Maybe the reason we’re not making progress on improving traceability in produce is because we are talking past each other. Given the number of times we’ve heard “I don’t know what the problem is. My company has great traceability,” paired with U.S. Food and Drug Administration (FDA) complaints that traceability in the produce industry is lacking, there is clearly a disconnect. This article seeks to view traceability from various perspectives, looking beyond record keeping to the physical commingling of product and the immense challenges that ensue when several foods may serve as vehicles for a common source of contamination.

Are Traceability Issues Unique to Produce?
Is traceability of fresh produce worse than other sectors? No. However, the reliance on traceability, as well as on people’s memories, is greater for fresh produce than for many other food items (especially shelf-stable ones). Early in the outbreak investigation process, individuals are interviewed to determine whether they consumed a common food item. Fortunately, fresh produce consumption is high, so it’s not surprising that ill (and well) individuals frequently report consuming produce. Because of the frequency of consumption, one produce item may be consumed as an ingredient in many meals. Iceberg lettuce, for example, may be eaten as part of a wedge salad (with cheese, dressing, tomatoes, and bacon), in a bagged blend of iceberg and romaine, or as a burger topping. This complicates the epidemiological investigation to tease out “iceberg” from the variety of other ingredients eaten concurrently. This isn’t different from many processed food products (flour, spices, etc.). However, it’s easier to verify and isolate the purchase of a shelf-stable product that lasts for a while (e.g., a bag of flour that’s only purchased a few times a year) compared with fresh produce (e.g., iceberg that may be purchased or eaten twice a week). Thus, components of epidemiology and traceback become intermingled early in the investigative process.

Figure 1 compares and contrasts hypothetical outbreak investigations associated with fresh produce versus a shelf-stable food. The key to solving an outbreak is knowing the lot number associated with the product that made someone ill. Having product in hand, which is more likely the case for shelf-stable products (peanut butter, frozen vegetables, flour, etc.) substantially increases the likelihood that the lot number can be determined, because it’s typically printed on the package. But when it comes to fresh produce, it’s typically printed on the package too! Sure, there is bulk produce, but today, many produce items are branded and packaged, putting them on par with other types of food products. The difference is, by the time an outbreak is realized (generally several weeks after the first illness), that package is long gone. Unless that lot number was captured, investigators are starting with a very generic description of an item and need to review paperwork to somehow, hopefully, figure out the lot number that was printed on that package.

If a jar of peanut butter or bag of frozen peas wasn’t stamped with a lot number, we suspect that outbreaks associated with those products would be just as difficult to solve.

If jelly is the suspect item, and investigators find two jars of “Jen’s Jelly” in an ill consumer’s home, and a test of an unopened jar is positive for the pathogen of interest, we can assume that the contents of the opened jar made someone ill. There is no reason to visit the grocery store that sold Jen’s Jelly and follow a complex supply chain to find the manufacturer. The manufacturer is clearly Jen. The FDA can go directly to Jen without securing the one-up/one-down records to show how the jelly moved from the manufacturer to the consumer’s home. There is a direct link—“a smoking gun”—case closed.

Challenges for Produce
Without a lot code, or even a brand, investigators pursuing a fresh produce-associated outbreak must go to the best-guess point of purchase and identify what product shipments could have been available for the consumer to have consumed or purchased. The more frequently the product is consumed, the more difficult it is to identify which eating occasion made someone ill. This may explain why restaurants are often the starting point of outbreak investigations involving fresh produce. When multiple ill people report dining at a specific location (or even a specific chain) in a narrow time frame, that is the smoking gun. Depending on the level of sophistication of that location, there could be very detailed records or maybe a vague receipt or invoice. Even with the most detailed records, it is likely that multiple shipment records must be collected to capture the possible dates that could line up with the patient consumption time frame. This is problem number two. As part of this, there are the issues with identifying what lots were on those shipments. Are lot codes included? Are brands named? Are product descriptions clear (e.g., romaine, Rom/Hrt, RomBlendIce, CelloRom, etc.)? Even if the product lot codes are available at the point of sale/service and naming is clear, one is still faced with multiple lots that could have been “the one” if the turn of the product is high. Today, this information is rarely available at the point of purchase or final delivery before the consumer. The journey backward consists of looking at the possible distribution points prior, and the various distribution points prior to that, and so on. Here, we have the same issues magnified, and thus, the list of suspects grows. Keep in mind that if other food items relied on this approach, the same problems would be faced. Traceability alone, especially in a lengthy one-up/one-down fashion, rarely definitively identifies the source and cause of a foodborne outbreak. Historically, foodborne illness investigations rely directionally on traceback, which can be verified by product testing.  

