The number as everyone hears it
In January 2026, a new federal dietary-guidelines campaign called RealFood.gov launched with a black-and-white Super Bowl ad featuring Mike Tyson. Visitors who scrolled the campaign site found this line: "90% of U.S. healthcare spending goes to treating chronic disease — much of which is linked to diet and lifestyle." The same figure appeared in the dietary guidelines themselves and on the CDC's website. Health and Human Services Secretary Robert F. Kennedy Jr. stated it plainly in a January 7 announcement: "The CDC reports that 90% of healthcare spending treats chronic disease."
It is a genuinely alarming number, and it has real institutional weight behind it — a Cabinet secretary, a federal dietary campaign, a decade of CDC materials. It is also, on inspection, not what it is being used to say.
Where the number actually comes from
The 90% figure traces to a 2017 RAND Corporation report, using 2014 survey data from the Medical Expenditure Panel Survey. RAND defined a "chronic condition" as any physical or mental health condition lasting more than a year that either limits a person's ability or requires ongoing treatment — a definition broad enough to include hypertension, high cholesterol, anxiety, arthritis, and asthma alongside heart disease and cancer. Using that definition, RAND estimated that about 60% of Americans had at least one chronic condition. Spending on that 60% of the population — all of their spending, not spending on the condition itself — came to roughly 90% of the total.
Notice the word that matters. The statistic measures spending on people with a chronic condition, not spending treating chronic conditions. A person with well-managed high cholesterol who breaks an arm skiing has the entire cost of that fracture counted in the 90% — because they are a person with a chronic condition, not because the chronic condition had anything to do with the bill. This is population attribution, not treatment accounting, and the two produce very different numbers. When PolitiFact worked through what treatment of chronic disease alone would actually cost against the current $5.3 trillion total, the honest figure came out closer to 42% — under half, not nine-tenths.
Christine Buttorff, the RAND researcher whose work is the actual source of the 90% figure, put it plainly when asked about its use in the RealFood.gov campaign: the underlying issue is real, she said, "even if their use of this statistic isn't quite right."
Why it doesn't scan
Set the definition next to the population, and the number stops looking like a discovery and starts looking like an artifact of how the bucket was built.
By RAND's definition, 60% of all Americans qualify as having a chronic condition — and that is not a figure concentrated in older or sicker Americans. CDC's own current breakdown shows 60% of adults ages 18 to 34 already meet the threshold, rising to over 75% in midlife and over 90% among adults 65 and older. A classification that sweeps in six in ten people in their twenties and thirties is not identifying a distinguishable "sick" population. It is close to identifying everyone who will eventually see a doctor for anything lasting longer than a year — which is to say, nearly the whole country, categorized by age alone.
Run the arithmetic implied by the headline figure. If 60% of people account for 90% of spending, the remaining 40% account for the other 10% — a per-capita ratio of roughly six to one between the two groups. That is not an unreasonable ratio for health spending, which is always skewed toward the people who use more care. But it means the 90% statistic is close to a mathematical near-certainty once you have defined "chronic condition" broadly enough to capture 60% of the population. It reveals far less about disease burden than it appears to, and far more about where the classification line was drawn.
Health economists who study spending concentration directly — rather than through a chronic-condition proxy — consistently find something more precise and more useful: the top 5% of patients by spending account for roughly half of all health care spending. The top 1% account for about 21%. The bottom half of the population, by contrast, accounts for only about 3%. This is not a new or contested finding; it shows up in the government's own Medical Expenditure Panel Survey data year after year.
That is the number that actually locates where the money goes — a small population of complex, often multi-condition patients, not "everyone who has ever had a cholesterol test." It is also a far more actionable number, because it points to precisely who benefits most from continuous, coordinated care: the patients drifting toward that top tier, before they arrive there.
There is another problem hiding in the way the number is now presented. The 90% figure comes from RAND's analysis of 2014 Medical Expenditure Panel Survey data. The $5.3 trillion figure comes from the much broader National Health Expenditure accounts. RAND did not find that 90% of $5.3 trillion was spent on people with chronic conditions. The percentage and the dollar total come from different spending universes. Attaching one to the other creates a dollar claim the underlying study did not measure.
Peeling the total apart
That denominator problem becomes clearer when the $5.3 trillion total is peeled apart. The National Health Expenditure figure is not synonymous with patient care, as if every dollar were spent addressing someone's medical need. A meaningful share of the total is profit, capital investment, administration, marketing, and the cost of care nobody paid for, all of it bundled into the same figure and then, by implication, pinned on the patients who happen to be sick. This site's own breakdown of where the money goes already draws this line once, in outline. Here it is drawn all the way through.
