Ohio Alone Has Now Reported Sixty-Four Times What It Reported in All of 2025. The Fifty-State Cyclospora Count Is 26,755, and the Federal Number Has Not Moved Since Wednesday.

The fifty state health departments now report 26,755 cases of cyclosporiasis this season, up 919 from 25,836. Two states moved and both are inside the federal outbreak. Ohio went from 4,013 to more than 4,900, and Arkansas from 205 to 237.

The Ohio figure deserves a moment. The state health department’s dashboard shows more than 4,900 cases confirmed since the beginning of the year, of which 3,399 were confirmed between July 4 and August 4 alone. In all of 2025, Ohio confirmed 76. That is not a bad year. That is sixty-four times a normal year, in a single state, in about five weeks. Ohio’s hospitalizations have also moved, from the 96 the state reported in mid-July to 172 now, and the department says more than a hundred people have been hospitalized there in the past month.

One caution for anyone reading Ohio coverage this week. The state is running two figures at once: a press-conference number of nearly 4,000 counted since June 1, and a dashboard number above 4,900 counted since January 1. They are not in conflict and they are not interchangeable. This chart has always used the calendar-year dashboard series for Ohio, so 4,900 is the like-for-like figure, and I count it at 4,900 rather than at more than 4,900, the same way I count every range on this chart at the bottom of its range.

Arkansas reported 237 laboratory-confirmed cases on Friday. CDC says 41 of them are tied to Taco Bell or the recalled Taylor Farms de Mexico lettuce, which is exactly the federal figure for Arkansas in the table below — a rare instance this summer of a state number and a federal number describing the same thing and agreeing.

The federal count has not moved since Wednesday evening. FDA and CDC both still say 6,358 illnesses, 278 hospitalizations and two deaths in fifteen states. Against the 24,539 those same fifteen states report to their own residents, that is 25.9 percent. Against the fifty-state figure it is 23.8 percent. CDC’s national surveillance page still reads 10,468 laboratory-confirmed plus more than 12,255 not confirmed, which totals 22,723 and has not changed since Tuesday. The states are now 4,032 cases ahead of that figure, a gap of nearly eighteen percent, and CDC does not update again until Tuesday.

The hospitalization picture moved with Ohio. The thirteen states that publish a hospital figure now total 680 between them. CDC’s national surveillance says 517 for the entire country. The federal figure for this outbreak is 278, and Michigan by itself reports 279. Four numbers, one disease, one week — and thirty-seven states that publish no hospital figure at all.

Here is the federal count for each of the fifteen outbreak states against what that same department tells its own residents.

And here is where all fifty stand, with the source for every number.

One closing note on what this figure is. 26,755 is a floor, not an estimate. It is the sum of what fifty health departments have published, on schedules that range from daily to never, with four states reporting only a range that I count at the bottom. It is not an estimate of how many people are sick. CDC’s own burden model assumes roughly eighty-three actual illnesses for every laboratory-confirmed case of this parasite, which applied to CDC’s 10,468 confirmed cases puts the real number near 870,000. That multiplier does not go on 26,755, because the state totals already contain the unconfirmed cases the multiplier exists to estimate. What both numbers agree on is the direction, and the direction has not changed once all summer.

This started (Well, it actually started with getting up at 4:00 AM and too much coffee) with a farm radio script ninety-second item on the Land and Livestock Report saying a produce group wants federal investigators to look beyond food as the cause of the Cyclospora outbreak. Behind it was an August 4 press release from the International Fresh Produce Association. Behind that was a federal docket — 91 FR 32972, Docket No. CDC-2026-0925, CDC’s request to keep using the Cyclosporiasis National Hypothesis Generating Questionnaire for another three years that I dug for. I found the comments.

There are three. That is the entire outside response to the federal government’s proposal to keep using, unchanged, the instrument that generates every hypothesis about where Cyclospora outbreaks come from, filed in a summer when the states have counted 25,836 Americans sick.

No state health department commented. Not Michigan, with 12,485 cases. Not Ohio, not Indiana, not Missouri. No consumer organization, no academic epidemiologist, no laboratory, no grower. The window ran sixty-two days, from June 2 through August 3, straight through the beginning of the largest outbreak of this disease in American history, and three people found it. I lately write about this parasite most days and I missed it too.

