I have been writing this blog since 2004. Something north of five thousand posts. Nearly all of them are about somebody getting sick from something they ate, and too many are about somebody dying from it.
Here is what I actually did yesterday on Labor Day. I reviewed a life care plan for a woman who will never walk again or care for herself, because of Listeria in deli meat. I worked up the lifetime costs for three children with hemolytic uremic syndrome, caused by E. coli in a romaine lettuce outbreak that FDA never told the public about. I spent an hour on the phone with a woman who is still testing positive for Salmonella after eating mayonnaise made with raw eggs, and who has had to delay her cancer treatment because of it. And I edited a dozen lawsuits to be filed in the ByHeart and Nara infant botulism outbreaks.
Four files, one holiday Monday. Cyclospora in lettuce, Salmonella in peppers and sprouts will wait until today. That is not a complaint. It is the argument for everything that follows.
We tell ourselves we have the safest food supply in the world. Maybe we do. No two countries count illness the same way, so the claim cannot really be tested, and in my experience, it gets used less as a fact than as a reason to leave things alone. What can be tested is our own number, and our own number has not moved.
In April 2025 CDC published its first full rebuild of the burden estimates since 2011, using 2019 data and seven major pathogens: 9.9 million domestically acquired foodborne illnesses, 53,300 hospitalizations, 931 deaths. Norovirus leads both illnesses, at 5.5 million, and hospitalizations, at 22,400. Then Campylobacter at 1.9 million, Salmonella at 1.3 million, Clostridium perfringens at 889,000, and STEC at 357,000. Salmonella kills the most, at 238.
CDC’s consumer page still carries the older 48 million figure, and it is not out of date. Four-fifths of that number was a bucket called unspecified agents — the gastroenteritis nobody ever identifies — and the new paper did not re-estimate it at all. Compare like with like and the picture is worse than a drop. In 2011 CDC put 31 known pathogens at 9.4 million illnesses, 55,961 hospitalizations and 1,351 deaths. Today six pathogens account for 9.9 million. Norovirus was about 5.5 million then. It is 5.5 million now.
That is the scoreboard, and an agenda is serious only if it is scored against it. Here is mine, ranked by what the data says rather than by what makes news.
1. Since it’s Labor Day — start with the people who handle the food.
Norovirus is the largest single line on the scoreboard and puts more people in the hospital than any other foodborne pathogen. CDC examined 800 restaurant outbreaks from 2017 through 2019: roughly 40 percent of those with a known cause involved a worker who came in sick, norovirus accounted for 47 percent of them, and only about 44 percent of managers reported offering paid sick leave. Eighty-five percent said they had a policy against working sick, which tells you the policy is not the binding constraint. The paycheck is.
The fix: paid sick leave for food workers, by state law where Congress will not act. It is the cheapest intervention available and aims at the biggest number on the board.
Two more belong in the same item, because a worker who cannot speak up and cannot get vaccinated is not much safer than one who cannot afford to stay home.
Whistleblower protection. FSMA gave food workers anti-retaliation rights in 2011, enforced by OSHA, and they are narrower than they sound. The complaint has to sound in the Food, Drug and Cosmetic Act, the filing window is 180 days, and workers in meat and poultry plants, which answer to a different statute, are largely outside it. Widen the coverage and lengthen the window. The people who see the problem first are standing on the line, and today most of them are better off saying nothing.
Hepatitis A vaccine. It is the only common vaccine-preventable foodborne disease in this country, and CDC still declines to recommend the shot for food service workers on the ground that they face no increased occupational risk. That answers a question nobody asked. They are not at greater risk of getting it. They are a risk of giving it, and about a quarter of infections are asymptomatic, so neither the cook nor the manager knows. St. Louis County has required the vaccine since 2000 and has not had a hepatitis A outbreak since. Famous Anthony’s in Roanoke had one infected cook: fifty-two sick, thirty-six hospitalized, four dead, six million dollars in acute medical bills and a chain in Chapter 11, over a shot that costs thirty dollars. I have written about this again this week.
2. Fix chicken.
Campylobacter and Salmonella together are roughly a third of the scoreboard, and Salmonella kills more Americans than any other foodborne pathogen. Interagency attribution work puts more Salmonella illness on chicken, at 19.1 percent, than on any other food. FSIS determined in 2024 that Salmonella at 1 colony-forming unit per gram in not-ready-to-eat breaded stuffed chicken is an adulterant, and it is still law. In December 2025 FSIS indefinitely delayed the sampling that would enforce it.
The fix: enforce the rule FSIS already wrote, extend it to chicken parts and ground poultry, and publish the validation data behind both positions. I drafted a FOIA asking for exactly that. Then make the worst FSIS performance category disqualifying for federal poultry purchases, with a defined path back. A purchasing specification takes a phone call. Rulemaking takes a decade.
3. Fix leafy greens, including the imported ones.
Attribution puts 67.9 percent of E. coli O157 illness and 27.4 percent of Listeria illness on vegetable row crops. That is lettuce and spinach, increasingly grown outside the United States. Guanajuato iceberg and Sinaloa jalapeños both landed here this summer.
The fix: pre-harvest agricultural water is the known route, and those standards have been delayed twice. Finish them, fund foreign inspection and FSVP audits, and restore the enforcement money Congress restricted for the Produce Safety Rule.
