Foodborne Illness Acquired in the United States—Major Pathogens

If you want to steer public health, you have to know where the road actually is. With foodborne illness, most of the road is hidden. People get sick at home, they ride it out, maybe they tell a friend, but they don’t file a laboratory report. So the basic question is deceptively hard: how many people in the United States get sick from food each year, and which pathogens matter most? Scallan and colleagues took on that invisibility problem. They focused on 31 pathogens and, crucially, on illness "domestically acquired," meaning infections people picked up in the United States, as opposed to while traveling abroad—and then asked: of those, what fraction came from food? The headline is blunt. Each year, those 31 pathogens are estimated to cause about 9.4 million foodborne illnesses, leading to fifty-five thousand nine hundred sixty-one hospitalizations and one thousand three hundred fifty-one deaths. That’s just the foodborne slice. If you broaden the lens to all transmission routes for these pathogens—person-to-person, water, animals—the total climbs sharply. The team’s best estimate is thirty-seven point two million illnesses overall, of which about thirty-six point four million are domestically acquired. Think of it this way: the agents we worry about in food also hurt us through other routes, but the foodborne part alone is a country's worth of sick days and a city’s worth of hospitalizations. It’s a lot of suffering that we can, at least in principle, prevent. Counting those cases isn’t as simple as tallying lab reports. Only a sliver of illnesses ever show up as laboratory confirmations. So Scallan and colleagues built two complementary ways to find the missing cases. One approach starts at the laboratory and scales up. For pathogens that are routinely detected—say, nontyphoidal Salmonella—they begin with confirmed cases from surveillance and then climb a "pyramid" of multipliers to estimate how many people were actually ill out in the community. The other approach starts from the population and scales down. For pathogens like norovirus, where lab confirmation is rare, they rely on community surveys and published studies to estimate how much illness is out there. Then they apportion that burden to foodborne transmission. Let’s walk that pyramid for a moment. The lab-anchored estimates pull in counts from several surveillance systems. These include the active Foodborne Diseases Active Surveillance Network, known as FoodNet; national passive systems like the National Notifiable Diseases Surveillance System; organism-specific systems such as the Cholera and Other Vibrio Illness Surveillance; outbreak data from the Foodborne Disease Outbreak Surveillance System; and, for Mycobacterium bovis, the National Tuberculosis Surveillance System. Where multiple sources overlapped, they favored FoodNet’s active case finding, except for pathogens like Vibrio that cluster outside FoodNet’s footprint and are better captured by their dedicated system. That’s the bottom of the pyramid. To climb to the top—how many people were really sick—they multiplied by a set of pathogen-specific factors that correct for the ways illness drops out of the pipeline: people don’t always seek care, they don’t always submit a specimen, laboratories don’t always test for the right pathogen, and the tests aren’t always sensitive. Those care and testing steps matter. In their model, a patient with bloody diarrhea—used here as a stand-in for more severe illness—sought care about thirty-five percent of the time and submitted a stool sample about thirty-six percent of the time. For nonbloody diarrhea, care seeking fell to about eighteen percent and stool submission to about nineteen percent. That’s a lot of potential cases that never even reach a lab bench. Now add laboratory practice. Not every lab runs every test. Testing coverage for particular pathogens can range from a minority of labs to nearly all of them. Test sensitivity can be as low as roughly a third or close to perfect, depending on the organism and method. The upshot is simple: a confirmed case is the tip of a tall, leaky pyramid. For a handful of clinically dramatic infections, they modeled behavior differently. If you have Listeria monocytogenes sepsis, Trichinella from undercooked game, or Vibrio vulnificus from raw oysters, you almost certainly get medical care. The team assumed care seeking and specimen submission at one hundred percent for Mycobacterium bovis and hepatitis A and at eighty percent for those other severe bacterial and parasitic infections. That makes sense—nobody toughs out listeriosis at home—and it changes the multiplier math. What about pathogens that show up mainly during outbreaks? There, Scallan and colleagues leaned on outbreak surveillance but recognized a major blind spot: most cases are sporadic and never linked to a known outbreak. To bridge that gap, they used an underreporting multiplier of twenty-five point five, essentially saying that for every outbreak illness we see, roughly two dozen went unreported through the outbreak system. And for pathogens that almost never get confirmed individually—norovirus is the classic case—they flipped to that second architecture, the population-down approach. They anchored norovirus to its share of acute gastroenteritis in community surveys, about eleven percent. Then they used FoodNet’s population data to estimate total norovirus illness, later apportioning the domestically acquired fraction that’s actually foodborne. Two more pieces round out the scaffolding. First, because the focus is domestically acquired disease, they separated illness picked up abroad by incorporating travel history and other data. What remained counted as U.S.