The public health impact of different microbiological criteria approaches for Salmonella in chicken parts

Non-typhoidal Salmonella is a significant foodborne pathogen causing over a million illnesses each year in the United States. Poultry is one of the food commodities most frequently associated with Salmonella infections. While government, research, and industry efforts have reduced Salmonella contamination in poultry to some extent, the incidence of salmonellosis has not changed significantly and is still above the public health goals of Healthy People 2020, and novel and more comprehensive approaches are needed. In this paper, the public health impact of implementing different microbiological criteria (MC) for Salmonella in chicken parts was evaluated using a quantitative risk assessment approach. Four hypothetical scenarios, including a no-action baseline and three alternative scenarios, were considered. Scenario 1 modeled a prevalence-based microbiological criterion based on the proportion of positive samples in an establishment, Scenario 2 modeled a microbiological criterion based on the concentration of Salmonella in samples, and Scenario 3 modeled a combination of the two. With exception of the baseline, all three scenarios assumed that different interventions would be adopted for non-compliant establishments (Scenario 1) or lots (Scenario 2), with Scenario 3 combining establishment-level and lot-level interventions. The product was assumed to be sold to consumers as raw, and contamination via undercooked product as well as cross contamination in consumer kitchens were considered as potential exposure routes. Risk was characterized by the probability of illness and the preventable fraction of risk, which was calculated for each scenario in comparison with the baseline. Simulation results show that, depending on the parameters of specific sampling strategies, both prevalence-based and concentration-based MC coupled with interventions could significantly lower risk (range of 60–88% in mean preventable fraction of risk). Overall, while the model is preliminary and subject to the stated limitations, it is likely that a combination approach including establishment-level and lot-level interventions would be highly effective in reducing risk and, therefore, benefit public health. The effectiveness of all MC was impacted by several assumptions and model parameters. In particular, the prevalence MC threshold and the concentration reduction associated with the establishment-level intervention impacted the preventable fraction of risk for Scenario 1, and the concentration MC threshold and the variability across lots impacted the risk outcomes for Scenario 2. Overall, high variance in risk outputs was observed, mainly associated with a high variance in concentration inputs. This model provides a risk-based approach to test different MC approaches for chicken parts at both lot and establishment levels, and over a wide range of scenarios of input contamination distributions, interventions, and consumer behaviors. Model estimates, as well as the ability to distinguish between variability and uncertainty, could be improved by additional data on the distribution of Salmonella concentrations across and within establishments.