Medicare pays a fixed price per inpatient stay. There are 773 of those prices. Here is every one of them, plotted against how often it gets used — and a slider for the cost of the new thing you want to put inside it.
On August 4 CMS published the FY 2027 inpatient rule. Buried in it: the repeal of the alternative pathway that let an FDA Breakthrough Device designation stand in for proof of substantial clinical improvement when applying for a New Technology Add-on Payment.
That sentence sounds like paperwork. It isn't. Here is the mechanism, which is the whole story.
Medicare pays a hospital one fixed amount per inpatient stay, set by the stay's MS-DRG. It does not matter what the hospital spent. If you sell a technology that gets used during that stay, its cost comes out of that fixed amount. NTAP is the escape hatch: an add-on payment on top of the DRG, calculated as the lesser of 65% of the technology's cost or 65% of the amount by which the case's cost exceeds the DRG payment — 75% for qualified infectious disease products.
Clearance decides what is legal to sell. Payment decides what gets used.
So: how big can a new technology be before the DRG can't hold it? Every dot below is one of the 773 national MS-DRGs in Medicare's 2023 fee-for-service inpatient file. Horizontal: the average Medicare payment for that DRG. Vertical: how many discharges it covered nationally that year. Drag the technology cost and the dots that can no longer contain it turn red.
Set the slider to $25,000 with NTAP off and the picture is stark: the technology costs more than the entire average Medicare payment in 620 of 773 DRGs, covering 6.09 million discharges — 88% of all Medicare fee-for-service inpatient volume. Turn NTAP on at 65% and the hospital's uncovered share drops to $8,750, and the underwater count falls to 220 DRGs and 2.22 million discharges.
That gap — 620 versus 220 — is what the add-on payment is. It is the difference between a technology being unusable in most of American inpatient medicine and being merely expensive. The FY 2027 rule did not remove NTAP. It removed the shortcut to qualifying for it.
Before you design your pivotal trial, find the MS-DRGs your technology will actually be billed inside and look up their average payment. If the number on the slider is bigger than the number in the DRG, your FDA endpoint and your CMS endpoint are two different studies and you need to design them as one. FDA wants safety and efficacy against a placebo or a sham. CMS wants substantial clinical improvement against current standard of care. Different control arm, same enrollment window — if you catch it early.
The FY 2027 rule carries a 2.3% net payment update and $779 million in new technology add-on payments, across 19 newly approved technologies and 41 continuing ones. Sixty products. Against 773 DRGs and 6.9 million discharges, $779 million is 0.75% of the $104.1 billion Medicare paid for those stays.
That is the size of the door. The repeal narrows who gets through it: everyone now demonstrates newness, cost, and substantial clinical improvement independently, with no Breakthrough designation standing in for the third. Reporting indicates the change applies to applications submitted on or after October 1, 2026 — the FY 2028 cycle, not some distant one. The practical translation is 12 to 18 additional months of comparative-effectiveness evidence before you can file, which is 12 to 18 months during which every hospital that uses your device eats the red dot.
Flip the axis toggle to Charges. The whole cloud slides right and the red mostly disappears. At a $50,000 technology cost, only 196 DRGs have average submitted charges below that line — against 729 DRGs when you measure by what Medicare actually pays. The median DRG's submitted charge is 6.0× its Medicare payment.
If you build your market model off chargemaster data — and a startling number of health-tech decks do — you will conclude there is roughly four times more headroom for your technology than exists. Charges are an opening ask nobody pays. Payments are the constraint.
Drag Min annual discharges up from zero. At a $50,000 technology, 729 DRGs are underwater; require 1,000 discharges a year and that drops to 482; require 5,000 and it drops to 243. The dots vanishing are real DRGs, but 263 of the 773 have fewer than 1,000 discharges nationally in a year and together account for 1.6% of all volume — 36 of them have fewer than 100. An average payment computed off 11 discharges (the CMS suppression floor) is not a price signal. Any "we address 700 DRGs" claim is counting rows, not patients.
The most important caveat, and the one this dataset cannot fix. NTAP is calculated off the hospital's cost for the case; the file behind this chart contains payments and charges and no cost column at all. So the red-dot test here is a ceiling test — can the payment even cover the technology on its own — not a margin calculation. A real case has other costs too, which makes the picture worse, not better. Read it as a floor on the problem.
Fee-for-service only. This file covers Original Medicare Part A at IPPS hospitals. It excludes Medicare Advantage, which now covers roughly half of Medicare enrollees and negotiates its own terms; it excludes Medicaid, commercial, uninsured, critical access hospitals and long-term care hospitals. 6.93 million discharges is not "US inpatient volume."
Averages hide the case you care about. Each dot is a national mean across every IPPS hospital. Wage-index adjustment, DSH, IME, and outlier payments move the actual payment for any single stay substantially away from it. The DRG that looks safe on this chart may not be safe in a rural hospital in Vermont ($9,357 average across all DRGs) or generous in Maryland ($21,790, and Maryland is on a global-budget waiver anyway).
The NTAP model here is a simplification. The rule pays the lesser of 65% of the technology cost or 65% of the amount by which case cost exceeds the DRG payment. This chart applies the first branch, which is the generous one. Where a case's total cost lands close to the DRG payment, the second branch binds and the hospital recovers less than the slider implies.
65% is current, not eternal. The marginal NTAP rate was 50% before FY 2020. If you have seen "up to 50%" in circulation, that is the pre-2020 rule; the tool defaults to 65% because that is what the FY 2027 regulation pays.
Vintage is a file date, not a performance year. The performance_year column in this table is null for every row; 2023-12-31 is the publication vintage used as the year. DRG definitions also change annually, so a code's meaning is not stable across years.
Suppression is invisible. CMS drops cells below 11 discharges entirely rather than nulling them, so the true count of low-volume DRG activity is higher than what is plotted. Do not read a missing DRG as a zero.
On July 8, FDA and CMS officials hosted a "clinical AI demo day" at White Oak that was never publicly announced; STAT reviewed the agenda and published it on August 5. Ten companies presented. The agenda covered both regulation and how Medicare should pay.
Scroll the chart. There are 773 rows on this ladder, and not one of them is an autonomous system evaluated the patient. There is no DRG, no code, no rate, no unit of service. Several of the companies in that room are building products for which no payment category exists, which means somebody has to invent one — and the definition of "a unit of service when no clinician performed it" sets the business model for the entire category.
The 773 prices on this chart were not written in a closed room. They were written in the Federal Register, over decades, in public, with comment periods almost nobody used. That process is slow and unglamorous and it is the only venue where a community health center's operations director and Microsoft get the same word count.