Food-Related Illness and Death in the United States

If you want to change something, you first have to count it. That’s the simple idea driving a classic analysis by Mead and colleagues on foodborne illness in the United States. When they stitched together the best surveillance we had, the picture that emerged was both enormous and oddly lopsided: about seventy-six million illnesses every year, three hundred twenty-five thousand hospitalizations, and five thousand deaths. Most people get sick and recover. Far fewer die. And among the deaths we can pin to known culprits, three names—Salmonella, Listeria, and Toxoplasma—do most of the damage, together accounting for roughly fifteen hundred lives each year. But here’s the twist. A surprisingly large fraction of that burden isn’t tied to a named pathogen at all. Known pathogens explain about fourteen million illnesses in that total; the other sixty-two million? Those are from unknown agents. Put differently, when you tally what’s clearly identified and what isn’t, the unknowns shoulder about eighty-one percent of illnesses and hospitalizations and about sixty-four percent of deaths. That doesn’t mean mystery monsters are on the loose. It means our surveillance misses a lot—mild cases that never see a doctor, viruses that labs don’t routinely test for, and causes that, even in a hospital, never get a precise label. So how did Mead’s team count what most of us can’t see? They built a surveillance mosaic. Start with FoodNet, an active surveillance network run by the Centers for Disease Control and Prevention with state and federal partners. FoodNet doesn’t wait for reports to trickle in; it goes out and pulls case data for a defined list of bacterial and parasitic pathogens from a defined population. It also calls people in that population and asks about diarrhea, whether they sought care, and whether a stool sample was ever tested. Layer on top of that the big passive systems that gather physician and lab reports nationwide, plus outbreak reports that link clusters of illness to a common food. Then bring in administrative datasets that tell you who showed up at a clinic, who was admitted to a hospital, and who died, along with the diagnosis codes assigned on discharge or on a death certificate. Now add one more idea that sounds simple but is critical: the surveillance pyramid. For every person who ends up in a lab database, there are many more who were sick at home, some who saw a doctor but never had a test, and some who had a test that wasn’t for the right pathogen. To climb back up that pyramid from what’s reported to what actually happened, Mead’s group used multipliers. For pathogens that usually cause nonbloody diarrhea, like Salmonella and Campylobacter, they multiplied reported counts by about thirty-eight to estimate what was missed. For those that often cause bloody diarrhea, like Escherichia coli O157:H7, closer to twenty. And for pathogens that typically land you in the hospital, like Listeria, the multiplier was much smaller, around two, because severe cases are more likely to be detected. Before they could say what share was foodborne, they needed to size the syndrome itself: acute gastroenteritis—diarrhea or vomiting from any cause. FoodNet’s nineteen ninety-six to nineteen ninety-seven population survey provided the backbone. People reported episodes, and when Mead’s team adjusted for respiratory symptoms that might masquerade as stomach bugs, the result was an average of zero point seventy-nine acute gastroenteritis episodes per person per year. Across the U.S. population in nineteen ninety-seven, that works out to about two hundred eleven million episodes annually. That’s the ocean. Foodborne illness is a big wave inside it. To figure out care patterns and severe outcomes in that ocean, they turned to physician and hospital surveys and discharge data. Pulling from the National Ambulatory Medical Care Survey and its hospital counterparts, and from the National Hospital Discharge Survey, they estimated roughly nine hundred thirty-six thousand hospitalizations each year for acute gastroenteritis, with about six thousand six hundred people dying in the hospital with a gastrointestinal diagnosis. Most of those deaths didn’t have a named cause—about three quarters carried undiagnosed labels—which tells you how often we treat the syndrome without ever learning the exact pathogen. Using those flows, and the pathogen-specific fractions that are known to be foodborne, they apportioned hospitalizations and deaths into the same known versus unknown buckets. When you cascade all those steps—community illness, care-seeking, lab confirmation, reporting, and attribution—you land back at the numbers we started with. About seventy-six million foodborne illnesses each year in the United States. Around three hundred twenty-five thousand hospitalizations. Roughly five thousand deaths. Within that, known pathogens account for about fourteen million illnesses, sixty thousand hospitalizations, and one thousand eight hundred deaths. Unknown agents make up the rest: roughly sixty-two million illnesses, two hundred sixty-five thousand hospitalizations, and three thousand two hundred deaths. The overall picture is a country living with a massive daily churn of nonfatal sickness and a smaller, concentrated death toll. Zoom in on the known pathogens and the landscape gets sharper. Norwalk-like viruses—what we now mostly call noroviruses—are the heavyweight in sheer illness. Mead’s team estimated about twenty-three million cases tied to these viruses, with roughly nine point two million transmitted by food. They lead to about fifty thousand hospitalizations and around three hundred deaths. That’s a lot of misery, but not a lot of mortality, which matches the lived experience: brutal forty-eight-hour bugs that rip through schools and cruise ships but rarely kill. If you shift to pathogens that send more people to the hospital or to the morgue, the names change. Nontyphoidal Salmonella sits near the top for known agents, with about one point four million illnesses and around sixteen