Author
![]() |
Hurd, Howard |
![]() |
SCHLOSSER, W - USDA-FSIS-OPHS |
![]() |
EBEL, E - USDA-FSIS-OPHS |
Submitted to: International Symposium on Epidemiology and Control of Salmonella in Pork
Publication Type: Abstract Only Publication Acceptance Date: 8/7/1999 Publication Date: N/A Citation: N/A Interpretive Summary: Technical Abstract: The objective of this study was to evaluate standard sample size recommendations for the prevalence estimation of intermittently shed and clustered infections, such as Salmonella. We developed a simple stochastic computer simulation model of 500 animals. Every iteration, each individual had a random chance of infection based on a set mean true prevalence (TP). The mean TP and its skewness were set at levels currently observed in U.S. finish hogs. Additionally, the possibility of clustering was included. Clustering reflects the expectation that infection is not homogeneously distributed among a population of 500 finish hogs. Instead, it may be grouped among pen mates or adjoining pens. We evaluated different clustering scenarios for each mean TP. At the beginning of the simulation, the sample size was set according to standard formulas. Each iteration, the model randomly tested a sample of animals. The model then estimated the apparent prevalence (AP) for that iteration. Test sensitivity and specificity were set to be 100%. Multiple simulations of 1,000 iterations were run for each scenario. Comparison of the TP and AP distributions suggest that skewness in the prevalence distribution can substantially affect the likelihood of accurately estimating Salmonella's true prevalence from a sample. For clustering scenarios, this comparison suggests the AP tends to underestimate the TP. The distribution of AP possibilities exceeds the predetermined acceptable precision boundaries. Therefore, the sample size determination is heavily affected by a priori estimates of the true prevalence and its distribution. |