies on vehicle movements should ensure vehicles are decontam-379 inated between farm visits and thus represent a low risk of disease transmission (Galvis and380 Machado, 2024a). However, this strategy hinges on the effectiveness of cleaning and disin-381 fection protocols, which are not always reliable (Boniotti et al., 2018; Mannion et al., 2008).382 Another proactive measure could involve rerouting vehicle movements during the epidemic383 based on the ASFV status of farms, to minimize contact with infected sites (Galvis and384 Machado, 2024b). These measures could significantly reduce the number of at-risk farms385 and lessen the burden of sample collection.386 The ASFV response scenario, which involves collecting blood samples from 31 animals387 per barn as the preferred sampling schema and implementing a 72-hour downtime between388 farm visits presents numerous logistical challenges. We demonstrated that in the worst-case389 scenarios, these requirements could delay farm sampling by up to 247 days (Figure 6). As390 a reflection, our findings suggested the need for the study region of 1,992 trained sample391 collectors (Table 2), which is indeed a large number and is expected to be costly for both392 the swine industry and the animal health officials who are in charge of certifying sampler393 collectors (Secure Pork Supply Plan, 2024). To reduce the number of sample collectors while394 maintaining efficient sampling times, our model shows that shorter downtime would have a395 notable benefit in completing sampling without delays. For instance, a downtime of 24 hours396 would necessitate between 8% and 49% fewer collectors, and eliminating downtime would397 require between 28% and 75% fewer collectors (Table 2). However, reducing or eliminating398 downtime requirements may increase the probability of between-farm disease transmission399 due to human-mediated transmission (Bellini et al., 2021). Another alternative solution400 17 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint could be to switch to oral fluid sampling, which requires fewer samples per farm and can401 be conducted more rapidly than blood sampling. For oral fluid sampling, the number of402 necessary collectors was reduced from 47% to 75% with a 72-hour downtime, from 69%403 to 75% with a 24-hour downtime, and from 83% to 91% with no downtime (Table 3). In404 addition to downtime and sample type, working schedules including a maximum of three405 sample collectors per farm with extended working hours—from eight to 12 hours—proved406 more effective, reducing collectors between 3% and 33% (Table 2). These alternatives not407 only streamline operations but could also ensure a faster, more cost-effective approach to408 managing ASFV outbreaks, ultimately enhancing our ability to control the spread of the409 virus promptly.410 Experimental studies have demonstrated the effectiveness of collecting oral fluid sam-411 ples via ropes in pens for identifying ASFV cases within 3-5 days of virus introduction412 (Goonewardene et al., 2021b). However, this method has noted limitations, such as variable413 interaction of pigs with the ropes limiting the ability to demonstrate if all pigs in a pen con-414 tributed to the sample (Grau et al., 2015; Guinat et al., 2014; Goonewardene et al., 2021b),415 environmental contamination that can skew results, false negatives before or within the first416 3–5 days after exposure and limited field validation studies that affect the reliability of this417 approach (Goonewardene et al., 2021b). Therefore, while oral fluid sampling is a promising418 sampling method, it should be used cautiously and ideally complemented by other sampling419 strategies until further clinical trials provide support for its use alone.420 Our findings reveal that the laboratory’s sample processing capacity cap leads to po-421 tential processing delays of up to 5.7 years for blood samples and up to one year for oral422 fluid samples without