CDC currently reports 4,173 laboratory-confirmed, domestically acquired Cyclospora cases since May 1.[1] The federal government’s own published model for foodborne illness burden — Scallan and colleagues, 2011, still the only Cyclospora-specific burden estimate ever published in this country that I could find — carries a credible interval whose upper bound implies that roughly 272 people are infected for every 1 a laboratory confirms.[2] Multiply it out and this outbreak has infected at least 1,135,056 Americans.
I do not believe that number, and I would not publish it as a finding. The top of an interval is not an estimate; it is the edge of what the model cannot rule out, and in this case the model is fifteen years old and the interval is enormous — the same table’s lower bound implies a multiplier of 1.0, which would mean every single Cyclospora infection in America gets laboratory-confirmed, which is plainly false. When the honest range on a parameter runs from 1 to 272, the honest thing to say is that we are guessing.
Scallan’s central estimate for the underdiagnosis multiplier is 83.1. Applied to 4,173 laboratory-confirmed cases, that is 346,776 people. That figure is not the edge of anything. It is the middle of the published distribution, produced by CDC’s own epidemiologists and used by CDC ever since to describe how much foodborne illness this country actually has.[3]
Walk it down, because even 346,776 is built on generous assumptions. The 83.1 was derived from data collected in the 2000s, when finding this parasite required a clinician to specifically request an unusual stain because the routine ova-and-parasite exam misses it entirely.[4] The first FDA-cleared molecular gastrointestinal panel with a Cyclospora target did not exist until 2014, and by 2018 and 2019 a survey of roughly seven hundred American clinical laboratories found 62 percent of panels included one.[5] Detection is far better now than when that multiplier was built. On top of that, an outbreak this size creates its own surge — physicians in Michigan and Ohio are testing for this parasite in July of 2026 in a way nobody does in an ordinary March. Both effects push the real multiplier down.
There is one more number that belongs on the table because it does not come from the government. bioMérieux, which makes the most widely used molecular gastrointestinal panel in the country, publishes surveillance drawn from its own installed base of instruments, and between June 26 and July 5 of this year it recorded Cyclospora detections running 10.8 percent above the same stretch of 2025 nationally, and 27.5 percent above it across the Midwest.[6] Those are not the numbers you would expect underneath a national count running many times last year’s at the same point in the season.[7] I think the window explains most of the gap — that stretch closes before this outbreak turned steep, and Michigan’s newly reported cases over the four weeks that followed ran 71, then 498, then 1,331, then 2,408 — and the platform sees only the laboratories that bought one manufacturer’s machine.
Hospitalizations are the least-undercounted thing in any outbreak. People sick enough to be admitted get tested. Scallan’s companion table put roughly 20 Cyclospora hospitalizations a year against 19,808 total annual illnesses — a hospitalization rate of about one tenth of one percent of all infections, not of confirmed cases.[8] There are 308 hospitalizations in the national count. Divide 308 by 0.001 and the implied number of infections is roughly 305,000.
What the hospitalization figure actually tests is whether this outbreak looks like the outbreaks the model was built from, and it does: Scallan’s table implies roughly 8.4 hospitalizations for every hundred laboratory-confirmed cases, and this outbreak is running 7.4. A model assembled from 2000s data is describing a 2026 illness with about the severity it predicted, which is exactly what you would want to check before applying a fifteen-year-old model at all.
Now cut it to the bone. Suppose the multiplier has fallen by a full order of magnitude since 2011 — from 83.1 down to 8.3. That is a more aggressive haircut than the arrival of PCR panels plausibly justifies, and it concedes essentially the entire argument to anyone who wants to say the old number is stale. It still puts this outbreak at about 35,000 people. That is more than twice what every state health department in America has managed to count between them, and it is roughly ten times the number of cyclosporiasis cases this country reports in a normal full year.[9]
Whatever you believe about the multiplier, every rung of that ladder sits above what is being counted. The disagreement is only about how far above.
Michigan offers a check on all of it. The state has reported 8,176 cases and 160 hospitalizations.[10] That is a hospitalization rate of 2.0 percent against the 6.5 percent Scallan found — though the comparison is not quite like for like, because Michigan’s total includes probable cases while Scallan’s rate was calculated among laboratory-confirmed ones, which pushes Michigan’s true rate somewhat higher than 2.0. Even allowing for that, the direction is unmistakable. A falling hospitalization rate means a state is catching the mild end of the illness spectrum, which means Michigan’s own undercount multiplier is far smaller than 83. Michigan is looking hard. Run the hospitalization arithmetic on that one state anyway and you get roughly 158,000 infections, which works out to a multiplier of about 19. That is the number I would defend. And then consider Pennsylvania, which has reported 28 cases in a nine-state outbreak and where reporting cyclosporiasis to the health department is voluntary. Nobody can calculate Pennsylvania’s multiplier, because Pennsylvania is not measuring anything.[11] That is the actual shape of the undercount: small where somebody is looking, unknowable where nobody is.
