A likelihood-based approach to identifying contaminated food products using sales data: performance and challenges

Area of Science:

• Food safety
• Epidemiology
• Supply chain management

Background:

• Foodborne disease outbreaks pose significant public health and economic risks.
• Interconnected supply chains can amplify the impact of contamination events.
• Rapid identification of the source is critical to mitigate harm.

Purpose of the Study:

• To develop and evaluate a likelihood-based approach for accelerating the identification of contaminated food products during outbreaks.
• To assess the method's performance using real-world sales data and simulated outbreak scenarios.
• To extend the method for binary classification and explore factors influencing identification difficulty.

Main Methods:

• Utilized a likelihood-based approach integrating food product sales data with foodborne illness case report distributions.
• Employed real-world food sales data and artificially generated contamination scenarios for validation.
• Extended the method to function as a binary classifier to identify sets of potentially contaminated products.
• Analyzed food distribution patterns and spatial sales data correlations to understand identification challenges.

Main Results:

• The likelihood-based method effectively identifies single contaminated food products in simulated outbreaks.
• The binary classifier extension accurately includes the true source in a set of suspect products.
• High spatial correlation in product sales data is identified as an indicator for difficult-to-trace contaminated foods.

Conclusions:

• The developed likelihood-based approach significantly accelerates the identification of contaminated food products during outbreaks.
• The method provides a valuable tool for public health officials and the food industry to manage food safety crises.
• Understanding product sales correlations can help proactively identify high-risk food items.