Towards resilient beef cattle production systems: Impact of truck contamination and information sharing on foot-and-mouth disease spreading

As cattle movement data in the United States are scarce due to the absence of mandatory traceability programs, previous epidemic models for U.S. cattle production systems heavily rely on contact rates estimated based on expert opinions and survey data. These models are often based on static networks and ignore the sequence of movement, possibly overestimating the epidemic sizes. In this research, we adapt and employ an agent-based model that simulates beef cattle production and transportation in southwest Kansas to analyze the between-premises transmission of a highly contagious disease, the foot-and-mouth disease. First, we assess the impact of truck contamination on the disease transmission with the truck agent following an independent clean-infected-clean cycle. Second, we add an information-sharing functionality such that producers/packers can trace back and forward their trade records to inform their trade partners during outbreaks. Scenario analysis results show that including indirect contact routes between premises via truck movements can significantly increase the amplitude of disease spread, compared with equivalent scenarios that only consider animal movement. Mitigation strategies informed by information sharing can dramatically improve the system resilience against epidemics, highlighting the benefit of promoting information sharing in the cattle industry. In addition, we identify salient characteristics that must be considered when designing an information-sharing strategy, including the number of days to trace back and forward in the trade records and the role of different cattle supply chain stakeholders. Sensitivity analysis results show that epidemic sizes are sensitive to variations in parameters of fomite survival time and indirect contact transmission probability and future studies can focus on a more accurate estimation of these parameters.


Introduction
developed a model to heuristically generate a dynamic hog production system while 73 accounting for truck contamination, demonstrating that producer specialization can increase system 74 vulnerability to disease outbreaks. Bernini

104
Traceability programs, however, have become common in the global beef market, and lack of such 105 programs may decrease export markets for the U.S. beef industry [31]. For example, if 25% of beef products 106 became unacceptable in international trade, then the U.S. economy would experience an estimated $6.65 107 billion loss [32]. An improved information infrastructure with traceability systems would yield many 108 benefits, including targeted and timely product recalls after a foodborne illness outbreak and increased 109 brand value for products due to quality assurance. Several pilot projects have recently developed and tested 110 purpose-built cattle disease traceability infrastructures, such as BeefChain and CattleTrace. In these projects, 111 cattle movement information is uploaded to a secure third-party database through radio frequency 112 identification tags or similar devices. Since the acceptance of such traceability programs depends on the 113 trust among stakeholders, these pilot programs have focused on persuading operators to participate. New 114 technologies such as Blockchain, which has been increasingly studied [33][34][35][36], are expected to build trust 115 among food partners, promote livestock traceability, and enhance food safety.

135
This study does not mean to predict the spread of FMD in a real-world production system, but it 136 facilitates a realistic epidemiological model to highlight the impact of indirect contact through truck 137 movement and the potential benefits of information sharing to the system regarding disease transmission.

138
The findings are expected to benefit existing disaster preparedness and promote the development of new  [20] Probability that truck will be contaminated upon visiting an infected producer/packer 0.15 Probability that contaminated truck will infect the subsequent producer/packer it will visit 0.15 Probability that infected cattle will contaminate packer/producer receiving area 0.75 Number of days to trace back and forward during the information sharing process (days) 14 Assumed

Epidemic initialization
163 All the cattle in the system are in the susceptible state before the epidemic is initialized. For each 164 simulation run, one randomly selected animal from outside SW KS will enter the infectious state on the 9 th 165 day midnight, and will be brought to a stocker inside the region on the 10 th day. If there are no cattle at the 166 border at this time, the simulation will stop and start the next simulation run. In different simulation runs, 167 the recipient stocker of the first infectious animal will be different due to the stochasticity of the model.

173
(1) Cattle agent 174 Each cattle agent is associated with a Susceptible-Exposed-Infectious-Removed compartment model.

175
Compared to the cattle in the infectious state, cattle agents in the exposed state have been infected by the 176 FMD disease, but are not contagious yet. Animals in the removed state have been infected before, and 9 become dead or immune, or are culled as part of control strategies. More specifically, in contact with an 178 animal in the infectious state, a susceptible cattle agent has a 95% chance to become infected and transition 179 to the exposed state. After a latent period, the cattle agent will transition to the infectious state during which 180 it contacts and infects other cattle of the same premises. Then, the cattle agent enters the removed state after 181 an infectious period.