Now the part that matters. What IFPA filed is not what IFPA announced. The August 4th press release says the association asked CDC to consider a broader range of exposure pathways. Reasonable, technical, unremarkable. The comment letter behind it opens with a recommendation headed: consider the likelihood that Cyclospora has established endemically in the United States, and the likely routes of exposure for an endemic waterborne parasite.

That sentence is not in the press release. It is not in the association’s consumer FAQ. It went where regulators read things and not where reporters do.

Let me state the argument at its full strength, because it deserves that and because the weak version of my objection would lose to it. Cyclospora cayetanensis has no host but humans. It reaches the environment through water carrying human waste — agricultural runoff, aging sewers, combined sewer overflows. The oocyst leaves the body unable to infect anyone and needs warmth and moisture over days or weeks to mature, which is why this has historically been a tropical and subtropical disease. IFPA’s question is whether a warming United States now supplies those conditions often enough to sustain a domestic cycle. It notes that CDC’s own spring-to-summer season is also the season of heavy rain and sewer overflows. And it makes a genuinely sharp point I wish I had made first: because molecular typing cannot independently identify linked clusters — a limit CDC concedes in its own notice — the interview steers the entire investigation, so if water is never asked about, it can never be ruled in or out.

IFPA then drafted nine questions and attached them. Main drinking water source. Water drunk away from home, with zip codes. Awareness of flooding, sewer overflow or a boil-water advisory. Septic system and last service date. Occupational contact with sewage treatment, stormwater, portable toilets, biosolids. Camping on private water. And recreational water contact, with a checklist that runs pool, hot tub, splash pad, lake, river, pond, irrigation canal or ditch.

Strip off the letterhead and the argument is this: twenty-five thousand Americans did not get sick from Taylor Farms lettuce. They got sick from water in Michigan. That compression is mine and not theirs — IFPA never names a body of water, and its letter is careful to frame everything as a question for CDC’s consideration. But the outbreak is half in Michigan, the checklist says lake, and I do not think I am misreading who the questions are for.

Here is the tell, and it is not in what they asked for. It is in what they did not. Nine questions, and not one of them would improve produce traceback. No brand fields. No lot or date codes. No package type, no purchase location, nothing that would move an investigator faster from a sick person to a shipment. Mitchell Berger — a private citizen with nothing to sell, writing on his own time — asked for exactly that: replace the free-text brand and lot fields with structured ones so traceback data is analyzable. The trade association whose members would have to produce those codes did not.

If your stated goal is finding the source quickly and accurately, you want both halves. You ask for the water questions and the lot-code questions. Asking only for the half that points away from your sector is the answer to the question of what this filing is.

Two other things point the same way. This is the same argument IFPA made against Michigan on July 15, when it faulted the state for discounting recreational water — and then removed from its public statement four days later. It has now been refiled somewhere permanent and almost entirely unread. And the association addressed CDC’s Items 1, 3 and 4, but not Item 2, which asks whether the burden estimate is accurate. CDC told the Office of Management and Budget this collection needs 2,800 interviews a year. The states have counted 25,836 people. Nobody on the docket said a word about it.

Now let me be careful, because the sloppy version of this critique loses. The word endemic is doing three different jobs in that letter. That this parasite turns up in American wastewater is probably true and I am not going to pretend otherwise. That there is sustained domestic transmission is unproven. That endemic American water explains this outbreak is very unlikely.

Let me say why, because I have been asked and because a one-line dismissal is not an answer. Six reasons, and the first one is the one that does the most work.

This parasite does not multiply outside a person. It has no animal reservoir, and it does not reproduce in water or soil. Every oocyst in an American lake came out of an American. An endemic cycle therefore has to be primed: you need a standing population of infected people shedding into the water before anyone can catch it from the water. Last year at this point in the season CDC had recorded 249 cases nationally. All of 2025 produced somewhere over 2,800. That is not a reservoir. Endemicity has to start somewhere, and in April of this year there was nothing here to start it with.

Water is local. This outbreak is not. Fifteen states are in the federal count. New Hampshire does not share an aquifer, a lake, a river or a utility with Michigan. To explain fifteen states with domestic water you need fifteen separate local water events arriving inside the same eight weeks, for the first time in the recorded history of this disease. That is not one hypothesis. It is fifteen coincidences. What does reach fifteen states inside eight weeks, routinely and by design, is a distribution network.

