Identification of critical actors and paths of contamination in the Dutch broiler meat chain: investigating the robustness of graph-based traceability system

Today's poultry supply chains are complex, involving a diversity of actors and product quality labels. Risks of contamination with bacteria, such as Salmonella and Campylobacter are high and do occur regularly. A traceability system that identifies the critical actors and paths of contamination is essential for all actors in the food system to ensure a safe and healthy food supply. In this study, we enhanced a gravity model used in previous studies, which was limited to the two parameters of distance between actors and the capacity of actors , using a resource allocation optimisation method that allows all parameters of interest to be incorporated in the estimation of the supply chain network. We applied the new gravity model to incorporate product quality label information to identify critical actors and pathways of contamination in the Dutch broiler meat supply chain network using publicly available trade and actor data. The results show that slaughterhouses and cutting plants have the highest betweenness centrality, making them critical actors for contamination spread across the supply chain, while two hatcheries were identified as top actors based on degree centrality, indicating their potential influence within their respective supply chains. The predicted sources and pathways of contamination in the Dutch broiler meat supply chain network were largely comparable with the previous study. Though the inclusion of product quality labels in the modelling in this study has limited impact on the estimated supply chain network compared with the previous study, as the critical actors process broiler meat of diverse quality labels, the results demonstrated the value of the resource allocation optimisation modelling for incorporating all relevant parameters to model the supply chain network. The results suggest that the graph theory approach is methodologically robust in its predictive power as more data and more categories of actors are included.