Against every one of those figures, here is what is actually counted. CDC’s outbreak investigation — laboratory confirmation plus a documented exposure — stands at 1,947 people in nine states, 98 hospitalized.[12] CDC’s national surveillance count of laboratory-confirmed domestic cases is 4,173, with more than 7,400 additional reported illnesses awaiting the analysis that would confirm them, upward of 11,500 cases in all across 41 states, and 308 hospitalizations. Add up what the individual state health departments have posted and the floor is near 15,500.[13] Every one of those is a floor, and every one of them describes a week ago or more, because CDC has said it can take as long as six weeks to determine whether a sick person belongs to an outbreak.
It is worth pausing on how large those floors already are. CDC’s outbreak reporting system holds 181 Cyclospora outbreak records covering 1990 through 2023 — thirty-three years — and together they account for 6,251 illnesses.[14] The 11,500 reported this summer is nearly twice the entire outbreak-associated total that system logged in three decades. The outbreak that taught this country the parasite existed, the 1996 Guatemalan raspberry outbreak, produced 1,465 reported cases and 978 laboratory-confirmed ones across twenty states, the District of Columbia and two Canadian provinces, and it is the reason cyclosporiasis became a nationally notifiable disease in 1999.[15] The confirmed cluster alone in this outbreak is larger than that.
The last question is why a plaintiffs’ lawyer (and his Epi and Nurse) is doing this arithmetic at all, and the answer is that nobody with a federal badge is. There has been no updated Cyclospora burden estimate since 2011. The 2025 revision of the federal foodborne illness estimates covers seven pathogens and this one is not among them.[16] And on July 1 of 2025, CDC narrowed FoodNet — the active surveillance system that generated the 239-cases-a-year denominator underneath every figure above — from eight pathogens to two, keeping Salmonella and Shiga toxin-producing E. coli and dropping Cyclospora, Campylobacter, Listeria, Shigella, Vibrio and Yersinia, citing inadequate funding.[17] Ten months later the first people in this outbreak got sick.
The realistic range in this post runs from 35,000 to 347,000. It is embarrassingly wide, and that width is not my doing. It is the width you get when a country decides that counting sick people is a discretionary expense. There may be a final official number for this outbreak eventually. It will be tidy, it will be footnoted, and it will be smaller than the truth.
[1]CDC, Cyclosporiasis Surveillance, cdc.gov/cyclosporiasis/php/surveillance/index.html (4,173 laboratory-confirmed domestically acquired cases; 7,400+ awaiting analysis; 11,500+ total; 41 states; 308 hospitalizations; data as of July 21, 2026).
[2]Scallan E, Hoekstra RM, Angulo FJ, Tauxe RV, Widdowson MA, Roy SL, Jones JL, Griffin PM. Foodborne Illness Acquired in the United States — Major Pathogens. Emerg Infect Dis. 2011;17(1):7–15, Table 2. doi:10.3201/eid1701.p11101. The Cyclospora cayetanensis row reports 239 laboratory-confirmed cases annually and 19,808 estimated total illnesses, 90% credible interval 239–65,135 — implying a multiplier ranging from 1.0 to approximately 272.5. Expanded tables at cdc.gov/foodborneburden/pdfs/major-pathogens-expanded-tables.pdf.
[3]Id. (underreporting multiplier 1.0; underdiagnosis multiplier 83.1; 11,407 domestically acquired foodborne illnesses, 90% CrI 137–37,673). The domestically acquired foodborne interval is a 275-fold spread and should not be presented as precise.
[4]Mathison BA, Pritt BS. Cyclosporiasis — Updates on Clinical Presentation, Pathology, Clinical Diagnosis, and Treatment. Microorganisms. 2021;9(9):1863. doi:10.3390/microorganisms9091863. See also American Society for Microbiology guidance that a normal ova-and-parasite examination does not rule out Cyclospora, and CDC guidance recommending three or more stool specimens collected more than 24 hours apart owing to intermittent shedding.
