How this was made

About

This site explains Medicare’s Hospital Readmissions Reduction Program and shows the FY2027 penalty for every hospital in it. It is built entirely from public federal files and is meant for anyone: patients and families, reporters, and the hospital teams who work with these numbers every day.

Built by Dalton Haslam, MD, MBA, a physician advisor, with Claude Code. It is an independent project, not affiliated with or endorsed by the Centers for Medicare & Medicaid Services. To report an error or suggest a change, use the feedback form (opens in a new tab).

The formula

For each of the six conditions with at least 25 cases, CMS compares the hospital’s excess readmission ratio with its peer group’s median. Each condition above the median contributes:

neutrality modifier × payment weight × (ratio − peer median)

A condition at or below its peer group’s median adds nothing.

The contributions are added, capped at 3%, and subtracted from one to give the payment adjustment factor. This year the neutrality modifier is 0.9628. The cut applies to every traditional-Medicare base inpatient payment the hospital receives from October 1, 2026 through September 30, 2027.

Recomputing every hospital from the values CMS publishes, read at full precision from the Excel version of its file, reproduces its factor exactly for all 2,912 hospitals.

Estimated dollars

CMS publishes each hospital’s percentage, not a dollar amount. This site estimates the base payments the cut applies to, then multiplies by the cut:

cases × case-mix index × (labor amount × wage index + non-labor amount × cost-of-living adjustment)

Cases and case-mix index are each hospital’s transfer-adjusted traditional-Medicare figures from the FY2027 IPPS impact file, which uses claims from two years earlier. The standardized amounts come from the final rule’s standardized-amount tables (correction notice), choosing the column that matches each hospital’s quality-reporting and EHR status as CMS flags it in the impact file.

Across all hospitals this gives about $91.0B in base payments and $342M in penalties. CMS’s own estimate of the penalties was $361M (CMS model of FY2027 penalties from preliminary Jan 2022–Dec 2023 data, current measure set (without sepsis), FY2027 IPPS final rule Table V.I.-05 (91 FR 49892)). It is a ballpark, not a check of these figures. CMS’s table (opens in a new tab)

Accuracy

In an October 2026 check against hospitals’ FY2024 Medicare cost reports, 85% of estimates land within 15% after allowing for payment-rate growth from FY2024 to FY2027 (74% for hospitals under 100 beds).

Assumptions

Where the files leave a choice, this site makes one. Each choice and what it does to the estimate:

Teaching, low-income-patient, low-volume and outlier payments are left out because the cut does not apply to them.

How to read a penalty

• It is relative. CMS sets each median over every hospital in the peer group, so about half of the hospitals measured on a condition sit above it, and most hospitals measured on several conditions get some cut.
• A readmission here is an unplanned return to a hospital within thirty days of discharge, for any reason. Not every readmission is preventable.
• A condition with fewer than 25 cases does not count toward a hospital’s cut.
• Medicare Advantage patients are in the ratios starting with FY2027, but the cut applies only to traditional Medicare payments.
• The data window is July 1, 2023 through June 30, 2025, so the cut reflects care from roughly one to three years before it takes effect.
• The Readmissions program covers the hospitals Medicare pays under its inpatient payment system, so critical access hospitals and the hospitals paid outside it (children’s, cancer, psychiatric, rehabilitation and long-term care hospitals) are not in it and have no page here.

Reading the figures

• 48 hospitals come in above a peer-group median on at least one condition yet take no cut: the condition’s share of the hospital’s Medicare payments is small enough that the reduction rounds to zero. The share cut by number of conditions measured counts them as not cut.
• What 0.01 of ratio is worth: what each 0.01 over the peer median costs, modeled, as on each hospital’s page, as the hospital’s base payments × the condition’s payment weight × the neutrality modifier × 0.01; a hospital under its median loses nothing from 0.01 more until it crosses it. A hospital already at the 3% cap loses nothing more, so it counts as $0; a condition with no payment weight in the file, or a hospital with no base payment, is left out (29 hospital-conditions in all).
• What peer grouping changed: each condition’s flags are drawn again against one median over every hospital measured on it in this file, so only the grouping changes. CMS sets its peer medians itself, so the medians here differ slightly from any CMS would set, and the counts are approximate. Only flags move; no dollars are modeled for a world without peer groups.
• Flagged by number of cases: a ratio is CMS’s predicted readmission rate over the rate expected at an average hospital with similar patients, not a raw rate, and the file gives no interval around it. The middle half shown is the spread of the hospitals observed in a band, not an interval; a band of fewer than 20 hospitals is left out.
• Years cut: the counts are of today’s hospitals, so a hospital that has left the program is missing and the shares can run above CMS’s for a year. A year a hospital was not in the file is left out, never counted as not cut, which is why 963 hospitals were cut in every year they appear but 936 in all 15 years.

Places

Map dots come from each hospital’s street address, geocoded with the U.S. Census Bureau Geocoder (2,526 hospitals). When an address does not match, or its point falls far outside the hospital’s county, the dot goes to the center of its ZIP code (371) or, failing that, its county (15).

Metro areas are the core-based statistical area where each hospital sits, from the IPPS impact file. Regions and divisions follow the Census Bureau.

History

Penalties for FY2013 through FY2026 come from CMS’s archived final-rule supplemental files.

The history’s shares and average cuts count every hospital in each year’s file except Maryland and Puerto Rico hospitals (FY2013–FY2015) and hospitals with no data for any condition (FY2013–FY2018), which the early files list and later files drop.

Annual dollar totals come from CMS, as a payment-rule estimate or, for FY2025, a later simulation of that year under the earlier method, or from KFF Health News reporting. The table gives each year’s total and its source. No total for FY2024 and FY2026.

Feedback, privacy and accessibility

The feedback link opens a Google Form, so Google sees what you send and your IP address. The link fills in only the site and the page you were on. Please leave out any patient information.

How the family of sites handles privacy, including how visits are counted, is on the privacy page, and its accessibility standard is on the accessibility page.

Accessibility

Every chart has a text description, and its numbers are also in a table on the same page. Selecting individual dots or map areas needs a mouse or touch; the tables, search and links reach the same pages by keyboard.

Pinned files

Every input is a public file or an extract of one, pinned byte for byte: the build reads only these copies. Each file’s size and the start of its checksum (a fingerprint of its exact bytes) are below.

Reuse

This site’s text, charts and derived data may be reused with credit to Paid & Penalized (CC BY); its code is MIT-licensed. CMS’s data are U.S. government works. Every table with a download is a CSV file. In those files the CCN is plain text, so open its column as text in a spreadsheet to keep the leading zero; headers name their units; and a condition’s penalized column is blank when the hospital was not measured on it (fewer than 25 cases, or no ratio published).

Every column of these files, and the files themselves, are on the data dictionary page.