A generalized linear mixed model approach to assess associations between Certified Food Protection Managers and inspection performance in Franklin County, Ohio

In the United States, regulations often require public-facing food establishments to employ Certified Food Protection Managers (CFPMs). CFPMs are individuals who have received food safety training and demonstrated their knowledge through passing a standardized exam. In this study, the effect of a recently enacted change that required food establishments in Ohio, United States of America, to employ a CFPM was evaluated using generalized linear mixed models (GLMMs). A list of non-compliance items found during 5,234 inspections were obtained for Risk Level III and IV food establishments operating in Franklin County, Ohio, for the year immediately following the implementation of the CFPM requirement. GLMMs were used to test the hypothesis that food establishments that complied with the CFPM requirement were more likely to earn a satisfactory inspection result (defined by the number of non-compliance items identified at an inspection) than those that did not comply. Analyses were stratified by establishment Risk Level, controlled for establishment type, and averaged across inspectors and geographical areas. Satisfactory inspection performance was significantly associated with compliance with Ohio’s CFPM requirement at Risk Level IV (OR: 2.166 (95% CI: [1.712, 2.740])) establishments. This analysis provides quantitative statistical support for an association between Ohio’s CFPM requirement and satisfactory inspection performance. Further, this study demonstrates how GLMMs can be used to evaluate changes to retail food safety regulations while accounting for correlation due to inspectors and multiple inspections at establishments. • Certified Food Protection Managers are associated with improved inspection performance • Inspection data can be used to evaluate impacts of changes to retail food safety policies • Inspectors are an important source of variation in inspection data • GLMM approaches can account for differences between inspectors • Appropriate statistical approaches for inspection data depend on research questions