New Hampshire is the cleanest example, and the state published it. Of its 69 cases, 32 are tied to recalled lettuce served at Cheshire Medical Center, 32 have no identified source, and 5 are travel-related. The largest identified cluster in the state is a group of people who ate Taylor Farms lettuce in the same building. They did not swim in the same building.

And then there is Minnesota. If recreational water were driving this, the map would follow lakes and hot weather. Minnesota, which puts the number on its license plates, reports 41 cases. Wisconsin reports 160. Florida, where people swim year-round, reports 254. California reports 41. Michigan reports 12,485. Michigan has lakes and so does Minnesota. What Michigan had that Minnesota did not was the product.

The line the cases fall on was drawn by a shipping manifest. Of the 25,836 cases the states report, 25,423 are in the twenty-nine states that received recalled Taylor Farms product. The twenty-one states that did not receive it hold 413 between them. That is ninety-eight percent of counted illness on one side of a distribution list. I have said before that a recall footprint is not a product footprint, and the recipient states skew large, so I would not hang a case on that number alone. But no water hypothesis predicts any correlation with a shipping list at all, and this one is not subtle.

Last, climate is a trend and this is a step change. American temperatures have been climbing for decades. Endemicity does not switch on in a single summer. 2025 produced more than 2,800 reported cases in the entire year. The states have counted 25,836 in the first week of August. Warming raises a baseline slowly; across seasons you can graph. A ninefold jump in one summer is not what a baseline does. It is what a contaminated shipment does.

None of that means water never matters, and none of it is an argument against adding IFPA’s questions. Add them. Add every one of the nine. The reason this is even a live dispute is exactly the reason IFPA gave: nobody asks, so nobody knows. What I am saying is narrower and I want to be held to it — endemic American water is a reasonable thing to investigate and a poor explanation for this outbreak, in these states, on this curve, in this year.

CDC has put in print that for every United States outbreak from the mid-1990s through 2015 in which both the vehicle and its source were identified, the produce had been imported — Guatemala, then Peru, then Mexico. Endemic domestic water does not produce that pattern for thirty years. It does not produce Michigan at 12,485 while Indiana sits at 1,705 and Ohio at 4,013; Michigan’s water is not unusual; Michigan’s distribution chain was. And the biology cuts the other way from how IFPA ran it: an oocyst needs one to two weeks in the environment before it can infect anyone, so a domestic cycle is inherently slow. That is a raised background, not a curve that climbs from a standing start in mid-May and is still climbing in August.

But their hypothesis is testable, and that is the whole point of this post. If these illnesses came from local water in dozens of watersheds, the parasites should be genetically diverse and should vary by region. If they came from one contaminated shipment, they should cluster. Those are different pictures and they do not look alike.

CDC can take that picture. It has been taking it since 2020. The single published study anybody cited on this docket is a CDC paper by Ahart and colleagues on the 2021 season, which built temporal-genetic clusters from sequenced stool isolates and then ran the epidemiology behind them — testing whether a genetics-first model would work for this parasite. Dr. Angel cited it for the proposition that the questionnaire separated one apparent outbreak into two. Having read it, the engine is the sequencer, not the questionnaire. I cited that same paper myself three weeks ago and did not notice it was the answer to the question I keep asking.

Which leaves IFPA and me wanting the same document for opposite reasons. They want the genotypes published because they expect diversity. I want them published because I expect a cluster.

One of us is wrong. The data that decides it was collected months ago, sits in Atlanta, and has not been published. CDC does not have to referee an argument between a plaintiffs’ lawyer and a trade association. It has to release a table.

And if the answer comes back diverse — if the sequencing says these are dozens of separate exposures and not one shipment — then IFPA is right, I am wrong, and I will say so in this space the same day. What I am not willing to do is let a hypothesis this consequential get settled by whoever files the better letter. It should be settled by the sequencer.

The questionnaire at the middle of all of this is a public document. Ten minutes will tell you which questions a sick person gets asked and which they do not. Three people read it this summer and wrote to the government. The season closes on August 31.

Tonight I went back through all fifty state health departments, both federal case pages, and every Cyclosporafile FDA has open. Here is where each of those numbers actually stands, and then the sentence I would like to measure them against.

