DHS software applied to poultry processing
What works for the Department of Homeland Security can be applied to control risks in the food industry.
Battelle has been very active for the past 10 years developing terrorism risk assessment methods for the Department of Homeland Security. Out of this research came a software modeling program designed to chart responses to incidents ranging from biological and chemical attacks in public gatherings to the purposeful contamination of a public water supply. Researchers have been able to apply this software, Probabilistic Risk-Informed Analysis, in the poultry industry to focus on the challenges of minimizing Salmonella contamination and understanding where processes are susceptible to Salmonella growth.
FE spoke with Dr. Brian Hawkins, Battelle senior research scientist, to learn how PRIA works in the poultry industry and how it might be applied to other food industry segments that battle Listeria, E. coli, Campylobacter and other dangerous contaminants.
FE: Can you explain how PRIA came to be?
Brian Hawkins: Going back about 10 years, Battelle began work for the Department of Homeland Security to assess the risk of chemical, biological, radiological and nuclear terrorism against the US. Evaluating this risk was really a difficult task because there wasn’t an existing pool of data like there is for other relatively common risks. For example, if you were trying to assess the risk of a fatal car accident in any one location, you’d have a ton of data, and you could look at the last one or two decades’ worth of data and estimate the likelihood of such an accident and its consequences and determine the risk very easily.
But terrorist attacks against the US homeland are quite rare, so you don’t have the data to make a good assessment. Therefore, you have to construct models to simulate potential attacks and look at other ways to quantify the likelihood of one of those occurring. Battelle adopted a Probabilistic Risk Analysis approach, which was researched and developed for the nuclear power industry back in the ‘70s. We applied that approach to help DHS make informed decisions, create policies and decide where the US Government should allocate its limited funds. For example, we can’t buy every possible antibiotic or countermeasure; we can’t invest in every type of chemical countermeasure; we can’t put in detectors for every possible threat. DHS has to make some risk-informed decisions regarding priorities because the amount of funding is limited.
We looked at modeling attacks for a lot of different venues including something being dispersed in the air—either indoors or outdoors—in municipal water or in the food supply through various scenarios. As we began doing that work for DHS, we transitioned from a static report to a dynamic software tool that DHS analysts could use to field questions quickly and analyze potential risks. We went from what you might think of as a status report to providing a real-time, software-based capability.
With more recent efforts on the food front, Congress passed the Food Safety Modernization Act, which calls for the industry to take a risk-based approach for food defense. As we looked at the food industry, we found there are some approaches it has used canonically for food safety and defense, but they aren’t as quantitative as our approach. The approaches are sound and involve a lot of experts, but all-in-all, they’re mostly semi-quantitative or maybe even qualitative. For example, they determine only a low, medium or high level of risk.
We decided we could take the work we did for DHS and adapt it to put food safety and defense on the same platform—the same basis—and provide a quantitative level of risk so the food industry could quantify the impact of their risk mitigation decisions. These decisions could include starting an initiative on a farm to reduce contamination of the chickens coming in [for processing], using a new antimicrobial dip or a change to processing—maybe an immersion chiller vs. an air chiller—looking at reducing temperature variations in distribution and retail or trying to improve food preparation in the home to eliminate cross-contamination or undercooked and dangerous food.
We started development in April 2013 on a software-based tool for poultry because Salmonella in poultry is a very prevalent problem, causing a fair number of illnesses and recalls each year. Pound for pound, it is the riskiest protein product.
We developed this software tool—a proof of concept, if you will—for poultry and then began talking to poultry processors in the US and abroad to make sure it would integrate with their existing systems. We want to pull in their live data streams and work with that data. For example, they know the weight of the birds coming in, the temperature in their scalders and the process. So, if they incorporate the software into their operations, it can provide them with the capability to determine outcomes. PRIA will provide results so processors can weigh variables on a quantitative basis and perform a cost-benefit analysis.
Processors can’t spend money willy-nilly trying to fix all the process variables at once because the moment that chicken costs $20 a pound, nobody will buy it. We’re trying to give them a quantitative basis so they can be informed, weigh their decisions and determine the priority actions to take.