Start with what CMS's own accounting already separates out, because these figures are official and they subtract cleanly from the $5.3 trillion total.
Those three categories — roughly $835 billion combined, close to CMS's own long-standing figure of about 15% of the total — are not treatment. Nobody's chronic condition consumed the cost of a new hospital wing. Subtract them from $5.3 trillion, and what remains is what CMS itself calls personal health care: roughly $4.5 trillion, about 85% of the total. That establishes the first boundary. Roughly 15% of the $5.3 trillion national expenditure total is not personal health care at all, before examining what is still embedded inside the remaining 85%.
What's still hiding inside the 85%
There is no national account labeled “money actually spent treating an individual patient or condition.” So the next step has to be an estimate. The purpose is not to manufacture a more precise number than the data allow, but to ask how much of the $5.3 trillion could reasonably remain once expenditures clearly surrounding care rather than treating an individual or condition are removed.
A hospital's per-visit charge does not arrive as a receipt itemizing "cost of your care" plus "cost of the lobby renovation" plus "the CEO's compensation" plus "this quarter's operating margin." It arrives as one number. The same is true of a prescription's price, which folds in the cost of the pill alongside the cost of the advertisement that made a patient ask for it by name. None of this is hidden by conspiracy. It is hidden by the structure of the accounting itself, which was built to track total spending, not to answer the question this primer is asking.
Three pieces of it are visible enough to put a number on, even if the number can only ever be illustrative rather than exhaustive:
Profit. The seven largest publicly traded health insurance companies reported a combined $71.3 billion in net profit in 2024 — a new record, and a figure large enough on its own to fund a serious expansion of primary care access. That number sits mostly inside the roughly $400 billion "administration and net cost of insurance" category above; it is the clearest single piece of that category anyone can point to and name.
Marketing. The pharmaceutical and healthcare industry spent an estimated $30 billion or more on advertising in 2024, close to 90% of it from pharmaceutical companies — over $10 billion specifically marketing prescription drugs. None of this is a separate NHE line item. It is folded into the price of every branded drug that gets bought because a patient saw the ad, which means it lives inside the $467 billion prescription-drug figure, indistinguishable from the cost of the medicine itself.
Unpaid care, repriced onto everyone else. The American Hospital Association reports that hospitals have absorbed roughly $745 billion in uncompensated care since 2000 — bad debt and charity care that was never reimbursed. That cost does not vanish. Hospitals price it into what they charge paying patients and insurers, a practice widely known as cost-shifting. A portion of every paying patient's bill is, quietly, someone else's unpaid one.
These three are the ones with a findable number attached. They are not the whole list. Facility and real-estate costs, executive compensation, private-equity return requirements layered onto physician practices and hospital chains, out-of-network markups, the administrative staff a clinic employs solely to fight insurance denials — all of it is real, all of it is priced into care somewhere, and none of it is separately reported anywhere in the national accounts. That is not because it is small. It is because the accounting was never built to make it visible, which is exactly the blind spot the measurement problem describes from a different angle: an instrument that was never pointed at the question cannot answer it, no matter how much data it produces.
So how much actually goes to treatment
The official accounts cannot answer this question precisely, so the next step is to construct a reasonable upper estimate from the categories most readily identifiable with treatment of individual patients and conditions.
Hospital care ($1.63 trillion), physician and clinical services ($1.11 trillion), and retail prescription drugs ($467 billion) together total roughly $3.2 trillion, about 60% of the $5.3 trillion national expenditure total. For the purpose of this exercise, that 60% provides a generous working ceiling for spending most readily identifiable as treatment directed toward individual patients and conditions. It is an estimate, not an official CMS category.
The estimate is intentionally generous. Those categories still contain profit, administration, facilities, marketing, cost shifting, and other expenditures surrounding treatment rather than constituting the treatment itself. Public data do not allow those costs to be cleanly removed. The point is therefore not that exactly 60% of national health expenditures are patient care. The point is that treating the entire $5.3 trillion as though it were money spent treating patients is not supportable, and even a generous attempt to reconstruct that denominator produces a dramatically smaller number.