On July 21, at a briefing about health care fraud, Secretary Kennedy was asked about this outbreak and said the federal government had it under control. He said extensive epidemiological forensics had identified the source and that the companies involved had implemented a recall. Asked separately whether staffing and program cuts had slowed the response, he told CNBC those criticisms were invalid, that there had been no cuts to surveillance, and that the FoodNet reductions were for redundant surveillance.

Here is what the federal government’s own pages said that day, and what they say tonight.

Every one of those movements happened after the sentence, not before it. The federal outbreak count has more than tripled. The hospitalization figure has nearly tripled. Two people have died, both in Michigan, and FDA now says on its own advisory that both had illness onset dates before the July 17 recall — which is the federal government stating that the recall arrived after the deaths it might have prevented.

I want to be fair about the part that is genuinely hard, because the argument is stronger with it than without it. This parasite is difficult. It has an incubation period of up to two weeks, a reporting lag CDC itself puts at six weeks, and no whole genome sequencing route — only targeted genotyping, which the agency’s own scientists describe as slower and less precise. None of that is anybody’s fault. A secretary saying the situation is hard would have been accurate on July 21 and would still be accurate tonight. What he said instead was that it was “under control.”

The claim about surveillance is checkable, and it did not hold up. FoodNet stopped requiring active monitoring for Cyclospora effective July 1, 2025, ten months before the first illness in this outbreak. Susan Mayne, who ran FDA’s Center for Food Safety and Applied Nutrition from 2015 to 2023, told CBS Sunday Morning that the characterization of those cuts as redundant is not accurate. A surveillance system that was switched off is not a redundant one; it is the reason the states are now the only place a national number can be assembled.

Now the piece I have not put in one place before. These are all seven Cyclospora investigations FDA has open, as the agency published them this morning. Not the one everybody is writing about. All seven.

Seven open investigations into this one parasite. Six of them have no food identified at all — six separate groups of Americans who got sick from Cyclospora and cannot be told what they ate. Those six carry 178 people. Add them to the lettuce outbreak and 6,536 people are attached to a numbered federal Cyclosporafile tonight, against the 25,836 the states report. FDA has named a food in one file out of seven.

Two rows on that table deserve to be read slowly. FDA opened file 1398 on August 5 — a brand new Cyclospora outbreak, ten people, no food identified, traceback just beginning. It opened the same day the government announced the lettuce outbreak had reached 6,358. And file 1385 went from ten cases to twenty-five, which FDA explains by saying it now includes additional illnesses based on similarities in reported exposures. That is the agency telling us it has found a common exposure pattern in a file whose food it still will not name.

Back to the states. Two moved since my last chart and both are inside the outbreak. Indiana went 1,349 to 1,705 in a single day, and Michigan went to 12,485 with its hospital count rising from 193 to 279 on its Thursday update. I re-checked Ohio, Iowa, Missouri, Oklahoma, Kansas, Illinois, Kentucky, North Carolina, West Virginia, New Hampshire, Texas, Arkansas, Massachusetts, Alabama and Wisconsin against their own departments and the same-day roundups, and none of them has posted a new figure that has reached the search index. The rest of the fifty have published nothing new since mid-July.

That Michigan hospital number deserves its own line. One state now reports 279 people hospitalized. The federal figure for the entire fifteen-state outbreak is 278. CDC’s national surveillance page says 517 for the whole country. The thirteen states that publish a hospital figure come to 604 between them, and thirty-seven states publish none at all. Four numbers, one week, one disease.

Here is the federal count for each of the fifteen outbreak states against what that same department tells its own residents.

And here is where all fifty stand, with the source for every number.

A note on what this figure is and is not. 25,836 is a floor. It is the sum of what fifty health departments have published, on wildly different schedules, four of them reporting only a range, which I count at the bottom of the range. It is not an estimate of how many people are sick. CDC’s own burden model assumes roughly eighty-three actual illnesses for every laboratory-confirmed case of this parasite, which puts the honest number somewhere near 870,000. I am not going to publish that as a count. I am publishing the floor, and the floor is 25,836.

I do not think the Secretary was lying on July 21. I think he was told the source had been identified and a recall had been issued, and he treated those two facts as the end of the problem. They are the beginning of it. A recall reaches forward only, from the day somebody notices; the people who got sick first had already eaten the evidence. Identifying a food is not identifying a cause — nobody has published a positive product sample, a water sample, an environmental assessment, or a genotype from the fields in central Mexico. And a case count that is still climbing sixteen days later is not what control looks like. It is what a lagging indicator looks like on a system that stopped watching.