182
Considering computational efficiency, we use a scaling factor of 10 to change all the parameters related 183 to the number of cattle, e.g., truck capacity and cattle capacity of each producer, such that one cattle agent 184 represents ten cattle. We assume that each cattle agent contacts, on average, 20 cattle agents in a day. 20

188
(2) Producer agent 189 Once there are infectious cattle in the producer agent, the producer will transition from the clean state 190 to the infected state and the c_infected variable becomes true. After a cattle infectious period, the first 191 infectious cattle agent will become removed, and if control strategies have been implemented on the 192 producer premises level, the producer agent will enter the under-control state. In the under-control state, 193 various mitigation strategies are implemented including movement bans and information-sharing 194 functionality, and more details will be described in the scenario analysis section. When all infected cattle 195 of an infected producer enter the removed state, the c_infected variable is set to be false. On the other hand, 196 when a producer becomes infected by fomite, the variable f_infected becomes true and will last for 14 days 197 (fomite survival time in Table 1).

(3) Packer agent 199
Once a cattle agent in the infectious state arrives at the packer, the packer agent will transition to the 200 infected state (c_infected=true), and after a contamination period T, the packer goes to the under-control 201 state, in which the packer stops requesting and transporting cattle from other producers to its location. To 202 be more realistic, we assume that more infectious cattle arriving at the packer will speed up FMD detection.

203
For example, if there are infectious cattle coming in on day d1 and day d2, the model will generate two 204 numbers, ct and ct', respectively, according to the distribution with a mean of 5 days in Table 1

215
More specifically, when the truck arrives at the origin premises, if there are infectious cattle loaded to 216 the truck at the origin premises, then there is a 75% probability that the receiving area of the origin producer 217 will become infected via fomite (f_infected = true). If the origin producer is fomite-infected and the truck 218 is not contaminated, the truck may become contaminated with a 15% probability; however, if the truck is 219 contaminated and the origin producer is not fomite-infected (f_infected = false), the truck will cause the 220 producer to become fomite-infected with a 15% probability.

221
When the truck arrives at the destination premises, if there are infectious cattle unloaded, there is a 75% 222 probability that the destination premises will become fomite-infected. Meanwhile, if the destination is 223 fomite-infected and the truck is clean, there is a 15% probability that the truck may become contaminated.

224
However, if the truck is infected and the destination is not fomite-infected, the truck may cause the 225 destination to become fomite-infected. Note that at the time the origin or destination producer becomes 226 fomite-infected, we will randomly select one cattle agent from the premises to become infected.  Table 2, and all the parameters follow the values described in Table 1

301
More specifically, for scenarios 1-5, scenarios 2-5 result in a smaller and similar epidemic size 302 compared to scenario 1 (no premises-level control strategies implemented). For scenarios 6-10, scenario 6 303 (no premises-level control strategies implemented) results in the largest epidemic size, followed by scenario 304 7 (producer isolation), and then scenario 8 (information infrastructure enabled on producers), while 305 scenarios 9 (information infrastructure enabled on producers and packers) and 10 (regional movement ban) 306 result in the smallest and similar epidemic sizes. There is a slight difference in terms of the median size 307 between scenarios 6 and 7, indicating that merely implementing movement bans on infected producers 308 cannot effectively contain the epidemic, when indirect contact is considered. Information sharing can 309 significantly reduce the epidemic size, compared to scenario 6. Particularly, scenario 8 has a similar median 310 number of infected producers with scenarios 9 and 10 but has a larger interquartile range with 15 as the 311 third quartile in Fig. 3. This is because information-sharing functionality is not enabled on packers in 312 scenario 8, and these infected packers can spread the infection by their contaminated trucks to other 313 producers in some simulation runs.

326
In the cattle supply chain, cattle flows are essentially driven by the normal operation of packers to meet 327 steady demand, so the total number of cattle in the system during the simulation period is strongly affected 328 by the packers' operation time. With infected cattle coming in, the packer will go to the under control state 329 after a contamination period, and then will stop receiving cattle from feedlots both inside and outside the 330 region. As a result, feedlots will stop requesting cattle from other premises, which will impact the number 331 of cattle stocker operations will request. Therefore, the system's cattle flow quickly stops once packers are 332 under control, affecting the total number of cattle agents in the system. As the producers and packers gain 333 revenues by marketing their cattle or cattle-related products, the total number of cattle agents in the system, 334 which include the cattle that are marketed and being prepared to be marketed, can be seen as a measure for