pooling. These results highlight the necessity of sample pooling to423 prevent delays, which can be conducted either in the laboratory or field as outlined in the424 certified swine sample collector program (Secure Pork Supply Plan, 2024). For instance, in425 our scenario with a pool of 10 samples, we observed a median processing delay of zero days.426 However, this strategy did not consistently ensure timely sample processing in the worst427 epidemic scenarios, leading to delays in ASFV detection and control measures. Thus, a428 pooling size of 10 samples may not be universally sufficient for all ASFV epidemic scenarios.429 Our results also suggest that the size of the sample pool should be adjusted based on the430 scale and timing of the epidemic and the number of at-risk farms throughout its course,431 starting with pools of one to five samples and increasing as necessary. However, it should be432 considered that larger pools are more likely to dilute the virus genetic material dramatically,433 impacting the performance of the test and thus the ability to accurately detect disease (Aira434 et al., 2019; Pikalo et al., 2021; López et al., 2021; Vilalta et al., 2019; Gallardo et al., 2019).435 An alternative approach could involve enhancing the laboratory’s daily processing capacity436 and redistributing unprocessed samples to other NAHLN laboratories (USDA, 2024). We437 demonstrated that in scenarios without pooling, the laboratory could reach its processing438 capacity from day one (Supplementary Material Figures S40 and S41), requiring a daily439 distribution of 10,062 (range = 2-67,940) samples (Figure 7). Such a large volume may440 potentially overwhelm the processing capacities of these other laboratories, which may also441 be receiving and processing samples from other states during an ASFV epidemic in the442 U.S. Additionally, assuming all other NAHLN laboratories have a similar capacity of 1,000443 18 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint samples per day, 50% of cases would require collaboration with up to 10 NAHLN labora-444 tories (Supplementary Material Figures S40), while in the worst epidemic scenario would445 require up to 50 laboratories (Supplementary Material Figures S41). Ideally, raising the446 daily laboratory capacity to 5,000 samples reduces the number of daily samples needing447 redistribution to 0 (range = 0-8,788) for pools of five samples and 0 (range = 0-1,894) for448 pools of 10 samples (7). Similarly, pooling samples also reduces the number of NAHLN449 laboratories needed during an ASFV epidemic (Supplementary Material Figures S42 - S45).450 Finally, portable devices capable of detecting ASFV directly on farms, despite potentially451 increasing processing costs, could significantly reduce sample processing delays (Chen et al.,452 2021; Ngan et al., 2023). Therefore, subsequent studies should include cost-benefit analyses453 to evaluate the effectiveness of these devices as part of an ASFV detection strategy while454 accounting for potential proficiency testing challenges with on-farm sample collectors and455 variations in diagnostic test performance.456

and further remarks457 This study has several limitations that require careful consideration when interpreting458 results. The ASFV model employed was calibrated using data from a previous epidemic of459 porcine epidemic diarrhea virus in the U.S., aimed at robustly predicting the spread of a460 new disease in a susceptible population. Despite this, the model may not accurately reflect461 the actual number of ASFV outbreaks; it could either overestimate or underestimate the462 number of at-risk farms and, consequently, the number of samples and sample collectors463 required (Sykes et al., 2023). We developed an algorithm to emulate real-world conditions464 and closely estimate the needs for sample collectors and laboratory capacity. However, we465 recognize that some conditions assumed in our methodology may differ during