The season CDC defines closes on August 31. That is twenty-five days. I would settle for someone in Washington saying, out loud, that this is not under control yet.

The documented chain, and the hypothesis this post is about.

Every time I raise irrigation water in this outbreak, the same sentence comes back at me. The water comes from deep wells. There is no Cyclospora in a well six hundred feet down. It is a good answer. It is also, on the government’s own record, an answer to a question nobody asked.

Open FDA’s environmental assessment of the Doctor Mora plant, published in November 2013 after cyclosporiasis sickened hundreds of people in Iowa and Nebraska. The report says all five assessed ranches use well water, and that the wells run roughly 650 to 1,000 feet deep with casing to the water table and cement surface seals. That is the sentence the industry quotes. Two sentences later comes the one it does not.

In 2013 three of the assessed ranches used elevated irrigation water holding ponds. Both lined and un-lined ponds were observed.

And the reason FDA gives is not a food safety reason at all. It is an electricity bill. Some of the wells lacked adequate flow rates for direct irrigation, and the power to run the well pumps was much cheaper at night than during the day. The growers therefore pumped after dark into open ponds and drew from the ponds when they needed to irrigate. At Ranch B the water from two wells was combined in a single elevated reservoir before use, and that reservoir was dosed with copper sulfate to keep the algae down. Nobody treats a sealed pipe for algae. Algae needs sunlight and open air, and if you are managing algae, you are managing an open body of surface water.

That is the whole of the deep-well defense, and it collapses on a definition rather than on a fact. Nobody has to argue that this parasite is swimming around in the Guanajuato aquifer. Water that leaves a six-hundred-foot well and then sits in an unlined pond under the sun is not well water anymore in any sense that matters to a parasitologist. The industry’s own rulebook says so. California’s leafy greens standards make growers categorize every source separately — well, reservoir, canal, municipal — and treat water from open sources such as a canal or reservoir when it is applied overhead in the twenty-one days before harvest. The distinction between a well and a pond is not something I invented for a blog post. It is a line the industry drew itself and then stopped mentioning.

Taylor Farms knew it too, and its own numbers are in the same federal report. In response to the 2013 outbreak the company ran its own assessment across the plant and thirty ranches and collected 793 samples. The categories are itemized: 81 well samples, and 48 from reservoir and storage. Six more reservoir and storage samples under environment. Three from a reservoir listed as no longer active. Ten from run-off and standing water. Eight from drain ditches. Seven from rivers, streams and canals. A company that sends technicians out to sample drain ditches and standing water is not a company that believes its irrigation water never touches the surface of the earth.

FDA did look at the ponds in 2013, and it is worth being precise about what the looking consisted of. The team walked the entire perimeter of each holding pond and found no indication of human or animal intrusion and noted that the ponds were fenced inside a fenced ranch under twenty-four-hour security. That is an intrusion check. It answers whether somebody climbed in. It does not address rain running off surrounding ground into an open pond, it does not address birds, and it does not address what an unlined pond exchanges with the soil water around it. The report’s one sentence about flooding is specific: the team found no indication of flooding or other potential irrigation water contamination at any of the wells. The wells. The ponds are not in that sentence.

So, what does 2026 look like?  So far, silence.

Now the part where I am speculating.

The first half of 2026 was the wettest in Guanajuato in nearly a decade. The Servicio Meteorológico Nacional logged 10.9 inches of rain accumulated by mid-year, with June the heaviest month at up to 7.6 inches, a figure the state had not seen since around 2018. The rainy season opened in the middle of May. The state’s own water engineers warned before it started that June would be the wettest month, and warned specifically about standing water, urban flooding and arroyos overflowing their banks. CONAGUA had forecast in April that El Niño would push 2026 past 2025, which was already one of the wettest years of the decade.

Now put that beside the illness curve. Back the six-week reporting lag off the late-June surge in reported cases and the excess illnesses in this outbreak begin in the back half of May. The rains began the week of May fourteenth. June, the month the state water engineers said would be the wettest, is the month the reporting exploded.

That is a temporal correlation and nothing more. Statewide rainfall is not Doctor Mora rainfall, and the station data that would settle it sits with CONAGUA. I have not found a single published report of a sewage system overflowing or a treatment plant bypassing in Doctor Mora or San José Iturbide this spring. I am not going to claim one happened. What I will say is that perhaps nobody has looked, and that the absence of a news story about a rural Mexican sewer is not evidence about a rural Mexican sewer.