FE: So, you can come up with a quantitative comparison, say, between the processing of air-chilled and water-chilled birds?
Hawkins: Yes. It’s a modeling approach that provides a quantitative prediction. You can look at the data that exists on modeling cross-contamination that might occur in different types of chillers and the temperature of the chicken [with the Salmonella on the skin], and the software can determine the cross-contamination impact, as well as the growth or decay of the Salmonella based on the temperature.
FE: Would it be possible to determine the risk of not sterilizing poultry pens or cages on the farm after a flock change?
Hawkins: Right now, we’re not modeling the farm; we’re modeling forward from the point where the chickens hit the processor. I know there are a number of USDA efforts to look at different ways to reduce contamination on the farms—controlling boundaries, making sure you have best practices in cleaning, inoculations, all those things. That’s being looked at across the globe, especially in Australia, England and China. They’re trying to limit the contamination problems they have on the farm. PRIA doesn’t go quite that far, although it possibly could. We just haven’t concentrated there at present. We’ve been focusing on trying to capture a lot of data at the producer-controlled processing steps.
FE: There have been a lot of complaints at the processing stage about the Salmonella levels of incoming birds, haven’t there?
Hawkins: Yes. Through data that’s been studied on incoming birds to be processed, we could set operational targets. With this knowledge, steps could be taken at the farm level to reduce the level of Salmonella. Many cleaning and everyday initiatives—good agricultural practices—can be adopted on the farm to reduce contamination, and the results can be quantified as the birds enter the plant. A lot of that has to do with feed withdrawal, how close to harvesting you feed the chickens. A lot of the contamination spreads through the gut.
FE: Are there other areas where this technology could be applied?
Hawkins: Absolutely. It could be extended to basically any food and contamination source, e.g., chemical, biological, radiological. The latter’s not too much of a concern in food at this time, although with Fukushima, it could potentially become an issue. But, the technology could be extended to any food sector relatively easily with a little bit of work—definitely doable. I think the ones that make sense would be the common food safety problems: Salmonella, Listeria, Campylobacter and E. coli are good examples. You could consider chemical contamination as well; there was a recent incident with poultry from Chile that had dioxin contamination.
FE: And there’s also been the discussion in the press about chicken that had high arsenic levels because it was used in feed as a prophylactic.
Hawkins: Any of these contaminations can be studied, and I think the question becomes: Which are the more impactful? You really have to decide where to put the effort—that which will keep the most people the healthiest. We started with Salmonella in chicken, but it could be extended to pretty much any food with any potential contaminant.
FE: Does this tool reach into the supply chain as well?
Hawkins: Absolutely. The tool can be customized to determine the product flow from processing facility to distribution center to retail outlets. Then, using basic data and assumptions on how far people drive to go to a grocery store, we can go by the projection or an estimate of what homes are going to be affected and where you might see illnesses. We can start to look at illness clusters compared to public health data or possibly use customer complaint data. If you have detailed tracking, we could update it in real time.
FE: In the food industry, we are trying to get to one-step back and one-step forward electronic track and trace, but some people are still using Excel spreadsheets, right?
Hawkins: Exactly. We’re trying to help the industry move away from Excel-based systems. Anybody who does risk calculations now pretty much uses Excel, possibly with a plug-in. You’re talking about a team of Ph.D. statisticians who calculate risks for a processor. You mentioned tracking. Processors are trying to implement track and trace now, and that’s really an important thing to do, but it has a lot of sensitivities because companies are trying to protect their supply chain data, not provide a competitive advantage to their various ingredient suppliers. That’s a tough challenge.
FE: You mentioned briefly FSMA, which is more proactive than older HACCP methods. Could this tool be used to help along those lines? What about preparation for GFSI certification?
Hawkins: Chicken, of course, isn’t covered by FSMA because FSMA is FDA. But it’s still absolutely relevant. I think the rules are still being commented on, and it’s not clear exactly what FSMA compliance will look like. To our knowledge, and from what we’ve seen so far, we take a more rigorous, documented approach. So, I would have to imagine our tool will go a long way toward documenting compliance with FSMA, providing an underlying quantitative mathematical basis for GFSI certification and things of that nature.
