Sensitivity analysis of a two-dimensional probabilistic risk assessment model using analysis of variance

Area of Science:

• Food safety science
• Risk assessment modeling
• Statistical analysis

Background:

• Microbial food safety process risk (MFSPR) models are crucial for assessing risks.
• These models often involve complex characteristics like nonlinearity, thresholds, interactions, and categorical inputs.
• Distinguishing between variability and uncertainty is essential for accurate risk assessment.

Purpose of the Study:

• To apply sensitivity analysis to two-dimensional probabilistic risk assessment models.
• To evaluate the effectiveness of Analysis of Variance (ANOVA) against correlation-based methods for identifying key controllable inputs and uncertainty sources.
• To address challenges posed by complex MFSPR model characteristics in sensitivity analysis.

Main Methods:

• Utilized a microbial food safety process risk (MFSPR) model as a test case.
• Compared Analysis of Variance (ANOVA) with correlation coefficient-based sensitivity analysis methods.
• Employed bootstrap simulation to quantify uncertainty in input ranks due to sampling error.

Main Results:

• ANOVA provided robust insights into controllable inputs for risk reduction, unlike misleading correlation-based methods.
• ANOVA correctly identified the top six important inputs, whereas correlation methods were less reliable.
• Sensitivity analysis identified home storage of ground beef servings as the most significant risk factor.

Conclusions:

• ANOVA is a superior method for sensitivity analysis in complex MFSPR models, offering reliable identification of key controllable inputs.
• Effective risk management strategies can be developed by prioritizing inputs identified through robust sensitivity analysis.
• Consumer advisories focusing on improved handling and storage of ground beef are recommended for enhanced food safety.