At some point in the record-keeping conversation, the word “commingling” might be used. In this context, it could mean that a retail store received a pallet that contained four cases of a fresh produce item, and that each case was associated with a different lot number. Two, three, or even all four lot numbers may have been available for sale during the time frame of interest (maybe even more depending on how the store managed its inventory). It’s important to clarify that in this context, the individual lots of product are still distinct. However, the records that allow one to discern what was shipped or sold may not be granular enough to differentiate one lot from another. Improved record keeping can help tease apart this commingling.

At the point that investigators have cobbled together enough information from records in the supply chain to generate a list of potential lots (for instance, in the example above, all four lots may be on the list, even though it’s possible, or perhaps probable, that not all four were contaminated), this may point toward one or more processors or packers. At this point, the word “commingling” enters the conversation again, but in a very different context. FDA may comment that multiple farm lots were used for a particular production run of salad. This is absolutely true for a salad blend containing multiple types of vegetables but not uncommon for single ingredients as well (e.g., chopped iceberg). It is not dissimilar to the manufacturing of a processed food product (multi-ingredient like jelly or single ingredient like flour). It’s unclear why there is much more emphasis on finding the exact farm location where contamination of fresh produce occurred compared with other products that are contaminated in the field. For example, in the case of flour, why aren’t we asking which wheat field was contaminated? Is the situation that dissimilar from romaine?

It’s critical to recognize that some fresh produce suppliers (packers and processors) are able to identify with certainty what farm lots went into the products, while others may not track this. This is a loophole in the current record-keeping requirements that we hope will be addressed as FDA proceeds in writing a traceability-oriented rule. Whether tracked or not, having multiple inputs associated with one finished product expands the possible culprits. Record keeping will not improve the inability to discern this commingling; changes in production processes would be required.

Another commingling variant has to do with the farm itself. FDA identifies traceability difficulties due to the fact that a single farm may supply multiple shippers/processors. Each shipper/processor may identify the farm differently, obfuscating any common source. Current record-keeping requirements exempt farms from keeping records (although most produce farms do have traceability records). If investigators stop at the last “registered facility,” they may not realize that the farm source named by one facility is in fact the same farm source named by a different facility. Speaking a common language is key to recognizing convergence.

The “equation” summarizes the issues that increase the amount of “noise” in the traceback of a fresh produce item, diminishing the “signal.” Addressing only one of these issues will help, but will not solve, our traceability challenges.

Packaging and Labeling Issues
We would be remiss to not mention the concept of provenance labeling, which was introduced for romaine following the November 2018 outbreak of Escherichia coli O157:H7. Provenance labeling is not traceability. It is a Band-Aid fix, and a poor one at that, which assumes the status quo will be maintained and that FDA will continue to issue broad advisories due to the inability to distinguish the “signal” from the “noise.” As described in the opening of this article, if, prior to an outbreak, bagged romaine had been labeled as “Romaine: Salinas,” it would not have aided in tracing the product because those packages would have long been discarded. The appearance of this information on packaging was quite valuable this fall—after an outbreak was identified that was tied to the Salinas region. This allowed retailers and consumers to differentiate sources of the product, so that romaine from other regions could continue to be enjoyed.

But FDA’s link to Salinas was not due to the provenance label on the package. It was due to the specific lot information printed on a salad bowl that tested positive in the home of a consumer. This was a rare instance where a product was available for testing. Why, with this smoking gun, was there still an advisory? As of the date of this writing, the investigation is still ongoing. However, we know that the lot code information on the package was adequate to identify a short list of growers and a handful of fields. However, since ill consumers in other states did not report eating this brand, the agencies fear that a contaminant may have affected more than one field in the region. If an environmental contaminant affected more than one type of product, then the traditional epidemiological approach of using food history questionnaires would never be able to identify a common product. Additionally, as noted earlier, it’s also possible that a grower also sold product to other salad brands.

