Associations of processing level variables with Salmonella prevalence and concentration on broiler chicken carcasses and parts in Canada

A national baseline study was conducted between December 2012 and December 2013 to determine the pre-packaging prevalence and concentration of foodborne pathogens on broiler chicken carcasses and parts at processing; a survey was implemented simultaneously to collect data on the processing practices used to control these pathogens. Thirty federally-registered Canadian poultry processing establishments completed the questionnaire. A total of 2,732 samples of carcasses and parts (breast and thigh pieces) were collected over the study period from these establishments. For Salmonella, the overall proportion positive was 0.22 (95% CI 0.20, 0.23), and the mean concentration was 0.67 (95% CI 0.51, 0.83) MPN/mL of rinse fluid. Multivariable regression models with random intercepts for the establishment and the date of sampling were used to identify associations between Salmonella prevalence and concentration and processing practices. In the final logistic regression model for the prevalence outcome (positive or negative sample), there were three statistically significant variables: product type (carcass or part); chilling method (water or air); and chlorine use in the establishment (chlorine, cetylpyridinium chloride, or neither). The likelihood of testing positive for Salmonella was higher on parts than carcasses (OR 3.03, 95% CI 2.38, 3.86), and higher when cetylpyridinium chloride was used (OR 2.00, 95% CI 1.36, 2.95), or when other processing aids were used (OR 1.99, 95% CI 1.26, 3.15), than when chlorine was used. Water chilling was negatively associated with testing positive for Salmonella when compared with air chilling (OR 0.68, 95% CI 0.48, 0.96). In the final linear regression model for the concentration outcome (log(10) MPN/mL), there was one statistically significant variable chilling method, where water chilling was associated with a decrease in concentration (13-0.23, 95% CI -0.38, -0.08 log(10) MPN/mL). The intraclass correlation coefficients for establishment and date sampling were 0.02 and 0.23 in the linear regression model, and 0.01 and 0.34 in the logistic regression model, respectively. Further studies to explore the methods to reduce microbial contamination during the air chilling and cut-up and boning processes in broiler chicken establishments in Canada are recommended.