Advancing food safety through simulation of school cafeteria share tables and produce systems

Reyes Polanco, Gustavo Abel

Permalink

https://hdl.handle.net/2142/121335 Copy

Description

Title

Advancing food safety through simulation of school cafeteria share tables and produce systems

Author(s)

Reyes Polanco, Gustavo Abel

Issue Date

2023-07-13

Director of Research (if dissertation) or Advisor (if thesis)

Stasiewicz, Matthew J

Doctoral Committee Chair(s)

Prescott, Melissa P

Committee Member(s)

• Banerjee, Pratik
• Davidson, Paul C

Department of Study

Food Science & Human Nutrition

Discipline

Food Science & Human Nutrition

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• Food Safety
• Modeling
• Produce Safety
• Microbiology

Abstract

Modeling strategies, such as quantitative microbial risk assessments (QMRAs), and process models are important for understanding food systems and their related microbiological processes. In this document, four food systems were modeled. (i) A QMRA was developed to understand the norovirus risk associated with share table (ST) implementation in a school cafeteria system and the effect of risk mitigation strategies. Process models for (ii) leafy greens & STEC, (iii) and tomatoes & Salmonella spp. were developed to understand the performance of sampling plans in these two systems. A Cyclospora cayetanensis testing model was developed to characterize current detection methods, and a growth cycle simulation was built to understand sampling plan performance for a cilantro growth cycle. Share tables are locations in school cafeterias where students can share unwanted food items, which would otherwise be discarded. While STs can provide food security benefits by increasing food availability, food safety concerns expressed by shareholders are a barrier to their implementation. Therefore, a QMRA was developed to assess the norovirus food safety risk associated with STs, and the effect of relevant risk mitigation strategies. 13 what-if scenarios were developed to assess the impact of implementing STs and the efficacy of mitigation strategies. Results showed that, after 1,000 iterated weeks, STs modestly increase the mean illness prevalence from 1.5% (2.5th – 97.5th percentile: 0.52%-2.7%) to 1.6% (2.5th – 97.5th percentile: 0.67%-2.8%), a 6.8% relative increase in by implementing STs. Yet, mitigation strategies that focus on managing incoming norovirus loads are predicted to be most effective: efficient student handwashing and hand sanitizing reduced the relative illness prevalence to 43.6% and 41.9%, respectively. Other strategies such as a one-way share table were able to mitigate the added risk. While STs are predicted to slightly increase the illness prevalence, the risk is manageable by applying mitigation strategies For produce, test-and-reject (sampling) plans are often required for market access. However, the effect that these sampling plans have on reducing the number of adulterants that reach the system endpoint is still unknown. In addition, these types of sampling plans often focus on single stages in the supply chain (preharvest or finished product). Limited evidence exists to address the power of sampling at these stages compared to other stages. Here, process models were developed to better understand the impact of sampling. The power (detection rate), and their relative efficacy in preventing adulterants from reaching the system endpoint were analyzed to quantify sampling performance. For leafy greens, to better understand the impact of sampling, we simulated the effect of sampling (from preharvest to customer) and processing interventions (such as produce wash with antimicrobial chemistry) on the microbial adulterant load reaching the system endpoint (customer). Seven leafy green systems, an optimal system (all-Interventions), a suboptimal system (no-Interventions), and five systems, where single interventions were removed to represent single process failures were simulated. A scenario analysis combined seven sampling plans, seven systems, and three contamination clustering spreads for 147 total scenarios. The all-interventions scenario (where all processing interventions were included) resulted in a 3.4, 95% CI (3.3 - 3.6) log reduction to the total adulterant cells (TACs) that reached the system endpoint (endpoint TACs). The analysis suggests sampling plans that happen before effective processing interventions (preharvest, harvest, and receiving) were most effective at reducing endpoint TACs, ranging between 0.05 and 0.66 log additional reduction compared to systems with no sampling. In contrast, sampling post-processing (finished product) did not provide meaningful additional reductions to the endpoint TACs (No to 0.04 log reduction). For tomatoes, A 42-day tomato season was simulated. scenarios included 96 combinations of the 4 contamination spreads (Widespread-100%, 10%, 1%, and 0.1% spreads), 4 sampling locations (preharvest, harvest, receiving, packing), and 6 sampling-testing methods (2, 6, 20, or 60 tomatoes, 20 or 60 tomato mash). The model suggests the best sampling location depends on the initial contamination spread. For the widespread-100%, 10%, and 1% spreads, sampling preprocessing (Harvest and Receiving) had the highest probability of detecting contamination, between 26% to 30%. This study also predicted that sampling-testing methods with higher tested mass (60 tomatoes: 15,600g and 20 tomatoes: 5,200 g) yield higher power when the contamination spread was larger (Widespread-100%, and 10% clusters). When the contamination spread was smaller, 1% and 0.1% clusters, sampling-testing methods that sample more tomatoes (60 tomatoes and 60 tomato mash), had higher detection power regardless of the testing method (mash or whole tomato enrichment). For cilantro, detection method performance and a 45-day growing cycle were simulated to assess sampling efficacy for Cyclospora cayetanensis contamination. Empirical data from two peer-reviewed studies was used to fit the method performance as a function of contamination levels. The results suggest contamination levels above 0.92 oocysts per liter for water testing, and 1.58 oocysts per liter for product testing, at least one single 10L or 25g sample will reliably detect contamination. Below these thresholds, increasing the number of samples will increase the probability of detection. The 45 days growth cycle simulation shows that daily production or water testing will more efficiently detect a random (once per season) contamination event while sampling once a season may not be enough to reliably detect this random contamination event.

Graduation Semester

2023-08

Type of Resource

Thesis

Handle URL

https://hdl.handle.net/2142/121335

Copyright and License Information

Copyright 2023 Gustavo Reyes Polanco

Owning Collections