Some of the poultry companies (especially international) have expressed that the PRIA tool would have helped them with some of their past contamination incidents, where they were in court documenting their safety processes, how they look at the risk and what it was. PRIA’s mathematical rigor, paired with some of the microbiological data they gather, would really have helped at that time.
FE: PRIA could go a long way if it could be adapted for other products. For instance, the Chinese melamine situation. The melamine was intended to boost the protein levels in dog food or whatever it was added to and increase the perceived value of the product. But it made animals and people sick. How would your software tool be applied to this type of situation?
Hawkins: It could definitely be applied to that type of contamination. But a certain amount of data would be needed including dose-response or the potential health effects of something that’s not well known. What meets a certain specification in China might not meet the specification here. So you might have an acceptable level in China, but it’s not acceptable here.
Recently, a terrible tragedy occurred in India when someone used a drum that had contained pesticide as a container for cooking oil, and a lot of people got sick and died. The software could be useful for helping to prevent any of these types of incidents—intentional, unintentional, EMA or otherwise. But again, we have to set priorities in terms of what we address.
FE: Is this software available now as a commercial product?
Hawkins: Yes, it is. To mitigate the hardware requirements, it’s supported as a web-based cloud solution, so all you need is an Internet connection and a Google Chrome browser. Users coordinate with us to get a log-in, and they have their own app in the cloud, so they are protected. It uses a SQL database and has all the security and encryption of a Microsoft Azure cloud. Users can do simulations and look at the results. Right now, some customization would be required, such as hooking up temperature data streams or connecting the microbiological data that people gather.
We [currently] have one template for tray-pack chicken. However, the specifics of processing [chicken] definitely change across the globe—for example, the prevalence of immersion chilling vs. air chilling—and, therefore, those operations are very different depending on the country. We’re working with a couple of companies, both domestically and internationally, to make sure the system will integrate into their systems and operations seamlessly, and it will have the impact we expect.
FE: Some of the large processors are more technologically advanced and are using track-and-trace systems and popular database software. Can PRIA integrate with these systems?
Hawkins: Absolutely. PRIA software already predicts both a level of contamination and a physical location for each package of chicken. It would take a little effort, but integrating PRIA with a track-and-trace system to provide that location rather than estimate it based on distribution patterns is definitely doable. The key to this integration is that PRIA already tracks a physical location; we would just need to marry up the track-and-trace system to provide that input to the software.
FE: It wouldn’t be much of a stretch to translate this to E. coli and the beef industry, then?
Hawkins: Not at all. It’s relatively straightforward. Right now, we’re focused on tray-pack chicken, but you could be looking at other products like ground turkey. There are endemic problems with contamination in this area, as well as ground beef and E. coli. There have been some truly massive recalls of ground beef, and typically they’ve been late enough that you have to think most of that beef has already been consumed [by the time of the recall].
FE: What do you see as the future for PRIA?
Hawkins: In the next five years, I seePRIA being used to inform food safety and defense decisions at companies across the globe, from Campylobacter in Australia to Salmonella in the states. I see PRIA being used to estimate the potential impacts of antibiotic-resistant strains of various organisms, quantify the potential impacts of unintentional chemical contaminants and inform epidemiological investigations. I see PRIA transforming the vast amount of raw data in the food industry into information that can inform decisions that prevent illnesses and save lives.
About Brian Hawkins Brian Hawkins, Ph.D., is a chemical engineer at Battelle, a worldwide, independent research and development organization. He specializes in quantitative probabilistic risk assessment), subject matter expert) elicitation, mathematical modeling and building air quality studies. Hawkins has served as the program manager and principal investigator for chemical terrorism risk assessment and chemical infrastructure risk assessment for the Chemical Security Analysis Center of the Department of Homeland Security. Hawkins holds a bachelor’s degree in chemical engineering with a specialization in environmental engineering from the Ohio State University. He also holds a doctorate in chemical engineering from the University of Wisconsin-Madison.
For more information: Katy Delaney ; 614-424-7208; firstname.lastname@example.org