Salmonella assessment along the Spanish food chain: Likelihood of Salmonella occurrence in poultry and pig products is maintained across the food chain stages

Abstract

Salmonellosis is one of the most important foodborne diseases worldwide, including the European Union. Despite the One Health approach measures for risk assessment and risk management implemented by the European Union, the occurrence of disease and disease outbreaks remains high (e.g. 694 outbreaks were reported in 2020), highlighting the need of new assessment methods. Herein we applied machine learning using the random forests method to evaluate and identify key points regarding the occurrence of Salmonella sp. along the Spanish food chain during 2015-2020, using data provided by the Spanish Agency for Food Safety and Nutrition. We compared the role of the three categorical variables [product (20 categories), region (18 categories) and stage (11 categories)]. Salmonella presence was influenced by the three explanatory variables considered: first by product, followed by region and stage. The most determinant product for Salmonella probability was 'meat', while the most important stage was 'slaughterhouse'. Specifically, the highest values were found in pig and poultry meats. In these products, the Salmonella probability was high at the early and final stages of the food chain, although not at intermediate stages. The presence of Salmonella in the final stages (retail) of the food chain is of concern, as it can cause human cases of salmonellosis, including outbreaks. This study demonstrates the utility of the random forest method to identify key points and evaluate the control efforts. We recommend improving the surveillance and control measures, especially in the product and stages pointed out by our analysis, and enhancing the data collection harmonization among the different autonomous communities.

Keywords: One Health; food-borne infections; random forest; salmonellosis; zoonosis.