Here is why the mechanism fits, though, and it fits better than anything else. This parasite has one known reservoir, and it is human feces. FDA’s own report says so in its second paragraph. Heavy rain is how human waste gets out of septic tanks and sewers and into arroyos and onto open ground — that is not a Mexican problem, it is a plumbing problem, and it happens in Iowa. An unlined elevated holding pond is a catchment. And then comes the detail that makes a pond different from every other point on this map: the oocysts shed in faces are not infectious when they leave the body. They need one to two weeks somewhere warm, between 72 and 90 degrees Fahrenheit, to sporulate. Northeastern Guanajuato sits inside that temperature window in every month of the year.

A holding pond is warm, still, open and full for weeks at a time. It is not merely a place where contamination could enter the irrigation system. It is a place with exactly the residence time and exactly the temperature this organism requires to become infectious. If you were designing an incubator for Cyclospora and then attaching it to a lettuce field, you would build something that looked a great deal like an unlined irrigation reservoir in central Mexico in June.

The plant is at kilometer 16 of the Doctor Mora to San Miguel de Allende highway. Seven hundred and seventy-six feet away sits Balneario Las Fajanas, a public thermal bathing complex the state of Guanajuato lists in its own tourism guide, with pools, water slides, grills, palapas and restrooms, open every day of the week. I have written about that before. Within nine miles there is a second bathing complex, a water park, two municipal treatment works and two towns. What I cannot show you is the thing this post is actually about, because the five ranches in FDA’s report are lettered A through E and their locations have never been published. The ponds are somewhere in that landscape and nobody outside the company and two governments knows where.

None of this is a novel theory. The Center for Produce Safety and USDA have been funding work on exactly this premise — faulty septic systems, compromised sewer pipes and treated effluent reaching surface water used to irrigate food crops. A two-year study in southeast Georgia confirmed the parasite by sequencing in nine percent of municipal wastewater sludge samples. In 2020 FDA came closer to proving it than it ever has, sampling a regional canal in Florida during the Fresh Express bagged salad outbreak and finding the parasite at two points, north and south of where the suspect farm drew canal water for irrigation. The genetic typing was not mature enough to match those finds to patients, and the investigation closed unresolved. And in 2015 FDA published an import alert on Puebla cilantro describing human feces and toilet paper in growing fields, and wash water vulnerable to sewage, at eight of eleven farms inspected.

There are four documents that would settle most of this and none of them requires anyone to speculate. The CONAGUA station records for Doctor Mora and San José Iturbide, which would say what rain actually fell where. The municipal discharge and bypass records held by the state water commission and the San José Iturbide utility, which would say whether anything overflowed. The locations, linings and sampling histories of the irrigation holding ponds serving the 2026 crop — Taylor Farms has these, and FDA has seen their 2013 ancestors. And the answer to the question FDA asked Taylor Farms thirteen years ago, in recommendation two of that same report: is Cyclospora a reasonably likely hazard in the Guanajuato growing region, and if it is, what happened to the wash step?

Twenty-two thousand people have been counted this season by their own state health departments. Two of them are dead. I am not in a position to tell you that a pond did this, and I have tried to be careful above about the difference between what a federal report from 2013 says and what I think it implies. But when the industry answers a question about irrigation water with the word well, somebody should point out that FDA already wrote down what happens to the water after it leaves the well. It goes into a pond. Sometimes the pond has no liner. And it sits there in the sun, in a valley where the temperature never leaves the range this parasite needs, until somebody opens a valve.

Twelve days after its last update, FDA posted a new advisory this evening that said nine-state outbreak now says fifteen. The total is 6,358 illnesses, up from 1,947. Hospitalizations are 278, up from 98. And after eleven days in which every federal page said zero, the deaths are on it: two, both in Michigan.

The number that matters most, though, is not any of those. It is the sentence explaining them.

FDA says that because of additional interview data, CDC expanded its confirmed cases to include people who reported exposure to Taco Bell or recalled Taylor Farms de Mexico iceberg lettuce. Until tonight the federal definition was Taco Bell exposure and nothing else. I wrote this morning that if New Hampshire joined this outbreak on the thirty-two illnesses it has tied to a hospital cafeteria in Keene, the case definition would have widened past a single restaurant chain for the first time since July 14, and that CDC could settle it in one sentence. That is the sentence. This is no longer a Taco Bell outbreak. It is a lettuce outbreak that ran through Taco Bell too.