-acquired. Second, they estimated the proportion of domestically acquired illness that’s truly foodborne. That attribution draws on surveillance, risk-factor analyses, and literature, and it’s allowed to vary by pathogen. It probably varies by age and setting as well—a daycare norovirus outbreak spreads differently than a cruise ship buffet—but the available data aren’t always rich enough to model all of that explicitly. All of these ingredients carry uncertainty, so the authors treated them as distributions rather than fixed numbers. They used a Bayesian-style simulation with one hundred thousand iterations to propagate uncertainty from each multiplier through to the final counts, and they reported posterior means with ninety percent credible intervals. One last anchor point: the estimates are tied to the 2006 U.S. population, about two hundred ninety-nine million, with data spanning roughly 2000 through 2008. That keeps the math internally consistent and the uncertainty honest. So, what’s actually making us sick? One virus towers over the rest. Norovirus accounts for about five point five million of those foodborne illnesses—that’s fifty-eight percent of the total from the 31 pathogens. After that, two bacteria—nontyphoidal Salmonella and Clostridium perfringens—each contribute on the order of one million cases, and Campylobacter adds about eight hundred thousand. The pattern is clear and a little humbling. A virus that many people dismiss as a "stomach bug" is the biggest driver of foodborne illness, by far, while a small set of bacteria rounds out most of the rest. But the picture flips when you look at severity. Across all transmission routes, these pathogens send about two hundred twenty-nine thousand people to the hospital each year; when you look only at the foodborne slice, it’s fifty-five thousand nine hundred sixty-one. And hospitalizations are dominated by different names. Nontyphoidal Salmonella accounts for roughly thirty-five percent of foodborne hospitalizations. Norovirus, despite its huge case count, comes next at about twenty-six percent. Campylobacter follows, and Toxoplasma gondii also contributes an important share. This isn’t just a ranking change—it tells you that the pathogens causing the most trips to the bathroom aren’t always the ones filling hospital beds. Deaths sharpen the split even more. The best estimate is that one thousand three hundred fifty-one deaths each year are due to foodborne transmission of these thirty-one pathogens, out of about two thousand six hundred twelve deaths from these pathogens across all routes. Here again, nontyphoidal Salmonella leads, responsible for about twenty-eight percent of foodborne deaths. Toxoplasma gondii is close behind at around twenty-four percent. Listeria monocytogenes, the pathogen notorious for contaminating ready-to-eat foods and causing devastating infections in pregnancy and older adults, is also a major contributor. Norovirus, while widespread, takes a smaller slice of the mortality pie. The big lesson is that "most common" and "most dangerous" are not the same category. Before we leap from numbers to action, a few guardrails. Scallan and colleagues are upfront that these estimates rest on surveillance systems with imperfect coverage and sensitivity. Most people who get diarrhea never see a clinician, much less a lab test, so the model depends on multipliers for care seeking, specimen submission, laboratory testing, and test sensitivity. Those multipliers aren’t pulled from thin air—they come from FoodNet surveys, lab-practice audits, and the literature. But they do carry real uncertainty. The foodborne attribution fractions add another layer of fuzziness; for some pathogens and age groups, data are sparse or unrepresentative. And because the methods in this analysis differ substantially from the United States’ 1999 burden estimates, you can’t line up the two sets of numbers and read a trend. They’re portraits painted with different brushes. International comparisons are tricky for the same reason. So how do you use a portrait like this? You match the intervention to the burden profile. If norovirus drives the bulk of illnesses, that argues for stronger prevention where it spreads: in restaurants and food service, in retail environments, and in the everyday hygiene of people who prepare food at home. That can mean strict sick-worker policies, better hand hygiene, and attention to environmental contamination in kitchens. If nontyphoidal Salmonella, Listeria, Toxoplasma gondii, and Campylobacter dominate hospitalizations and deaths, that points to targeted controls along supply chains. These include on-farm biosecurity, processing-plant sanitation, temperature control, and consumer education for high-risk foods—plus protections for groups most likely to suffer severe outcomes. Those are different levers. The value of these numbers is helping agencies and industry pull the right ones. There’s also a budgeting angle. Public health is always about trade-offs. Knowing that two or three pathogens account for most severe outcomes lets you justify investments in specific surveillance enhancements, outbreak response capacity, or regulatory standards. And because Scallan and colleagues carried their uncertainty into every estimate, you can track changes over time within this framework. You can ask if new policies are moving the credible intervals in the right direction. It’s not perfect, but it’s a common ruler. Could we do even better? Almost certainly. Better data streams would shrink those multipliers and their uncertainty. Electronic laboratory reporting and culture-independent diagnostic tests can widen the window into community illness. That is, if they’re paired with smart ways to attribute cases to foods. Genomic subtyping can link sporadic cases that once looked unrelated, tightening the connection between a patient and a product. And for attribution—the hardest part—more focused case control studies in different age groups and settings could sharpen the foodborne fraction pathogen by pathogen. That’s the wish list. The core findings stand without it. Let’s leave it with the picture we started with. In any given year, in the United States, roughly nine point four million people get sick from food because of a known set of thirty-one pathogens. A virus most of us have had and never named causes the majority of those illnesses. A smaller cast of bacteria and a parasite you can’t see in the grocery aisle account for a disproportionate share of hospitalizations and deaths. And we have a map—built from surveillance, multiplied thoughtfully, and bounded by uncertainty—that tells us where to act first. As Scallan and colleagues showed, when you can finally see the road, you can drive policy with purpose.