an actual466 ASFV outbreak, potentially altering the number of sample collectors needed and sample467 processing time. Additionally, we excluded several key variables that could impact the re-468 sults. These include the use of auxiliary drop-off locations to reduce travel times for sample469 collectors, modeling samples redistributed to certified laboratories from the NAHLN includ-470 ing their processing capacities and locations, and a validation period during which farms,471 once tested, are not required to be retested for a specified number of days. Furthermore, we472 omitted sampling strategies that involve on-site trained farm workers, such as caretakers,473 despite their training being a key objective of the certified swine sample collector program to474 meet the high demand for manpower during an epidemic (Secure Pork Supply Plan, 2024).475 By enabling farm workers to collect samples on-site, the need for additional travel and exter-476 nal sample collectors is reduced, thereby streamlining the sampling process and saving time.477 Incorporating these scenarios into future studies will yield more accurate results regarding478 the sampling process.479 Another significant limitation is that the ASFV model was not run simultaneously with480 the sampling estimation algorithm. This absence of integration directly affects the trans-481 mission dynamics of ASFV, as delays in sampling can decrease the number of infected farms482 detected by active surveillance during an outbreak. Consequently, the effectiveness of control483 measures is compromised, allowing infected farms to spread the virus over longer periods484 and thereby extending the time required to eradicate the disease. Thus, a future approach485 19 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint requires an integration of these two models to provide more valuable results for the ASFV486 surveillance, control, and eradication plans.487 Although the oral fluid sampling strategy demonstrated promising results, it alone is488 not considered in the current ASFV response plan (USDA, 2023). Therefore, it is necessary489 to model more realistic scenarios that combine oral fluid and blood sampling strategies490 across various use cases throughout the simulated epidemic. It is important to note that491 our study does not account for the sampling of wild boars, which play a crucial role in the492 spread of ASFV (Sauter-Louis et al., 2021; Sánchez-Cordón et al., 2019). Including wild493 boar sampling would require estimating sample sizes and wild boar locations, as well as494 employing personnel specifically trained for wild boar sampling, which differs from farm495 sampling techniques (Engeman et al., 2013). Additional sampling strategies should also496 consider fomites and vehicles that have been exposed, as they can serve as long-term virus497 reservoirs (Gebhardt et al., 2022). Moreover, the barn-level sampling strategy presents498 limitations, particularly in barns with multiple rooms. The heterogeneous distribution of499 infected animals across different rooms could increase the likelihood of false negative results500 due to inadequate sampling (Lamberga et al., 2022). Therefore, it is crucial to compare barn501 and room-level sampling strategies to determine the most effective approach for identifying502 ASFV-positive farms. Our results highlight the challenges sampling strategies would face503 should ASFV be introduced into the country. Addressing these challenges necessitates the504 development of additional strategies to optimize resource utilization. Consequently, our505 future efforts will focus on enhancing the sample collector allocation process through the506 use of advanced optimization methods, including deep reinforcement learning algorithms to507 further extend the current sampling and diagnostic capacities, utility attempting to identify508 the best set of tactics capable of fulfilling the needed actions in stamping out any ASFV509 