[5]Evaluating Foodborne Cyclosporiasis Using Foodborne Diseases Active Surveillance Network and Foodborne Disease Outbreak Surveillance System Data, 2015–2019. Am J Trop Med Hyg. 2025;112(2):319 (1,376 FoodNet cases; 21.1% travel-associated; 90.9% summer onset; 71.7% outbreak-associated). The 62 percent figure appears in that paper, which cites an underlying survey of approximately 700 U.S. diagnostic laboratories; it is not the authors’ own finding.
[6]bioMérieux, Tracking the Cyclospora cayetanensis Outbreak in the United States (July 2026), reporting BIOFIRE FIREWORKS surveillance data drawn from its installed base. The same post states that several FDA-cleared multiplex panels available for clinical diagnostic testing do not include Cyclospora as a target.
[7]CDC Health Alert Network Advisory No. 00531, Domestically Acquired Cyclosporiasis Cases in Multiple U.S. States, 2026 (July 14, 2026) (249 cases reported nationally May 1–July 16, 2025). The 2026 and 2025 reporting windows are not precisely aligned.
[8]Scallan et al., supra note 2, Table 3 (approximately 20 Cyclospora hospitalizations annually, 90% CrI 0–190; 6.5% hospitalization rate; zero deaths). CDC’s expanded tables note that hospitalization and death rates were doubled to adjust for underdiagnosis before being applied to the number of laboratory-confirmed cases — so the hospitalization estimate does not run through the 83.1 underdiagnosis multiplier, while the illness estimate does. Twenty hospitalizations against 19,808 estimated illnesses yields approximately 0.101% of all infections. The upper bound of 190 hospitalizations would imply 0.96%, and a correspondingly smaller total near 32,000.
[9]CDC cumulative reported cases through report week 53: 2,707 (2022); 4,200 (2023); 3,719 (2024); 3,955 (2025) — an average of roughly 3,645 per year, inclusive of travel-associated cases.
[10]Michigan Department of Health and Human Services, reported case total as of July 24, 2026; hospitalizations as of July 23, 2026. The state total includes probable as well as laboratory-confirmed cases.
[11]Casillas SM, Hall RL, Herwaldt BL. Cyclosporiasis Surveillance — United States, 2011–2015. MMWR Surveill Summ. 2019;68(SS-3):1–16. doi:10.15585/mmwr.ss6803a1 (cyclosporiasis not reportable in Alabama, Idaho, Maine, Mississippi, Nevada, North Dakota or Pennsylvania during the surveillance period; reporting to CDC voluntary even where the condition is reportable by law).
[12]CDC, Cyclosporiasis Outbreak Investigation, cdc.gov/cyclosporiasis/outbreaks/07-26/investigation.html, and Where People Got Sick, cdc.gov/cyclosporiasis/outbreaks/07-26/locations.html (nine states; at least 1,947 cases; at least 98 hospitalizations; no deaths).
[13]State-by-state totals compiled from individual state health department postings; see also CNN’s running case tracker, cnn.com/2026/07/21/health/cyclospora-cases-count (refreshed July 24, 2026). Michigan and Ohio figures per MDHHS and CNN, July 24, 2026; NBC News reported 2,596 Ohio cases the same day and the Ohio Department of Health has not posted the figure itself.
[14]CDC National Outbreak Reporting System public data file, 181 Cyclospora outbreak records covering 1990 through 2023, totaling 6,251 illnesses and 198 hospitalizations.
[15]Herwaldt BL, Ackers ML; Cyclospora Working Group. An outbreak in 1996 of cyclosporiasis associated with imported raspberries. N Engl J Med. 1997;336(22):1548–56 (1,465 cases reported by 20 states, the District of Columbia and 2 provinces; 978 laboratory-confirmed). Cyclosporiasis became nationally notifiable in 1999.
[16]Scallan Walter EJ, et al. Foodborne Illness Acquired in the United States — Major Pathogens, 2019. Emerg Infect Dis. 2025;31(4). Cyclosporais not among the seven pathogens re-estimated.
[17]CDC reduced FoodNet’s pathogen panel to Salmonella and Shiga toxin-producing E. coli effective July 1, 2025. The CDC quietly scaled back a surveillance program for foodborne illnesses, NBC News (Aug. 26, 2025), nbcnews.com/health/health-news/cdc-quietly-scaled-back-surveillance-program-foodborne-illnesses-rcna227089; CDC confirms cuts to program that tracks foodborne pathogens, Food Safety News (Aug. 2025) (quoting CDC spokesman Gabrial Alvarado). CDC told the Connecticut Department of Public Health that funding had not kept pace with the resources required to maintain FoodNet surveillance for all eight pathogens.