Probabilistic modelling of events at evisceration during slaughtering of pigs using expert opinion: Quantitative data in support of stochastic models of risk of contamination

Abstract The evisceration stage is one of the most critical steps in the slaughtering process of pigs when considering the risk of carcass contamination. Unfortunately, it is also characterized by a number of fundamental quantitative data gaps preventing modellers from reproducing events in probabilistic terms. Recognising the practical difficulties that a systematic data collection would imply, in this study we modelled the answers of structured questionnaires submitted to eleven veterinarians (official veterinarians/meat hygiene inspectors) working in pig abattoirs to provide ready-to-use probability distributions in support of future quantitative risk assessments. The questions were aimed at modelling the occurrence of ruptured gut ( P GUT ) and gallbladders ( P GALL ) during evisceration procedures, the amount of faecal ( F L ) and bile ( B L ) contamination dropping on the carcass, the probability of internal cavities ( P I F − B ) and external surface ( P E F − B ) being contaminated and the conditional probability of partial condemnation of the carcasses (as unfit for human consumption) as a function of the level of contamination ( P C S a ). The answers were weighted according to the level of confidence each expert had in their own estimation. Out of 10,000 simulated values, P GUT and P GALL were higher in small (Mean = 0.048 and 0.035) compared to high (Mean = 0.021 and 0.016) or middle (Mean = 0.025 and 0.019) throughput abattoirs. The cumulative distributions describing F L and B L produced 50th and 90th percentile values of 24.5 g and 19.9 g (50th percentile) and 88.7 g and 68.8 g (90th percentile), indicating the level of contamination is generally low. The distributions describing both P IF and P EF and those describing P IB and P EB show comparable shapes suggesting there are no significant differences in the likelihoods of those events when considering the faecal and bile contamination respectively. Finally, the results obtained for P C S a suggested that common linear or nonlinear relationships are not adequate to describe the probability of a carcass being partially condemned as a function of the dose. Highly contaminated carcasses are not unlikely to be detained for manual removal of visible contamination rather than partially condemned, indicating that factors other than the amount of contamination are driving this relationship. With this study, we made use of the experience of eleven meat hygiene inspectors/official veterinarians to provide quantitative information on the key events occurring during evisceration. As presented, the probability distributions can be directly used to inform and integrate probabilistic models aimed at estimating to the risk of human exposure to foodborne pathogens through consumption of pork products.