Bayesian predictive risk modeling of microbial criteria for Campylobacter in broilers

Microbial Criteria define the acceptability of food products, based on the presence or detected number of microorganisms in samples. The criteria are applied at the level of defined food lots. Generally, these are interpreted as statistical batches representing the production [1]. The batches not complying with a criterion can then be e.g. rejected. A risk reduction for consumers is therefore expected. However, a quantitative estimate of the implied risk reduction is non-trivial, because it depends on many unknown parameters. The quantity and quality of data lead to uncertainties which can be assessed by computing posterior distribution of the parameters - a Bayesian evidence synthesis. The outcome of a defined Microbial Criterion (MC) for a batch provides additional evidence concerning the batch. Posterior predictive consumer risk (probability of illness) was computed for such batch(es) with the given outcome (MC met / MC not met / MC not applied) with OpenBUGS.