What Needs to Change to Improve Produce Traceability?
If we define “good traceability” as being able to rapidly link the product that caused illness to the location where contamination occurred, what changes need to occur to enable this?

•    Digitizing data

•    Using standards to communicate (e.g., GTIN); also standardizing locations (e.g., ranches)

•    Capturing lot data at the point a case/box is opened

•    Limiting the number of input lots into an output (of the same product) and/or limiting the run size of a lot of finished product (realizing the inefficiencies in doing this)

There are several key improvement opportunities:
1. Capture the original lot code information or any other data that can link the final destination of the product to the product origin and retain those data in digital format. This is easier said than done and does not exist for any food but would allow for quicker assembly of potential product information. Currently, there is a tremendous amount of data and information exchanged between trading partners. We don’t necessarily need to collect more information. We need better information. And linking origin to destination theoretically cuts out the need to collect data at the “in-between” points of distribution.

If lot-specific information is not available or is not captured by the terminal location, then each point would need to capture data that would allow for more rapid capture of the potential lot(s) involved. To facilitate communication within the chain, data standards should be utilized. This approach may result in some uncertainty, but the number of “possible sources” should be finite. When aligned with patient point-of-purchase and illness onset date information, one could rule out lots, brands, or regions that were not involved.  

2. Public health agencies should get more specific sooner in their patient interviews. “Leafy greens” and “romaine” are essentially catchalls. We need to give the American consumer more credit. They know whether they eat organic, what brands they prefer, where they eat, how often they eat what, and what they eat. No one eats just “romaine.” If we take the rather unique circumstance of the “Thanksgiving” strain of E. coli O157:H7 that seems to reappear each fall, as soon as the whole-genome sequence is one that has a previous product association, food history questionnaires should begin to get specific very quickly. While there is concern about drawing conclusions prematurely, we must also use historical knowledge to attempt to reach a resolution more quickly.

3. We need to rethink the traceback approach when it comes to produce. Common distribution patterns, specific markets, source locations, and regionality can all be incorporated at the front end of an investigation. Recognizing that FDA may be trying to put a legal case together throughout the investigation, there is an opportunity to identify the source from a different angle if we begin to approach investigations with a public health focus. Rather than trying to work backward step by step, why not try to figure out what supply went into the point-of-purchase locations involved from the front end? By and large, the supply chain knows where their product goes from the moment it ships out the door. It may seem daunting at first glance, but it’s worth a shot. In fact, it is a shotgun approach and something will hit the target. If it doesn’t line up, it doesn’t line up. This does not mean that the food industry must constantly upload all traceability data to a central repository on an ongoing basis. Rather, as an investigation unfolds and illnesses occur only in certain areas (e.g., only restaurants in the Midwest), supply chain knowledge can begin to help narrow the possible supply chains involved.

Conclusion
Whole-genome sequencing lets us link illnesses and recognize outbreaks that surely would have been missed just a few years ago. Whole-genome sequencing represents a powerful new tool, and perhaps the epidemiology and traceback approaches need to catch up.    

Consumers are eating a greater diversity of foods and in different ways, and these trends challenge food surveys to keep up with consumption practices. The sourcing and production of food continues to evolve to meet consumer demands, which sometimes results in complex supply chains, particularly when weather events disrupt expected yields or timing of harvest. Tracing foods, especially perishable foods, back to their source is not easy.

Obviously, preventing contamination must remain the top priority for the food industry. Avoiding outbreaks obviates the need for a debate about traceability. But recognizing that outbreaks may occur from time to time, we need to reconsider how we use the tools at our disposal so that our ability to respond to outbreaks keeps up with our ability to detect them.   

Drew McDonald is the vice president, quality and food safety, at Taylor Fresh Foods, Salinas, CA. He has over 20 years of experience in fresh produce and fresh foods. He oversees the quality and food safety programs across the foodservice, retail, and deli operations. Drew received his education from Lawrence University in Wisconsin.   

Jennifer McEntire, Ph.D., is vice president of food safety and technology at United Fresh Produce Association. She was previously vice president of science operations at the Grocery Manufacturers Association. She has also had roles as VP and chief science officer at The Acheson Group and as the senior staff scientist and director of science and technology projects at the Institute of Food Technologists. Jennifer earned a Ph.D. in food safety from Rutgers University.