Now put the numbers side by side. The public claim invites the reader to hear that roughly 90% of $5.3 trillion, nearly $4.8 trillion, is associated with chronic disease. RAND did not measure that. If its 90% finding were approximately transferable to a patient-directed spending pool closer to the generous 60% estimate above, the comparable amount would be roughly 90% of $3.2 trillion, about $2.9 trillion. That is not offered as a new measured national statistic. It is an illustration of the denominator problem: once expenditures that clearly do not represent treatment of an individual patient or condition are removed, the dollar implication changes enormously.
The difference between $4.8 trillion and roughly $2.9 trillion is not statistical noise. It is what happens when a percentage derived from one spending universe is attached to a much larger one.
The cost of getting it wrong by accident
Start with the honest version of the error, because it is the more common one and it deserves to be treated as such. "For people with" is an easy phrase to compress into "spent treating" — it is a shorter sentence, it fits a soundbite, and the compression feels harmless because the underlying concern is real. Chronic disease genuinely is expensive, genuinely is linked to diet and lifestyle in many cases, and genuinely deserves attention. Nobody needs a motive to make this error. It is the kind of simplification that happens whenever a nuanced population statistic gets handed from a research report to a communications team to a Cabinet announcement to a thirty-second ad.
But the compression is not free, and this is the point worth sitting with. "Ninety percent is spent treating chronic disease" points a reader toward a specific kind of solution: better treatments, better drugs, better disease-specific interventions — the product-shaped fixes this framework calls horizon-chasing. "Ninety percent of spending happens to fall on people who have a chronic condition, defined broadly enough to include most of the adult population" points toward a completely different and far less quotable conclusion: that the way to change the number is to change how care is delivered to nearly everyone, continuously, before conditions compound. The inaccurate version is not just wrong. It is wrong in a direction — toward the products this framework has always argued distract from the harder, unglamorous work of delivery.
It also does something subtler. A statistic that reads as "sick people are 90% of the cost" quietly reframes people with common, manageable conditions as the source of a financial problem — the exact cost-center framing this site has argued against from its first page. Nobody intends that reading. It happens anyway, every time the number is repeated without its context.
The cost of getting it wrong on purpose
Now the harder version, and it needs to be said carefully. This is not a claim about conspiracy. Nobody needs to coordinate for a shaky number to get used self-interestedly — they only need a reason to reach for it, and a number this large, this alarming, and this loosely defined gives everyone with a policy to sell a reason.
In 2024 congressional testimony supporting the Chronic Disease Flexible Coverage Act, a Ways and Means Committee chairman cited the same statistic — then paired at $4.1 trillion — to argue for legislation expanding how employer health plans could structure chronic-disease coverage. Around the same time, an industry coalition wrote to the House Budget Committee in support of a different bill, the Preventive Health Savings Act, citing a related but distinct cut of the same underlying data: that patients with two or more chronic conditions account for 84% of spending, and the healthiest half of the population accounts for under 3%. Different bill, different framing, same well of ambiguous population data — each advocate reaching for the version that best supported the specific policy already in front of them.
Notice what almost never appears in a floor speech or an ad campaign: the top-5%-of-patients figure, which is the more accurate and more actionable one. That is not an accident of taste. Nobody has a bill or a product built around "identify the patients drifting toward the top of the spending distribution and hold them in continuous primary care before they get there." It doesn't sell a supplement, justify a coverage-design change, or fit a thirty-second spot. The blunt, alarming, endlessly reusable number survives because it is useful to many different agendas. The sharp, accurate number does not survive, because it is useful mainly to the agenda nobody is currently funding — which is this one.
The framework's reading
Every source behind this number was generated somewhere other than the relationship. A survey classification decided who counts as chronically ill. A billing and claims system decided how to attribute a fractured arm to a person's chronic-conditions bucket. A communications office decided which nine words would fit on a campaign website. A committee decided which cut of the data would support a bill already being drafted. None of these processes involved anyone watching what happens between a patient and a care team — which is exactly the blind spot the measurement problem describes from a different angle. This primer is that same blind spot, traced through one specific number, from the moment it was generated to the moment it reached a Super Bowl audience.
And here is the turn this framework insists on making, because naming a bad number is not the same as having nothing to say. Properly read, this data is still the strongest possible case for continuous primary care — just not the case as commonly stated. Not because 90% of spending treats chronic disease, which is not true. Because a small, identifiable population of complex patients drives roughly half of all spending, and continuous, arc-oriented care is exactly the mechanism built to catch that population before they arrive there — in the silent years when a drifting condition is still cheap and simple to address, rather than in the expensive years after it has compounded. The accurate number is less quotable. It is also the one worth building a system around.