Flexible risk process models to quantify residual risks and the impact of interventions

The produce industry needs a model to (i) identify the most important risks in a supply chain and (ii) identify which practices and control strategies appropriately reduce risks of contamination events that could lead to product recalls and illness outbreaks. This could mean which pathogen is most important for a commodity, or which practice represents the largest risk for a given supply chain. Our project meets that need. We do this by first modelling the risk in a supply chain for leafy greens contaminated by two important pathogens, either Shiga toxin–producing Escherichia coli or Listeria monocytogenes. In future work, we would expand the model to accommodate additional pathogens, practices, and commodities, as identified by feedback from CPS stakeholders, and then show how to use the model to assess the impact of newly identified risks such as newly identified problematic practices, emerging pathogens, or products. In all our analyses, we will measure the impact of newly identified risks or newly modeled control strategies by how they change the total supply chain risk as compared to the risk uncontrolled by current practices. [Note: This award is for year 1 of a planned maximum three-year renewable period. Continuation of funding will be based on successful mid-term and year-end review of progress and accomplishments in years 1 and 2.]

We propose to build an adaptable supply chain risk model with steps universally applicable to most produce, parameterize this model with academic and industry data and expert opinion, and then distribute the model for both research into specific supply chains and as an adaptable stakeholder tool. Specifically, we will use previously published language to describe a generally applicable produce supply chain. We will then mathematically describe unit operations that can affect the safety of the produce, incorporating analyses to determine the marginal food safety gains of different operations. In the first year, we will build a supply chain risk model (SCRM) for Shiga-toxin producing Escherichia coli (STEC) and Listeria monocytogenes (LM) simultaneously in leafy greens. Major outputs will be a comprehensive model of residual risk for STEC and LM from current practices from preharvest to consumer, and sensitivity analysis of factors influencing risk to guide focus of future work. To advance to year 2, CPS stakeholders suggest “what-if?” scenarios for modelling additional commodities, hazards, and practices. In the second year, we will improve the STEC and LM model with feedback from CPS stakeholders and adapt the SCRM to accommodate additional pathogens and commodities. A major output will be the ranked risk posed by what-if scenarios, beyond the initial STEC and LM in leafy greens scope. To advance to year 3, CPS stakeholders suggest factors to focus on to reduce uncertainty in revised models and generate lists of candidate under-determined factors to assess with future research. In the third year, we will define an adaptable SCRM system for assessing the importance of previously unmodeled candidate risk factors. A major output will be ranked risks from both explicitly modeled factors and the aggregate risk posted by other factors not explicitly modeled at each step. At this point, stakeholders will have a system to bound the influence of newly identified risk factors, such as a newly uncovered harborage site. One could assess that risk using a what-if scenario. If important model outputs, such as residual risk of illness, were highly sensitive to this change, that increased risk is important in the context of the other known risks. Also critical, if model outputs show low sensitivity, the current supply chain likely has sufficient downstream risk reduction strategies to manage the increased risk, or this specific risk is minor.

1. Review of contemporary STEC and L. monocytogenes risk assessments and process models in leafy greens to collect (i) relevant process steps, (ii) model parameters, and to (iii) identify data needs for future, improved risk assessments. 

2. Build flexible supply chain process models for STEC and L. monocytogenes in leafy greens to evaluate the effect of literature-based and industry-suggested contamination scenarios and management strategies on the risk of a product recall.

Summary of Findings (Y1–Y3)

• We systematically reviewed 11 and 7 recent risk assessments for Shiga toxin–producing Escherichia coli (STEC) and L. monocytogenes in leafy greens. That review extracted 70 unique parameters that can support the development of future leafy greens risk assessments.

• A retail positive test was an outcome measure that industry stakeholders were willing to engage with.
– Highest-risk lots were those with >1 in 10 chance of a retail positive test and considered lots of “public health risk.”
– The overall probability of a positive test across all lots was defined as “recall risk.”

• The Supply Chain Risk Model (SCRM) was used to simulate a leafy green supply chain contaminated with Shiga toxin–producing Escherichia coli (STEC), and we assessed tradeoffs between (i) improved process controls and (ii) additional product testing.
– Rare, high-level contamination events drove recall risk.
– Implementing improved process controls and additional product testing each reduced public health and recall risks.
– However, additional product testing reduced recall risk at the expense of rejecting many lower-risk lots.

• Additional scenarios were developed to assess the impact of small-scale failures resulting in contamination or deviations from food safety protocols in leafy green supply chains.
– Modeling inadequate process wash water control (specifically, maintaining adequate free chlorine only about half the time) increased both public health and recall risks, even when initial contamination variability was low.
– Modeling preharvest failures resulting in contamination, specifically agricultural water treatment failure, small animal fecal contamination, or inadequate harvest sanitation between harvests, resulted in small changes to public health risk and recall risk.

• Preharvest testing plans with more total mass are more effective at detecting and managing risk from contamination that is spread uniformly across a field, whereas plans with more total grabs are better for managing risk from contamination clustered in specific areas of a field.

• Altering aspects of irrigation water testing plans (such as increasing sample volumes or numbers) can improve the likelihood of detecting contamination in irrigation water and minimize produce safety risks.

Recommendations

Industry experts and association representatives could work together to