4. Fix the two recurring Listeria sources: dairy and deli.
Dairy carries 31.9 percent of attributed Listeria illness, and ready-to-eat deli meat is the other source that keeps coming back. It is why I spent part of Labor Day reading a life care plan. CDC researchers calculated that unpasteurized dairy causes 840 times more illnesses and 45 times more hospitalizations than pasteurized, out of a sliver of consumption. I am handling raw goat milk and raw milk cheese cases right now.
The fix: enforce the existing interstate raw milk ban, stop the state-by-state expansion of herd shares and retail sales, and require ready-to-eat processors to publish environmental sampling results. Adults may decide for themselves. A four-year-old on dialysis decided nothing.
5. Count the sick.
FoodNet is the only active foodborne disease surveillance we have, and on July 1, 2025 it dropped from eight pathogens to two. Cyclospora and Listeria were both cut. Barb Kowalcyk put it more precisely than I would have: it was never funded well enough to carry eight in the first place.
The fix: fund it for eight and fund the state labs that feed it. CDC asked for roughly 72 million dollars for food safety in fiscal 2026. You cannot take down a number you have stopped counting.
6. Keep the isolates.
Clinical laboratories have moved to culture-independent tests. They are faster and better for the patient, and they produce a result without producing a bacterial isolate. PulseNet needs the isolate. No isolate, no genome, no cluster, no outbreak. CDC says plainly that without isolates PulseNet cannot keep detecting clusters, and clinical microbiologists have called the transition a threat to foodborne disease surveillance since 2016. Whole genome sequencing is how the patients in that romaine outbreak were tied to one another in the first place. That only works if somebody cultured the specimen.
The fix: require reflex culture and isolate submission on CIDT-positive specimens for the PulseNet pathogens, and pay the laboratories for it. The unreimbursed cost is the real obstacle and it is a small one. Counting the sick is worth very little if you cannot tell which sick people belong together.
7. Turn the traceability rule on.
Traceability does not prevent the first illness. It prevents the three hundredth, and it keeps fear from pricing a whole category when the problem sits in one facility. Congress set a January 2013 deadline. The rule issued in 2022 only after a nonprofit sued and a judge set a schedule, compliance was pushed to July 2028, and Congress barred FDA from enforcing it. A Congressional Research Service report this January, citing a Government Accountability Office finding, says FDA still has not built the internal tracing system the statute required. Deadline to enforcement: fifteen and a half years.
The fix: repeal the funding prohibition and build the system.
8. Tell the public what you know, when you know it.
The romaine in my opening paragraph was never announced. Eighty-nine sick across fifteen states, thirty-six hospitalized, seven with hemolytic uremic syndrome, one dead, and in January 2025 FDA and CDC closed the investigation recording the vehicle as romaine from a common supplier. No grower, no processor, no advisory. FDA’s position is that it names firms when there is actionable advice to give consumers, and that the product was gone before investigators confirmed the source. Those families learned what happened from a lawyer. It took a records request and our litigation to get the names out from behind the gray boxes.
The fix: name the firm and the product when the agency knows, even after the product is gone, because the public record is what makes the next one preventable. Publish retail consignee lists the day a recall is announced. Transparency is the cheapest enforcement there is, and an item here that needs no new statute.
9. Fix GRAS and be honest that it is a different number.
Self-affirmation lets a company decide privately that a novel substance is safe and never tell FDA. That is how tara flour reached Daily Harvest Crumbles. We represented 339 of the roughly 470 people sickened, and nearly forty of them lost their gallbladders. FDA’s proposed rule uses tara as its lead example, and it is progress that stops short: mandatory notification is not premarket approval, and a company may still ship first and notify later.
The honest part: the scoreboard above is pathogens. Chemical exposure is a real problem measured in chronic disease over decades, not in emergency rooms this week. Both deserve fixing, and neither should be funded out of the other’s budget.
10. Set the target and make somebody answer for it.
Federal reduction goals exist for four pathogens. None has been met, and Campylobacter and STEC appear to have gone the wrong way. Nobody loses a job over it, because no single official owns the number. Thirteen members of Congress wrote to FDA, CDC and HHS about this summer’s outbreak. Not one published answer.
The fix: one published national figure, one annual public accounting, and one agency answerable for it. I have argued for a single food safety agency for twenty years without success, so here is the fallback — pass the bills already written. H.R. 7867, the Infant Formula Safety Modernization Act, is one of them, and I have said so to the Energy and Commerce Committee.
What is not on this list
Plenty. Chronic disease and diet are not here, and they belong in any honest accounting of what food does to Americans. Neither is antibiotic resistance, or the safety of the people who work in the plants and the fields. I picked these ten because they are the ones where the evidence is clear, the fix is known, and the obstacle is a decision rather than a discovery. I would rather be argued with about the ranking than be vague about the method.
The method is the whole point, and it is not complicated. State the problem in plain terms. Propose a solution a reasonable person can defend. Execute it. Then measure whether the number moved. That work does not belong to any one of us: industry, government and consumers all have to want a fact and science based answer more than they want to win the argument about the last outbreak. Almost none of the last thirty years of that argument has been about preventing the next one. Reactive is how we got a number that has not moved since I started writing.
None of this is exotic. E. coli O157:H7 became an adulterant in ground beef in 1994, and that one decision took a large part of my practice away. I have never been happier to lose business, and it is still the only item on this list that has actually been done.
Twenty-two years. Five thousand posts. Four files on a holiday Monday. Do something, and one of these years I will take Labor Day off.