thousand hospitalizations each year. It claims roughly five hundred eighty-two lives. Most of those infections are foodborne, crossed wires somewhere between farm and fork, which is why poultry, eggs, and produce safety get so much attention. Campylobacter is another workhorse of bacterial gastroenteritis. Mead’s estimates put it at about two point forty-five million illnesses annually, with around thirteen thousand hospitalizations and about one hundred twenty-four deaths. Taken together with Salmonella, you get a sense of where the front lines are: common, often self-limited infections that still put thousands in hospital beds and, in some cases, trigger longer-term problems like Guillain–Barré syndrome. Escherichia coli O157:H7, by contrast, is less common but notorious. Think of undercooked ground beef in the nineteen nineties or leafy greens in the two thousands. The team estimated about seventy-three thousand illnesses and a little over two thousand hospitalizations each year, with around sixty-one deaths. Small numbers next to norovirus, huge numbers if you’re thinking in terms of preventable tragedies and product recalls. Then there are the quiet killers. Listeria monocytogenes doesn’t infect many people—roughly two thousand five hundred illnesses in these estimates—but it hospitalizes almost all of them and kills around five hundred each year. It’s devastating in pregnancy and in people with weakened immune systems, and it’s the pathogen behind some of the most consequential ready-to-eat food recalls. Toxoplasma gondii, a parasite many of us carry without knowing it, shows up here not for garden-variety stomach upset but for the severe disease it can cause in fetuses and immunocompromised adults. Mead’s group estimated about two hundred twenty-five thousand infections annually, with on the order of five thousand hospitalizations and roughly seven hundred fifty deaths, about half of those infections and fewer of the outcomes attributed to food. That’s a sobering reminder that foodborne isn’t just about diarrhea; it’s about who gets exposed and what happens next. Other parasites contribute meaningful, if smaller, slices. Giardia likely causes on the order of two million infections each year, with a minority acquired through food; Cryptosporidium, too, appears in this bucket, reflecting the mix of water, person-to-person, and food transmission that defines many protozoal diseases. And for viruses beyond norovirus—rotavirus and astrovirus in children, hepatitis A in the broader population—the foodborne share is real but generally smaller, with hepatitis A showing a detectable contribution from contaminated foods in the era before widespread vaccination. All of these pathogen-specific estimates rest on the attribution framework we talked about earlier: what fraction of total infections are believed to come through food. That’s where uncertainty bites hardest. Mead and colleagues called out norovirus in particular. Small tweaks to the assumed foodborne fraction—say, nudging it up or down a few percentage points—can shift the national illness totals by millions. Yet even with that sensitivity, the broad pattern holds. Unknown etiologies make up the bulk of foodborne illness and hospitalization, and among the things we do recognize, a short list of pathogens dominate the deaths. It’s also where the data systems show their seams. Surveillance historically treated bacteria, parasites, and viruses very differently. Lab capacity for viruses lagged behind bacteria in the nineteen nineties. Some organisms, or exposures like marine toxins and certain metals, didn’t have good reporting pipelines at all. And because the whole exercise relies on the right multiplier for the right pathogen, an overestimate or an underestimate in one corner can ripple through to the national picture. That’s why FoodNet’s active case finding and its population surveys were so important: they anchor the pyramid in observed behavior—how often people get sick, how often they seek care, how often a sample is tested—rather than guesswork. How does this analysis stack up against older attempts to count the same problem? In the nineteen eighties, Archer and Kvenberg compiled what they could and landed on about eight point nine million illnesses from known pathogens, and a very wide band—twenty-four to eighty-one million—for all agents combined. Bennett and colleagues produced similar broad strokes. Mead’s numbers are higher on total illnesses and more precise overall: seventy-six million illnesses, three hundred twenty-five thousand hospitalizations, and five thousand deaths, with fourteen million illnesses from known pathogens and sixty-two million from unknowns. That’s not evidence that America suddenly got much sicker. It’s a reflection of better source data, new recognition of viruses like norovirus, and a more disciplined way to climb the surveillance pyramid. There’s a policy message lurking in here too. If the overall death toll is lower than some earlier estimates, that argues for targeted control against the pathogens that do most of the killing—Listeria in ready-to-eat foods, Salmonella across animal products and produce, and Toxoplasma in settings where exposure can be catastrophic. If the illness count is as high as it is, that justifies investments in outbreak detection, worker policies that keep sick people from handling food, and sanitation along the supply chain. And if the unknowns dominate, then the only way to shrink that dark matter is through sustained surveillance—more active systems like FoodNet, broader viral testing, and laboratory methods that don’t miss what isn’t on a standing order set. Mead and colleagues close with that point. The landscape keeps moving as new pathogens emerge, as industry practices evolve, and as our tools get sharper. But we’ve seen what success looks like. When you put rules in place—cooking standards, pasteurization, environmental monitoring for Listeria—and you back them with good data, illnesses and deaths fall. Counting isn’t glamorous. It’s how you find the signal, and it’s how you prove you moved it.