incursion.510 Our study remarks on the importance of optimizing sample collector allocation and lab-511 oratory capacities for efficient and strategic resource management in the surveillance and512 control of ASFV. Similarly, we emphasize the certified swine sample collector program’s513 crucial role in training individuals for effective sample handling during large-scale epidemics514 and encourage regulatory agencies and the swine industry to combine efforts to ensure that515 sample collectors can continue to be trained, certified, and tracked for ready mobilization516 (Secure Pork Supply Plan, 2024). Efficient sample collection is pivotal not only for early de-517 tection but also for the effective containment and mitigation of outbreaks. As demonstrated518 by our results, improving the precision and responsiveness of sample collector deployment519 can significantly strengthen disease surveillance systems and expedite emergency responses.520 This becomes even more crucial in scenarios where rapid response and adaptability are nec-521 essary to manage dynamic and potentially widespread epidemics. Ultimately, investing in522 and refining these aspects of surveillance infrastructure are vital steps toward safeguarding523 animal health and, by extension, related agricultural and economic sectors.524 20 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint Conclusion525 We developed a strategic sampling model tailored to real-world conditions to accurately526 estimate the required number of sample collectors and necessary laboratory capacities in527 response to an ASFV introduction in the U.S. Our findings indicate that the number of528 samplers varies significantly mainly due to the size a possible epidemic, the sampling strate-529 gies employed and sample collector downtime. Depending on the sampling conditions, 238530 sample collectors are enough in 50% of the ASFV epidemic scenarios, but in the worst-531 case scenarios, 1,992 sample collectors are needed to complete sampling the same day the532 sample is required. However, it can be reduced between 28% and 75% with more flexible533 conditions that eliminate downtime or between 47% and 75% when compared to oral fluid534 sample collection. Our study also highlighted significant challenges in processing samples535 when laboratory capacity was capped at 1,000 samples per day, with processing times ex-536 tending over several years when not pooling. This underscores the urgent need for sample537 pooling, increasing laboratory capacities, or redistributing unprocessed samples to USDA-538 certified laboratories from the NAHLN. Therefore, it is crucial to revisit and refine ASFV539 sampling strategies to ensure efficient resource allocation and enhance disease preparedness540 and management capabilities for potential ASFV outbreaks.541 Acknowledgments542 The authors would like to acknowledge the valuable advice provided by participating543 companies and members of the Animal and Plant Health Inspection Service (APHIS): Barb544 Porter Spalding, Columb Rigney, Holden Hutchinson, Lindsey Holstrom, Lisa Rochette,545 Lydia Carpenter, and Sasidhar Malladi.546 Authors’ contributions547 JAG, DR, and GM conceived the study. JAG, MYS, ALS, KCO, LR, DR, and GM548 participated in the study design. JAG and ALS prepared population and movement data549 and developed the initial modeling computer code adapted here for the ASF. ALS designed550 the ASFV model and simulated scenarios. MYS designed a sample collector model and551 computational analysis. JAG, MYS, and GM wrote and edited the manuscript. All authors552 discussed the results and critically reviewed the manuscript. GM secured the funding.553 Funding statement554 This project is funded by USDA’s Animal and Plant Health Inspection Service through555 the National Animal Disease Preparedness and Response Program via a cooperative agree-556 ment between the Animal and Plant Health Inspection Service (APHIS) Veterinary Services557 (VS) and North Carolina State University, USDA-APHIS Award: AP23VSSP0000C088.558 The findings and conclusions in this document are those of the author(s) and should not be559 construed to represent any official USDA or U.S. Government determination or policy.560 21 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint Data A vailability Statement561 The data supporting this study’s findings are not publicly available and are protected562 by confidential agreements; therefore, they are not available.563 References564 Aira, C., Ruiz, T., Dixon, L., Blome, S., Rueda, P., Sastre, P., 2019. Bead-Based Multiplex Assay for565 the Simultaneous Detection of Antibodies to African Swine Fever Virus and Classical Swine Fever Virus.566 Frontiers in Veterinary Science 6, 306. URL:https://www.frontiersin.org/article/10.3389/fvets.567 2019.00306/full, doi:10.3389/fvets.2019.00306.568 Alarcón, L.V., Allepuz, A., Mateu, E., 2021. Biosecurity in pig farms: a review. Porcine Health Man-569 agement 7, 5. URL: https://porcinehealthmanagement.biomedcentral.com/articles/10.1186/570 s40813-020-00181-z, doi:10.1186/s40813-020-00181-z.571 Alvarez, E., Bielska, I.A., Hopkins, S., Belal, A.A., Goldstein, D.M., Slick, J., Pavalagantharajah, S.,572 Wynfield, A., Dakey, S., Gedeon, M.C., Alam, E., Bouzanis, K., 2023. Limitations of COVID-19573 testing and case data for evidence-informed health policy and practice. Health Research Policy and574 Systems 21, 11. URL: https://health-policy-systems.biomedcentral.com/articles/10.1186/575 s12961-023-00963-1, doi:10.1186/s12961-023-00963-1.576 Bellini, S., Casadei, G., De Lorenzi, G., Tamba, M., 2021. A Review of Risk Factors of African Swine577 Fever Incursion in Pig Farming within the European Union Scenario. Pathogens 10, 84. URL:https:578 //www.mdpi.com/2076-0817/10/1/84, doi:10.3390/pathogens10010084.579 Blome, S., Franzke, K., Beer, M., 2020. African swine fever – A review of current knowledge. Virus580 Research 287, 198099. URL:https://linkinghub.elsevier.com/retrieve/pii/S0168170220304019,581 doi:10.1016/j.virusres.2020.198099.582 Boniotti, M.B., Papetti, A., Bertasio, C., Giacomini, E., Lazzaro, M., Cerioli, M., Faccini, S., Bonilauri,583 P., Vezzoli, F., Lavazza, A., Alborali, G.L., 2018. Porcine Epidemic Diarrhoea Virus in Italy: Disease584 spread and the role of transportation. Transboundary and Emerging Diseases 65, 1935–1942. URL:585 https://onlinelibrary.wiley.com/doi/10.1111/tbed.12974, doi:10.1111/tbed.12974.586 Campler, M.R., Hall, M., Mills, K., Galvis, J.A., Machado, G., Arruda, A.G., 2024. Description of swine587 producerbiosecurityplanning forforeign animaldiseasepreparedness using theSecure Pork Supplyframe-588 work. Frontiers in Veterinary Science 11, 1380623. URL: https://www.frontiersin.org/articles/589 10.3389/fvets.2024.1380623/full, doi:10.3389/fvets.2024.1380623.590 Carriquiry, M., Elobeid, A., Swenson, D., Hayes, D., 2020. Impacts of African Swine Fever in Iowa and591 the United States. Technical Report. URL:https://www.card.iastate.edu/products/publications/592 pdf/20wp600.pdf.593 Center for Food Security and Public Health, 2017. INFORMATION MANUAL FOR ENHANCED594 BIOSECURITY FOR PORK PRODUCTION: ANIMALS RAISED INDOORS. URL: https://www.595 securepork.org/pork-producers/biosecurity/.596 Checkoff, P., 2023. U.S. Pork Exports. URL: https://www.porkcheckoff.org/markets/597 us-pork-exports/.598 Chen, L., Wen, K., Chen, F.E., Trick, A.Y., Liu, H., Shao, S., Yu, W., Hsieh, K., Wang, Z., Shen, J., Wang,599 T.H., 2021. Portable Magnetofluidic Device for Point-of-Need Detection of African Swine Fever. Analyt-600 ical Chemistry 93, 10940–10946. URL:https://pubs.acs.org/doi/10.1021/acs.analchem.1c01814,601 doi:10.1021/acs.analchem.1c01814.602 Cochran, H.J., Bosco-Lauth, A.M., Garry, F.B., Roman-Muniz, I.N., Martin, J.N., 2023. African Swine603 Fever: A Review of Current Disease Management Strategies and Risks Associated with Exhibition Swine604 in the United States. Animals 13, 