Monday, July 23, 2012
Exhibit Hall (Rhode Island Convention Center)
Joost Smid, National Institute for Public Health and the Environment, Bilthoven, The Netherlands
Rob de Jonge, National Institute for Public Health and the Environment, Bilthoven, Netherlands
Arno Swart, National Institute for Public Health and the Environment, Bilthoven, Netherlands
Annemarie Pielaat, National Institute for Public Health and the Environment, Bilthoven, Netherlands
Arie Havelaar, National Institute for Public Health and the Environment, Bilthoven, The Netherlands
Introduction: The transfer ratio of bacteria from one surface to another is often estimated from laboratory experiments and quantified by dividing the expected number of bacteria on the recipient surface by the expected number of bacteria on the donor surface. It can only be estimated with limited precision and its estimate may exceed 1 if real transfer is close to 100%. In addition, transferred fractions may vary over multiple experiments but it is unclear, using this approach, how to combine uncertainty and variability into one estimate for the transfer ratio.
Purpose: To develop a method by which uncertainty within one experiment is combined with variability over multiple experiments and by which inappropriate values for the transfer ratio are prevented.
Methods: A Bayesian Network model was developed for this purpose. The model was tested using data from a laboratory experiment in which the transfer of Salmonella from contaminated pork meat to a butcher’s knife and from the knife back to pork meat was determined. Recovery efficiency of bacteria from both surfaces was also determined and accounted for in the analysis.
Results: The functionality of the model was demonstrated. The transfer ratio probability distributions were shown to have a large variability, with a mean value of 0.11 for the transfer of Salmonella from pork meat to the knife and 0.36 for the transfer of Salmonella from the knife to pork meat.
Significance: The proposed Bayesian model can be used for analyzing data from similar study designs in which uncertainty should be combined with variability.
