Studying supply chains to predict food safety failures

Sink your teeth into this: much of what we eat comes almost exclusively from other countries — like 80 percent of the honey you stir into your tea, and 90 percent of the shrimp in your cocktail. And more often than not, it’s coming from developing countries that lack the basic quality controls we have here at home.

In 2015, the United States imported more than 40 million individual shipments of food. The Food and Drug Administration has the resources to check only a very small fraction of those shipments to ensure they do not have harmful bacteria, contaminants, or additives. “They have to rely on risk models to make decisions,” said Retsef Levi, a professor of operations management at MIT Sloan. “On the one hand, they can’t allow harmful stuff to penetrate our food supply. On the other hand, they can’t delay commercial activity. That’s a tension the FDA has to balance, and it’s a very demanding and challenging task.”

Three years ago, FDA officials engaged MIT in a contract to support and improve the risk management system of the FDA related to economically motivated food adulteration.  A team led by Levi drew from MIT Sloan, the MIT Center for Biomedical Innovation, and across campus. With additional funding from MIT’s Abdul Latif Jameel World Water and Food Security Lab, the team has conducted a systematic study of how food is grown, processed, and shipped to the U.S. — and of the risks involved. 

Now, even as the team makes its recommendations to the FDA, it’s delving deeper. In a new multiyear, multimillion dollar partnership with the Walmart Foundation, Levi and colleagues are launching a new study of supply chains in China, the world’s third largest food exporter. Recently, Levi explained the quest to protect our edible imports.

What’s risky about the way our global food supply chains work today? There’s a continuum of risks. Food can become contaminated with bacteria, like listeria or salmonella. That’s typically the non-intentional scenario, and that happens when you don’t have good practices or are negligent. The other extreme is terror-motivated, when an organization or individual has malicious intent like putting anthrax into the food system. And in between, there’s the economically motivated scenario. That’s mostly about financial benefits of some sort rather than intentionally causing harm. The FDA — and governments around the globe — are worried about all of the above.

What’s the worst thing that could happen? There have been incidents. Several years ago, in 2008, there was a big scandal around milk in China. That came about due to regulatory pressure. The Chinese government started to require a level of protein in milk, which prompted farmers and other stakeholders to add the compound melamine, which is essentially a poison. In tests, the melamine faked the level of protein. That ended up in baby formula. It adversely affected hundreds of thousands of babies in China, with some serious kidney and neurological effects and six deaths. That’s one example.How did you wrap your brains around this complex challenge?

We were able to assemble a very multidisciplinary team at MIT including Yasheng Huang, Tauhid Zaman, and Yangchong (Karen) Zheng, all at Sloan. We’re collaborating with Anthony Sinskey, Stacy Springs and Michael Strano from the Center for Biomedical Innovation, and people from electrical engineering and computer science. It’s a big group of faculty that over the last three years has taken a new approach to thinking about risk. We were able to create predictive risk models of how different product categories are vulnerable to what extent of risk, and which are more likely to be engaged.

What did you discover? Our research indicated that a key issue to understand is the risk related to the product category and its supply chain. We know that, for example, the melamine case was motivated by external pressure on the milk supply chain — regulatory pressure. In other cases, our research focused on poultry in China, and there, in response to outbreaks of avian flu, there was an increase in extensive use of antibiotics, antiviral, and herbal medicines to fight the outbreak. In the shrimp case, we’ve seen that a disease called early mortality syndrome killed shrimp in Asia. There was big outbreak in 2009, and in again 2013, that lead to increased use of antibiotics and other additives. All of these cases indicate that you might be able to assess whether risk is likely to increase if you understand what kind of socioeconomic and environmental drivers could increase the level of risk. Our research gave rise to a much more comprehensive, holistic understanding of the risk factors of a product category.In addition, you have to look within the product category. We were able to look at massive amounts of shipping data. We looked at bills of ladings — papers that shippers have to file with the weight of shipment, content — and used them to reconstruct the different layers of supply chains. That kind of info has pointed out companies that are more likely to be engaged in the adulteration of food. If you look at economically motivated adulteration, about intentionally doing something, people who are intentionally involved behave in similar patterns. We can pick up those patterns and signs.

What are some of the signs? They might use a third-party country and cross in and out of borders. They tend to ship with a lower number of shipments, or maybe have many transient partners.Even the very structure of a supply chain is predictive of risk. For example, in China you can have “dragon head” companies, which are large companies contracting with thousands of household farmers. Each farmer is going to bear a high level of risk. That’s a big issue in the sense that if they are under pressure, they are going to do whatever it takes, including engaging in bad practices, like giving chickens antibiotics or chemotherapy drugs. We have mapped in China well over 95 illegal drugs farmers are buying publicly and using. There are cases where one chicken had residuals of 19 types of antibiotics. We see similar phenomena in other farming supply chains and this is suggesting that the fundamental structure of the farming supply chain is a major risk factor.

Then there are the importers. In 2011, President Obama signed the FDA Food Safety Modernization Act. It aims to make the way food products are managed more similar to pharmaceuticals. In the past, importers dismissed responsibility; in the near future they will not able to get away with that. They will be held accountable to verify their supply chain. Our research indicates that that’s a key aspect of being able to manage risk.
 

What happens next? Now our project is just about to be done. We’re transitioning the tools and models we have developed to the FDA. At the same time, the Walmart Foundation has started a food safety innovation center in China this fall. In doing their due diligence, they met our team and saw that we were doing some unique work in this space. So we’re launching with them an even more ambitious project to go underground and understand to a greater degree of detail the structure of supply chains, the dynamics, and the way they are regulated, and to leverage our existing knowledge to be able to be really informed. And we’re going to complement it with new concepts of how to test food. The current approach is based on testing each time for a specific compound, and you need to know what it is. That is a very big problem in food safety. The range of contaminants is quite unbounded. The only way to address that is with a systematic approach to actively predict what compounds are likely to be present, and couple that with testing that is far more robust and can detect multiple compounds at the same time. We really are able to integrate disciplines that reside in different parts of the university.