U.S. Should Improve Estimates of Illnesses Linked to Particular Foods

More accurate analyses of the sources of foodborne diseases help food safety agencies better target their efforts to prevent illnesses. However, federal researchers need to address methodological weaknesses that can make their findings less dependable and actionable.

Getty Images

Researchers for public health agencies have a vital and difficult task when estimating how many foodborne illnesses are associated with different food types—from fruits and vegetables to dairy, eggs, fish, and meat products. In December, the Interagency Food Safety Analytics Collaboration (IFSAC)—a partnership that includes the Centers for Disease Control and Prevention, Food and Drug Administration, and U.S. Department of Agriculture—released new illness estimates for 17 food categories and four major disease-causing bacteria. The data and methods used in the study demonstrate the complexity of the task, but the results also show that IFSAC’s approach needs to be improved.

This kind of research—known as foodborne illness source attribution—matters because the results inform the options that policymakers pursue to reduce infections from E. coli O157, Salmonella, and other microorganisms that can contaminate food. The more accurate the government’s estimates, the more effectively agencies such as FDA and USDA can target their strategies and resources to decrease the food safety risks that threaten Americans’ health.

IFSAC’s report examined data from outbreaks (i.e., multiple reported cases linked to a common food source) between 1998 and 2013 in which one of four bacteria was identified or suspected as the cause. Among other conclusions, the researchers found that:

• Salmonella illnesses were spread across a wide variety of foods, with seven categories—such as seeded vegetables and eggs—each contributing about 9 percent or more of the total.
• E. coli O157 infections were mostly attributed to beef and vegetable row crops, such as spinach.
• Listeria monocytogenes cases were often linked to fruits and dairy.
• Campylobacter illnesses were predominantly caused by chicken consumption.

Because FDA and USDA officials will use this information as they determine priorities for their respective food safety policies, inspection activities, and research, the estimates must be reliable. However, weaknesses in IFSAC’s methodology and data choices make some findings less dependable and actionable than others. For example:

1. Because of the scarcity of Listeria outbreak data, IFSAC’s estimate that 50 percent of foodborne illnesses from this bacterium were attributed to fruits carries a large caveat: According to the report, the actual share could range from 5 percent to 77 percent. Such uncertainty makes it difficult for food safety agencies to decide whether to focus more or fewer resources on efforts to prevent Listeria infections resulting from fruit consumption.
2. IFSAC’s analysis of Campylobacter outbreaks excluded those associated with dairy products including unpasteurized milk, which are significant sources of illness. The researchers noted that unpasteurized milk, although linked to many of the outbreaks, is not widely consumed by the general population. Furthermore, they explained that an analysis of sporadic Campylobacter cases implied that outbreak data over-represented dairy as a source of infection. But the decision to omit dairy-linked illnesses affects estimates of the percentage of illnesses linked to other food categories. The exact effect on each category’s estimate is not well described in IFSAC’s report and is difficult to infer, leaving policymakers with an incomplete and potentially incorrect picture of Campylobacter sources.

Alternative approaches for future source attribution analyses could help address these and other limitations. With Listeria, for example, using other available data could increase the validity of food source estimates, because only a small fraction of illnesses from this bacterium are part of recognized outbreaks. Attribution methods that include sporadic illness data in addition to outbreak data could provide a more representative estimate for Listeria’s foodborne sources.

For Campylobacter, analysts could include dairy-associated illnesses and adjust the estimates according to Americans’ consumption patterns. More fundamentally, a combined approach using both outbreak and sporadic illness data could help overcome a unique source attribution challenge for campylobacteriosis infections: The foods that have been linked to outbreaks are often different from those associated with isolated cases.

Encouragingly, IFSAC’s Strategic Plan for 2017-2021 indicates a continued commitment to explore and improve its analytical approaches, including development of methods to incorporate sporadic illness surveillance information. The Pew Charitable Trusts supports this goal and others outlined in the plan, including the use of enhanced methods for attributing illnesses to foods with multiple ingredients—something the most recent report did not attempt.

IFSAC researchers should collaborate with other leading domestic and international experts in source attribution, integrate multiple data sources to inform estimates, and more transparently communicate the assumptions used in their statistical models. Steps such as these would increase the robustness and reliability of the research and provide crucial information to strengthen food safety systems and reduce foodborne illnesses.

Karin Hoelzer, a veterinarian, works on The Pew Charitable Trusts’ safe food and antibiotic resistance projects.