Files
Abstract
Foodborne pathogen contamination can occur at many points along the broiler supply chain. As a fast-growing alternative to conventional broiler production system, the pastured poultry system allows chickens to graze freely on pasture with a movable pen, in which more stringent rearing practices are applied. As a result, food safety concerns may differ from those found in conventional production. Currently, the food safety risks associated with alternative production strategies are little understood. The goal of the current study was to fill in information gaps about food safety risks present in alternative broiler production system. Preharvest (feces and soil) and postharvest (ceca and whole carcass rinse) broiler samples were collected from pastured poultry farms and processing plants in the southeastern United States. Machine learning models, such as random forest, least absolute shrinkage and selection operator (LASSO), and regression tree (RT) were constructed to predict Campylobacter prevalence and E. coli concentration in pre- and postharvest samples based on farm management practices. The RMSE (in log10 scale) under LASSO was 0.974 and 1.437, while under RT it was 1.032 and 1.476 for feces and soil samples, respectively. It was found that the source of animal feces was the important factor in predicting Campylobacter prevalence and E. coli concentration in feces and soil samples. Furthermore, logistic regression, a statistical model, was used to predict presence and absence of foodborne pathogens by the indicator microorganism generic E. coli. Additionally, the association between pastured-related microbiome and Campylobacter and Salmonella presence was investigated. Samples were assessed for Campylobacter and Salmonella using selective media and molecularly using microbiome relative abundances via 16S rRNA amplicon sequencing. In addition, Linear discriminant analysis (LDA) effect size (LEfSe) was used to identify taxa significantly enriched in Campylobacter and Salmonella positive samples or negative samples. Finally, a retail-to-consumption risk assessment model was used to assess seasonal effects of the risk of having contaminated broiler meat produced and prepared in-home in the United States. The model showed that higher number of infections and illnesses were estimated in the summer and fall months. These findings will aid in the broiler industry's risk-based decision making.