Participatory regional surveillance to establish regional FAD status within the U.S. and improve swine movement decisions

Scientific Abstract

Implementing an effective, sustainable, and affordable surveillance system capable of early detection of transboundary pathogens is crucial for enabling a rapid and effective response.  Previously, it was shown that an active participatory design based on collecting and testing a few targeted samples from many participating farms, could achieve detection at a low prevalence and sustainable cost (Trevisan et al., 2024).  In this study, we expanded the concept to the 48 contiguous states. 

Using the 2017 USDA Census of Agriculture, we created a dataset of 66,637 farms in the contiguous 48 U.S. states (8,080,470 km2).  Each farm was identified by geolocation, production type, and inventory.  The spread of a “generic” pathogen within this region was simulated for 180 days over a range of spread parameter values (6,075 scenarios) using the Animal Disease Spread Model (ADSM) software (v3.5.10.0).  The output dataset for each scenario reported the infection status of each farm by day post-outbreak (DPO).  Using this dataset, an R function was used to calculate the probability (P) of detecting ≥ 1 positive farm by DPO as a function of the percentage (%) of herds participating in the surveillance program, site-level detection sensitivity, and disease prevalence.  The effect of sampling interval was evaluated by comparing P for a specific DPO vs the aggregate P (Pa) based on sampling every 14 days (Equation 1). 

(Equation 1) Pa = [1 − (1 − P₁₄) × (1 − P₂₈) × … (1 − P_DPO)]

Detection simulations revealed that the probability of detection increased with both higher producer participation and site-level detection sensitivity.  However, even with low site-level sensitivity, high detection probabilities at a low prevalence were achieved as producer participation increased.  Further, the results showed that sampling at 2-week intervals significantly improved detection at a low prevalence versus a single sampling at a specific DPO.  For example, given 40% producer participation (26,654 participating herds) and 10% site-level detection sensitivity, a single sampling event on DPO 42 (0.04% prevalence/29 infected herds) provided a 70% probability of detecting at least one positive herd.  In contrast, sampling at 2-week intervals (DPOs 14, 28, and 42) provided a 91% probability of detection at DPO 42. 

Thus, a surveillance design based on testing a few samples from poor-doing pigs in each of many herds can provide low-cost, highly sensitive detection.  It should be noted that this surveillance sampling design is similar to that proposed for the U.S. SHIP program.