Submitted to: Salin Pork
Publication Type: Abstract Only
Publication Acceptance Date: 7/1/2001
Publication Date: 9/2/2001
Citation: Bahnson, P.B., Cray, P.J., Mateus-Pinilla, N.E., Grass, J., Gray, J.T., Fransen, L., Ladely, S.R. Herd-level risk for salmonella culture postive status in slaughtered pigs. Salin Pork. pg. 244-249
Technical Abstract: A risk factor assessment was conducted on 65 pig herds participating in a slaughter health monitoring program. Salmonella was isolated using conventional culture methods designed to optimize Salmonella isolation rates. Fecal samples were collected less than 48 hours before shipment to slaughter. Caudal mesenteric lymph nodes and colonic and cecal content were collected immediately after slaughter. The proportion of Salmonella detected was 0.277. A survey (72 questions) recorded information on pig health, production, farm design and management, medication, and transport and lairage. Responses to the survey were used to construct 14 variables representing hypothesized risk factors for further analysis. The outcome variable was the arcsine transformed proportion (ASP) of culture positive pigs. Multiple linear regression was used to construct the best subsets of all possible models. The Adj. R2 statistic suggested that at most 10 variables should be included. Within each subset of the number of explanatory variables, many models fit the data approximately equally well. A total of 56 models had 95% or higher of the explanatory power, measured by the R2 statistic, of the best model for the subset. In addition, a single model was developed using backward stepwise regressions, with a statistical significance for inclusion set at p less than 0.1. The stepwise regression resulted in a four variable model, including factors for substandard (slow growth) pigs, pig flow/hygiene, presence of birds in the pig space and the interaction of the latter two variables. In this model, substandard growth was associated with an increased in ASP of 0.466, corresponding to an increase from mean proportion of Salmonella fro 0.277 to 0.502. Similarly contrasting herds with batch pig flow, good hygiene and no access to birds to herds with either no batch flow or poor hygiene and access to birds was associated with a change in ASP of 1.135 and a corresponding increase in Salmonella for 0.277 to 0.954. Although, Mallow's Cp statistic suggested that the four variable model was underparameterized, the statistical evidence to include additional variables was weak. However, review of these models indicated that regression coefficients were stable, indicating that confounding may not be present. Further, the more complex models may be used to help prioritize future research efforts.