Date Full Report Received
Date Abstract Report Received
As a consequence of spatial disease transmission, the distribution of PRRSV will not be uniform, but rather more concentrated in the center of the barn (Figure 1). Given the heterogeneity among pens, it is clear that random sampling will not be the optimal approach. In other words, classical surveillance sampling designs, e.g. simple random sampling, do not account either for spatial correlation or for temporal relationships among pens’ disease status. Hence, we proposed the simulation-base Stepwise algorithm to solve these problems and to find the optimal sampling locations in an effective manner.
Firstly, PRRSV status and sample test outcomes were simulated based on the data-validated spatial piecewise exponential model. 1,000,000 datasets were generated to guarantee convergence and high precision. The optimal sampling locations are affected by multiple factors, including: (a) starting prevalence, (b) transmission rate, (c) sensitivity of the diagnostics, (d) sample time point, and (e) sample size limitation. Simulation studies of disease status and sample test outcomes were performed for each combination of factors (a), (b), (c), and (d). Next, based on the simulation results, exhaustive search for the highest power of detection among all unique samples was implemented. It turned out that the exhaustive method is extremely inefficient, hence replaced by the newly proposed Stepwise method. Staring from the given sampling locations, the Stepwise method optimizes the power of detection in an iterative way, thus it successfully avoids a large amount of computation. The simulation-based Stepwise algorithm makes it possible to take into account the spatial and temporal correlations occurred during the PRRSV transmission. It is also more efficient in time and computation compared with the exhaustive search.
Since the optimal sampling strategy is determined by the above-mentioned factors, it is not possible to report all the results in tables or figures of a manageable size. However, it is extremely important to provide clear answers to practitioner and others who need the information but may not be familiar with mathematical formulas. Therefore, we developed a user-friendly web interface (https://minz.shinyapps.io/sampling/) that displays the optimal locations to sample, given certain values of the factors, for convenient and free access to the outcome of this project.
The results of our study indicate that the improvement in the power of detection from random sampling ranges from 0% to 7.4%. The improvement is large when the sampling time point is during the mid-phase of the study (the third to fifth week) and when the sample size is small (no more than 10 pens). In practice, there is always a limitation of the sample size. Hence, our Stepwise sampling strategy is meaningful in terms of improving the power of detection for PRRSV with limited sample size.
Our research provides the producers and swine veterinarians with more efficient procedures for PRRSV detection and convenient access to the sampling results. Further application of our new algorithm to other swine pathogens will also improve the efficiency of the disease surveillance and diagnostic testing.
Dr. Chang Wang (Principle investigator)
3413 Snedecor Hall firstname.lastname@example.org (515) 294-8047
Min Zhang (Graduate student)
3414 Snedecor Hall email@example.com