3713. URL:https://www.mdpi.com/2076-2615/13/23/3713, doi:10.605 3390/ani13233713.606 Cwynar, P., Stojkov, J., Wlazlak, K., 2019. African Swine Fever Status in Europe. Viruses 11, 310. URL:607 https://www.mdpi.com/1999-4915/11/4/310, doi:10.3390/v11040310.608 22 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint Dixon, L., Sun, H., Roberts, H., 2019. African swine fever. Antiviral Research 165, 34–41. URL:609 https://linkinghub.elsevier.com/retrieve/pii/S0166354219300968, doi: 10.1016/j.antiviral.610 2019.02.018.611 Engeman, R.M., Massei, G., Sage, M., Gentle, M.N., 2013. Monitoring wild pig populations: a review of612 methods. Environmental Science and Pollution Research 20, 8077–8091. URL:http://link.springer.613 com/10.1007/s11356-013-2002-5, doi:10.1007/s11356-013-2002-5.614 Gallardo, C., Fernández-Pinero, J., Arias, M., 2019. African swine fever (ASF) diagnosis, an essential tool in615 the epidemiological investigation. Virus Research 271, 197676. URL:https://linkinghub.elsevier.616 com/retrieve/pii/S0168170219303843, doi:10.1016/j.virusres.2019.197676.617 Galvis, J.A., Machado, G., 2024a. Theroleofvehiclemovementinswinediseasedissemination: novelmethod618 accounting for pathogen stability and vehicle cleaning effectiveness uncertainties. Preventive Veteri-619 nary Medicine , 106168URL:https://linkinghub.elsevier.com/retrieve/pii/S0167587724000540,620 doi:10.1016/j.prevetmed.2024.106168.621 Galvis, J.O.A., Machado, G., 2024b. Mitigating between-farm disease transmission through simulat-622 ing vehicle rerouting and enhanced cleaning and disinfection protocols URL:https://osf.io/576qb/,623 doi:10.17605/OSF.IO/576QB. publisher: [object Object].624 Gebhardt, J.T., Dritz, S.S., Elijah, C.G., Jones, C.K., Paulk, C.B., Woodworth, J.C., 2022. Sampling625 and detection of African swine fever virus within a feed manufacturing and swine production system.626 Transboundary and Emerging Diseases 69, 103–114. URL:https://onlinelibrary.wiley.com/doi/627 10.1111/tbed.14335, doi:10.1111/tbed.14335.628 Goonewardene, K.B., Chung, C.J., Goolia, M., Blakemore, L., Fabian, A., Mohamed, F., Nfon, C., Clavijo,629 A., Dodd, K.A., Ambagala, A., 2021a. Evaluation of oral fluid as an aggregate sample for early detection630 of african swine fever virus using four independent pen-based experimental studies. Transboundary and631 emerging diseases 68, 2867–2877.632 Goonewardene, K.B., Chung, C.J., Goolia, M., Blakemore, L., Fabian, A., Mohamed, F., Nfon, C., Clavijo,633 A., Dodd, K.A., Ambagala, A., 2021b. Evaluation of oral fluid as an aggregate sample for early detection634 of African swine fever virus using four independent pen-based experimental studies. Transboundary635 and Emerging Diseases 68, 2867–2877. URL:https://onlinelibrary.wiley.com/doi/10.1111/tbed.636 14175, doi:10.1111/tbed.14175.637 Grau, F.R., Schroeder, M.E., Mulhern, E.L., McIntosh, M.T., Bounpheng, M.A., 2015. Detection of638 African swine fever, classical swine fever, and foot-and-mouth disease viruses in swine oral fluids by639 multiplex reverse transcription real-time polymerase chain reaction. Journal of Veterinary Diagnos-640 tic Investigation 27, 140–149. URL:http://journals.sagepub.com/doi/10.1177/1040638715574768,641 doi:10.1177/1040638715574768.642 Guinat, C., Reis, A.L., Netherton, C.L., Goatley, L., Pfeiffer, D.U., Dixon, L., 2014. Dynamics of African643 swine fever virus shedding and excretion in domestic pigs infected by intramuscular inoculation and con-644 tact transmission. Veterinary Research 45, 93. URL:http://www.veterinaryresearch.org/content/645 45/1/93, doi:10.1186/s13567-014-0093-8.646 Harlow, M., Torremorell, M., Rademacher, C.J., Gebhardt, J., Holck, T., Linhares, L.C.M., Main, R.G.,647 Trevisan, G., 2024. Biosecurity Insights from the United States Swine Health Improvement Plan: Analyz-648 ing Data to Enhance Industry Practices. Animals 14, 1134. URL:https://www.mdpi.com/2076-2615/649 14/7/1134, doi:10.3390/ani14071134.650 