You can see it in FDA’s own lists. Its restaurant line says the Taco Bell locations that received recalled lettuce include at least thirteen states — Arkansas, Iowa, Illinois, Indiana, Kansas, Kentucky, Michigan, Missouri, Nebraska, Ohio, Oklahoma, Pennsylvania and West Virginia. The outbreak has fifteen. The two that are in the outbreak and not on that list are New Hampshire and North Carolina, which are precisely the two states whose illnesses never ran through that chain. New Hampshire’s cases came out of a hospital cafeteria.

FDA also says, twice and without being asked, that many of these illnesses began before the July 17 recall. On the deaths it goes further: based on information Michigan provided, both people who died had illness onset dates prior to the recall. That is the federal government stating on its own advisory that the recall arrived after the deaths it might have prevented.

Here is the federal count for each of the fifteen states, alongside what that same state health department is telling its own residents this evening.

The first is North Carolina: two. The federal outbreak count for North Carolina is two people. North Carolina tells its own residents 867, and its dashboard still says the increase in the state does not appear to be associated with the Taco Bell outbreak. Both of those statements are now on government websites at the same time.

The second runs the other way, and it is the more useful one. In Nebraska and Pennsylvania, the federal number is now larger than the state number — 92 against 91, and 132 against 82. Neither state was undercounted by Atlanta. Both were undercounted by themselves. Nebraska’s figure is a spokesman’s number from July 20 that nobody has refreshed, and Pennsylvania publishes no dashboard at all and takes cyclosporiasis reports voluntarily. When the federal government can find more sick Pennsylvanians than Pennsylvania can, the 

The fifty-state floor stands at 25,213 tonight, across every state health department that publishes anything. The hospitalization picture is stranger than it was this morning. FDA now says 278. CDC’s national surveillance page says 517. The thirteen states that publish a number of their own come to 518 between them. Three federal-adjacent figures, three different answers, same week — and thirty-seven states publish no hospitalization figure at all.

One more thing changed quietly. FDA’s distribution list grew from twenty-seven foodservice states to twenty-eight, and the state it added is New York. New York has been reporting 753 cases while sitting outside every distribution list published so far. It is inside one now. The recall footprint is twenty-nine states counting the Walmart retail list, and the recall notice itself has not otherwise moved.

And one thing did not change at all. The outbreak still officially begins on June 22. FDA’s own July 16 advisory reported 1,644 illnesses with onset dates running back to May 13. The June 22 floor has now survived three updates and a full expansion of the case definition, and those six weeks of illness are still outside the outbreak. Nobody has explained where they went.

Where each of the 50 states stands this evening, with the source for every number. Ten rows changed: Alabama, Arkansas, Illinois, Kentucky, Massachusetts, Michigan, New Hampshire, North Carolina, Texas and West Virginia.

One more number, because 25,213 is a floor and not an estimate. In 2011 Elaine Scallan and seven colleagues at CDC published the agency’s foodborne burden model in Emerging Infectious Diseases, and its Table 2 still carries the only published United States burden estimate specific to Cyclospora. The underdiagnosis multiplier for this parasite is 83.1. For every laboratory-confirmed case, CDC’s own model assumes about eighty-three people actually got sick. Applied where it belongs, to the 10,468 laboratory-confirmed cases CDC reports this season, it puts the real number near 869,891. It does not go on 25,213. The state totals already include the cases the multiplier exists to estimate, so multiplying those counts the same people twice.

That is arithmetic rather than a measurement, and it should be read as a ceiling. The model’s own credible interval for Cyclospora runs from 137 to 37,673 domestically acquired foodborne illnesses a year, a 275-fold spread and the least precise line in the table. The 83.1 was built from data gathered before multiplex PCR panels carried a Cyclospora target at all, which happened in 2014, and an outbreak that has doctors ordering the test drives the true multiplier down further. Nobody has published an updated one; the 2025 revision of that paper covers seven pathogens and this parasite is not among them. What the figure is good for is the shape of the problem. The same table puts Listeria at 2.1, because severe illness gets diagnosed and a miserable, self-limiting one does not.