Tracking animal movements over time may fundamentally determine the success of disease control interventions. In commercial pig production growth stages determine animal transportation schedule, thus it generates time‐varying contact networks showed to influence the dynamics of disease spread. In this study, we reconstructed pig networks of one Brazilian state from 2017 to 2018, comprising 351,519 movements and 48 million transported pigs. The static networks view did not capture time‐respecting movement pathways. For this reason, we propose a time‐dependent network approach. A susceptible‐infected model was used to spread an epidemic over the pig network globally through the temporal between‐farm networks, and locally by a stochastic model to account for within‐farm dynamics. We propagated disease to calculate the cumulative contacts as a proxy of epidemic sizes and evaluate the impact of network‐based disease control strategies in the absence of other intervention alternatives. The results show that targeting 1,000 farms ranked by degree would be sufficient and feasible to diminish disease spread considerably. Our modeling results indicated that independently from where initial infections were seeded (i.e., independent, commercial farms), the epidemic sizes and the number of farms needed to be targeted to effectively control disease spread were quite similar, indeed this finding can be explained by the presence of contact among all pig operation types The proposed strategy limited the transmission the total number of secondarily infected farms to 29, over two simulated years. The identified 1,000 farms would benefit from enhanced biosecurity plans and improved targeted surveillance. Overall, the modeling framework provides a parsimonious solution for targeted disease surveillance when temporal movement data is available.

Infectious diseases in livestock are well-known to infect multiple hosts and persist through a combination of within- and between-host transmission pathways. Uncertainty remains about the epidemic dynamics of diseases being introduced on farms with more than one susceptible host species. Here, we describe multi-host contact networks and elucidate the potential of disease spread through farms with multiple hosts. Four years of between-farm animal movement among all farms of a Brazilian state were described through a static and monthly snapshot of network representations. We developed a stochastic multilevel model to simulate scenarios in which infection was seeded into single host and multi-host farms to quantify disease spread potential, and simulate network-based control actions used to evaluate the reduction of secondarily infected farms. We showed that the swine network was more connected than cattle and small ruminants in both the static and monthly snapshots. The small ruminant network was highly fragmented, however, contributed to interconnecting farms, with other hosts acting as intermediaries throughout the networks. When a single host was initially infected, secondary infections were observed across farms with all other species. Our stochastic multi-host model demonstrated that targeting the top 3.25% of the farms ranked by degree reduced the number of secondarily infected farms. The results of the simulation highlight the importance of considering multi-host dynamics and contact networks while designing surveillance and preparedness control strategies against pathogens known to infect multiple species.

Post-mortem inspection (PMI) of routinely slaughtered cattle in abattoirs is an extremely valuable tool for detecting bovine tuberculosis (bTB) infected herds that can supplement active surveillance activities. However, its true performance is difficult to assess due to the multiple factors that may affect it. Here, we determined relative efficiencies in the detection of bTB-compatible lesions and probabilities of subsequent laboratory confirmation of abattoirs located in Castilla y Leon, one of the regions with the largest cattle population in Spain, between 2010 and 2017. The slaughtered animal population was split based on the results of the ante-mortem tests (reactors or non-reactors), and two generalized linear multivariable mixed models were fitted to each subpopulation to calculate the risk of lesion detection and laboratory confirmation per abattoir while accounting for the effect of potential confounding variables. Throughout the 8-year period, ∼30,000 reactors and >2.8 million non-reactor animals in the ante-mortem tests were culled in the abattoirs under study. Bovine TB compatible lesions were detected in 4,710 (16%) reactors and 828 (0.03%) non-reactor animals, of which >95% were confirmed as infected through bacteriology. The probability of disclosure of bTB-like lesions was associated with the animal subpopulation, type of source unit, the herd size, the year of slaughter, the breed and age of the animal, and/or the season of slaughter. The probabilities of detection of bTB-like lesions varied largely depending on the abattoir in both subpopulations, ranging from 603 to 3,070 per 10,000 animals for the reactors and 0.2-16.1 per 10,000 animals for the non-reactor animals. Results obtained here will help to quantify the performance of PMI in abattoirs in Castilla y Leon and the between-abattoir variability, and to identify animals at increased risk of having bTB-like lesions detected during PMI based on animal- and farm-related factors.