Describing Uncertainty in Salmonella Thermal Inactivation Using Bayesian Statistical Modeling

Area of Science:

• Microbiology
• Food Safety
• Statistical Modeling

Background:

• Uncertainty analysis is crucial for evaluating scientific conclusions, especially in microbial risk assessment.
• Deterministic models provide point estimates, failing to quantify prediction uncertainty.
• Predicting microbial behavior accurately requires methods that capture uncertainty.

Purpose of the Study:

• To employ Bayesian statistical modeling for quantifying uncertainty in the thermal inactivation of *Salmonella enterica* Typhimurium DT104.
• To compare the predictive performance of two-step and global Bayesian regression approaches.

Main Methods:

• Utilized thermal inactivation data for *Salmonella* in broth (aW 0.75) from the ComBase database.
• Applied a log-linear model for primary inactivation and a linear model for secondary temperature-dependent inactivation.
• Performed data fitting using two-step and global Bayesian regression techniques.

Main Results:

• Bayesian modeling generated probability distributions for cell density, reduction time, and inactivation rate.
• The global regression approach yielded less uncertain predictions than the two-step method.
• Model validation confirmed its ability to describe thermal inactivation uncertainty, with most data within 95% prediction intervals.

Conclusions:

• The developed Bayesian model successfully quantifies uncertainty in microbial thermal inactivation predictions.
• This approach enhances microbial risk assessment and processing design by providing probabilistic outputs.
• Global Bayesian regression offers more robust and less uncertain predictions for thermal inactivation studies.