Hayes, B.H., Andraud, M., Salazar, L.G., Rose, N., Vergne, T., 2021. Mechanistic modelling of African swine651 fever: A systematic review. Preventive Veterinary Medicine 191, 105358. URL:https://linkinghub.652 elsevier.com/retrieve/pii/S0167587721001021, doi:10.1016/j.prevetmed.2021.105358.653 Hernando, L., Mendiburu, A., Lozano, J.A., 2018. Hill-climbing algorithm: Let’s go for a walk before finding654 the optimum, in: 2018 IEEE Congress on Evolutionary Computation (CEC), pp. 1–7. doi:10.1109/CEC.655 2018.8477836.656 Jean-Pierre, R.P., Hagerman, A.D., Rich, K.M., 2022. An analysis of African Swine Fever conse-657 quences on rural economies and smallholder swine producers in Haiti. Frontiers in Veterinary Sci-658 ence 9, 960344. URL: https://www.frontiersin.org/articles/10.3389/fvets.2022.960344/full,659 23 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint doi:10.3389/fvets.2022.960344.660 Jungnickel, D., 1999. The Greedy Algorithm. Springer Berlin Heidelberg, Berlin, Heidelberg. chapter 5. pp.661 129–153. URL:https://doi.org/10.1007/978-3-662-03822-2_5, doi:10.1007/978-3-662-03822-2_662 5.663 Lamberga, K., Depner, K., Zani, L., Ol ,ševskis, E., Seržants, M., Ansonska, S., Šteingolde, , B¯ erzin,š,664 A., Viltrop, A., Blome, S., Globig, A., 2022. A practical guide for strategic and efficient sam-665 pling in African swine fever-affected pig farms. Transboundary and Emerging Diseases 69. URL:666 https://onlinelibrary.wiley.com/doi/10.1111/tbed.14582, doi:10.1111/tbed.14582.667 Luxen, D., Vetter, C., 2011. Real-time routing with openstreetmap data, in: Proceedings of the 19th ACM668 SIGSPATIAL International Conference on Advances in Geographic Information Systems, ACM, New669 York, NY, USA. pp. 513–516. URL:http://doi.acm.org/10.1145/2093973.2094062, doi:10.1145/670 2093973.2094062.671 López, W.A., Gauger, P.C., Harmon, K.M., Holtkamp, D.J., Cano, J.P., Macedo, N., Zhang, M., Silva,672 G.S., Angulo, J., Zimmerman, J.J., Linhares, D.C., 2021. Probability of PRRS virus detection in pooled673 processing fluid samples. Veterinary Microbiology 261, 109190. URL:https://linkinghub.elsevier.674 com/retrieve/pii/S0378113521002133, doi:10.1016/j.vetmic.2021.109190.675 Machado, G., Galvis, J., Freeman, A., Sanchez, F., Mills, K., Fleming, C., Hong, X., 2023. The Rapid Ac-676 cess Biosecurity (RAB) app™ Handbook URL: https://osf.io/z5wbj/, doi:10.17605/OSF.IO/Z5WBJ.677 publisher: Open Science Framework.678 Mannion, C., Egan, J., Lynch, B.P., Fanning, S., Leonard, N., 2008. An Investigation into the Effi-679 cacy of Washing Trucks Following the Transportation of Pigs—ASalmonella Perspective. Foodborne680 Pathogens and Disease 5, 261–271. URL:http://www.liebertpub.com/doi/10.1089/fpd.2007.0069,681 doi:10.1089/fpd.2007.0069.682 Mighell, E., Ward, M.P., 2021. African Swine Fever spread across Asia, 2018–2019. Transboundary683 and Emerging Diseases 68, 2722–2732. URL:https://onlinelibrary.wiley.com/doi/10.1111/tbed.684 14039, doi:10.1111/tbed.14039.685 Mur, L., Gallardo, C., Soler, A., Zimmermman, J., Pelayo, V., Nieto, R., Sánchez-Vizcaíno, J.M., Arias, M.,686 2013. Potential use of oral fluid samples for serological diagnosis of African swine fever. Veterinary Micro-687 biology 165, 135–139. URL:https://linkinghub.elsevier.com/retrieve/pii/S0378113513000436,688 doi:10.1016/j.vetmic.2012.12.034.689 Ngan, M.T., Thi My Le, H., Xuan Dang, V., Thi Bich Ngoc, T., Phan, L.V., Thi Hoa, N., Quang Lam,690 T., Thi Lan, N., Notsu, K., Sekiguchi, S., Yamazaki, Y., Yamazaki, W., 2023. Development of a highly691 sensitive point-of-care test for African swine fever that combines EZ-Fast DNA extraction with LAMP692 detection: Evaluation using naturally infected swine whole blood samples from Vietnam. Veterinary693 Medicine and Science 9, 1226–1233. URL: https://onlinelibrary.wiley.com/doi/10.1002/vms3.694 1124, doi:10.1002/vms3.1124.695 Njau, E.P., Machuka, E.M., Cleaveland, S., Shirima, G.M., Kusiluka, L.J., Okoth, E.A., Pelle, R., 2021.696 African Swine Fever