P3-116 Validation of Predictive Risk Tools Applied to Strategic Facility Investments

Wednesday, August 3, 2016
America's Center - St. Louis
Anthony Pavic, Birling Avian Laboratories, Bringelly, Australia
Ashley Kubatko, Battelle Memorial Institute, Columbus, OH
Regina Gallagher, Battelle Memorial Institute, Columbus, OH
Eric Johnson, Battelle Memorial Institute, Columbus, OH
Brian Hawkins, Battelle Memorial Institute, Columbus, OH
Introduction:  Strategic investment decisions regarding plant improvements across the enterprise are challenging.  At any given time, many potential improvements or updates could be undertaken at any one of a company’s multiple facilities; however, financial resources to support these investments are limited.  Extensive data is likely available through various systems, including temperature logs and microbial sampling, but how to use that data to inform decisions to best reduce risk to the enterprise is far from clear.

Purpose: The purpose of this study was to validate predictive risk tools for the purpose of evaluating strategic investments across a poultry enterprise by comparing predicted microbial levels of packaged product against laboratory measurements.

Methods: Historical plant improvements and their impact on processing conditions were quantified for three facilities over a five year window. Ambient temperature data, process logs (e.g., chlorine levels and core temperatures measured at chilling), and the prevalence and severity of incoming contamination were also mined to inform the predictive modeling. A predictive cloud-based software tool was utilized to perform ensembles of simulations for each quarter of the time window using these data as inputs in order to produce predicted contamination levels as a function of time for each facility.

Results: Comparison of model predicted microbial levels and laboratory measured microbial levels of packaged products show reasonable agreement for air packed poultry (favorable comparisons and consistently similar trends).  These results indicate that potential strategic investments can be evaluated in silico based on currently available data in a manner that informed decisions that reduce risk to the enterprise.

Significance: This presentation provides a quantitative comparison of predictive model results with actual laboratory measurements that illustrate how quantitative software-based modeling tools can be applied to inform strategic facility investments based on readily available and reasonably predictable data.