Characterizing the concentration of pathogen occurrence across meat and poultry industries

Raw meat and poultry contamination frequency demonstrably varies across establishments within an industry. The degree to which more contaminated establishments contribute to the overall production of contaminated products by the entire industry is rarely explored in quantitative microbial risk assessments. We use Lorenz curves and Gini coefficients to describe the degree of concentration of contamination within the industry for several product-pathogen combinations. The analysis is based on beta distributions fit to Salmonella sampling data for comminuted chicken, comminuted turkey, chicken parts, and comminuted beef and beef carcasses, as well as for Campylobacter in chicken and turkey. We also explore empirically-derived Lorenz curves. Lorenz curves for nine product-pathogen pairs suggest the pattern of contamination ranges from highly dispersed across the industry (e.g., Salmonella-comminuted chicken, Gini=0.19) to highly concentrated within a small part of the industry (e.g., Salmonella-beef carcass pre-chill, Gini=0.77). Generally, an inverse relationship between an industry's average contamination frequency and its Gini coefficient is observed across these examples. Also, illustrative empirical Lorenz curves are biased relative to the fitted curves because of a substantial number of empirical results with point estimates of zero. Large Gini coefficients may suggest that risk management should focus on just that part of the industry with high contamination frequencies, while small Gini coefficients might suggest that risk management should be more dispersed and holistic across the industry. We discuss monitoring Gini coefficients across time, as well as other applications of these methods, as useful adjuncts to food safety risk assessments.