Virus (ASFV): Biology, Genomics and Genotypes Circulating in Sub-Saharan Africa.697 Viruses 13, 2285. URL:https://www.mdpi.com/1999-4915/13/11/2285, doi:10.3390/v13112285.698 Pikalo, J., Deutschmann, P., Fischer, M., Roszyk, H., Beer, M., Blome, S., 2021. African Swine Fever699 Laboratory Diagnosis—Lessons Learned from Recent Animal Trials. Pathogens 10, 177. URL:https:700 //www.mdpi.com/2076-0817/10/2/177, doi:10.3390/pathogens10020177.701 Sauter-Louis, C., Conraths, F.J., Probst, C., Blohm, U., Schulz, K., Sehl, J., Fischer, M., Forth, J.H.,702 Zani, L., Depner, K., Mettenleiter, T.C., Beer, M., Blome, S., 2021. African Swine Fever in Wild703 Boar in Europe—A Review. Viruses 13, 1717. URL: https://www.mdpi.com/1999-4915/13/9/1717,704 doi:10.3390/v13091717.705 Schambow, R., Colin, Y., Dave, W., Schettino, D.N., Perez, A.M., 2022. Enhancing passive surveillance for706 African swine fever detection on U.S. swine farms. Frontiers in Veterinary Science 9, 1080150. URL:707 https://www.frontiersin.org/articles/10.3389/fvets.2022.1080150/full, doi:10.3389/fvets.708 2022.1080150.709 Schambow, R.A., Hussain, S., Antognoli, M.C., Kreindel, S., Reyes, R., Perez, A.M., 2023. Epidemiological710 24 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint Assessment of African Swine Fever Spread in the Dominican Republic. Pathogens 12, 1414. URL:711 https://www.mdpi.com/2076-0817/12/12/1414, doi:10.3390/pathogens12121414.712 Secure Pork Supply Plan, 2024. Certified Swine Sample Collector Training. URL: https://www.713 securepork.org/training-materials/disease-monitoring-sample/.714 Sykes, A.L., Galvis, J.A., O’Hara, K.C., Corzo, C., Machado, G., 2023. Estimating the effectiveness of715 control actions on African swine fever transmission in commercial swine populations in the United States.716 PreventiveVeterinaryMedicine217, 105962. URL: https://linkinghub.elsevier.com/retrieve/pii/717 S0167587723001265, doi:10.1016/j.prevetmed.2023.105962.718 Sykes, A.L., Silva, G.S., Holtkamp, D.J., Mauch, B.W., Osemeke, O., Linhares, D.C., Machado, G., 2022.719 Interpretable machine learning applied to on-farm biosecurity and porcine reproductive and respiratory720 syndrome virus. Transboundary and Emerging Diseases 69. URL:https://onlinelibrary.wiley.com/721 doi/10.1111/tbed.14369, doi:10.1111/tbed.14369.722 Sánchez-Cordón, Nunez, Neimanis, Wikström-Lassa, Montoya, Crooke, Gavier-Widén, 2019. African723 Swine Fever: Disease Dynamics in Wild Boar Experimentally Infected with ASFV Isolates Belong-724 ing to Genotype I and II. Viruses 11, 852. URL: https://www.mdpi.com/1999-4915/11/9/852,725 doi:10.3390/v11090852.726 Sánchez-Cordón, P., Montoya, M., Reis, A., Dixon, L., 2018. African swine fever: A re-emerging viral disease727 threatening the global pig industry. The Veterinary Journal 233, 41–48. URL:https://linkinghub.728 elsevier.com/retrieve/pii/S109002331730268X, doi:10.1016/j.tvjl.2017.12.025.729 USDA, 2020. African swine fever Response Plan: The Red Book. URL: https://ppainteragencial.730 iica.int/sites/default/files/2022-07/African%20Swine%20Fever%20response%20plan.%20The%731 20red%20Book..pdf.732 USDA, 2023. African swine fever Response Plan: The Red Book. URL:https://www.aphis.usda.gov/733 animal_health/emergency_management/downloads/asf-responseplan.pdf.734 USDA, 2024. National Animal Health Laboratory Network. URL:https://www.aphis.usda.gov/labs/735 nahln.736 Vilalta, C., Baker, J., Sanhueza, J., Murray, D., Sponheim, A., Alvarez, J., Sylvia, F., Polson, D., Tor-737 remorell, M., Corzo, C., Morrison, R.B., 2019. Effect of litter aggregation and pooling on detection of738 porcine reproductive and respiratory virus in piglet processing fluids. Journal of Veterinary Diagnos-739 tic Investigation 31, 625–628. URL:http://journals.sagepub.com/doi/10.1177/1040638719852999,740 doi:10.1177/1040638719852999.741 25 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 29, 2024. ; https://doi.org/10.1101/2024.09.26.615206doi: bioRxiv preprint