How the ratings are built, and what they cannot tell you

The data

Five sources. All free, all public, none requiring a signup or a paid licence.

1. CMS HCRIS, form HOSP10, FY2024 file. Every hospital that bills Medicare files an annual cost report. Worksheet C states, for each service line, the cost the hospital allocated to it and the charges it billed against it. We use the most recent report per hospital: 6,025 hospitals, of which 4,509 contained usable cost and charge totals. This drives the dollar ratings.

2. NPPES, August 2026. The national provider registry, for surgery centres, imaging centres and urgent care: 42,131 facilities.

3. Hospital price transparency files. Federal rules require every hospital to publish a machine-readable file of its standard charges. Collecting several thousand of those directly is a large scraping job, so we build on the open archive maintained at nathansutton/hospital-price-transparency (Apache 2.0), which has already gathered and normalised them. It covers 3,323 of our hospitals with real published cash and list prices. This drives the self-pay discount and the cash price pages.

4. CMS Care Compare, vintage 2026-05-13. Star ratings and hospital characteristics, plus ambulatory surgery centre case volumes. Provenance matters here and we will not blur it: the CMS host that publishes this data blocks our network, so these files are retrieved from the Internet Archive's mirror of it, captured 2026-09-07. The content is unmodified public-domain CMS output, but it is not a live CMS feed and we do not pretend otherwise. 3,055 of our hospitals carry a star rating; the rest are shown as unrated because CMS itself publishes no rating for them.

5. US Census 2024 Gazetteer and 2023 CBSA delineations. Postcode centroids and metro area definitions, which let us rank a hospital inside its own market rather than only nationally.

Why there are no CPT codes here

Most hospital prices in America are keyed to CPT codes, which the American Medical Association owns and licenses. About four in five rows in the price data are CPT-coded. We do not hold that licence, so we do not publish those rows. Everything on the cash price pages is coded in HCPCS Level II, which CMS owns and which is free to reproduce, and every procedure category elsewhere on the site comes from CMS cost-centre definitions. This costs us coverage and we accept that, because the alternative is republishing licensed content we have no right to.

The same logic applies to Medicare's ambulatory surgery centre rate tables. They sit behind an AMA licence gate, so we do not fetch them, and our fetchers block those paths outright.

The rating

For a service line, markup is charges divided by the hospital's own reported cost for that same line. A hospital that spent one dollar delivering MRI scans and billed thirteen dollars for them has a markup of 13. We compute that per hospital per category, rank all hospitals nationally within each category, and cut the field at the 25th, 50th and 75th percentiles. Bottom quarter gets one dollar sign, top quarter gets four.

As a check that the pipeline is sound: the national median cost-to-charge ratio comes out at 0.261 against a published literature value of about 0.25; government hospitals rate cheapest and large hospitals priciest, both matching known findings.

What this is not

Not a price. Markup is a ratio, not a dollar figure. We deliberately do not publish an estimated dollar amount for your procedure, because we cannot know your plan, your deductible or your case.
Not your bill. What a hospital charges and what an insurer actually pays are different numbers. Insured patients rarely pay charges. Uninsured and out-of-network patients often are billed against them, which is exactly who this is most useful for.
Not procedure-level. Categories are cost-centre level, so "surgery" means the operating room as a whole, not your specific operation.
Not local. Ratings are national percentiles. A four-dollar hospital in an expensive metro may still be the cheapest option near you.
Not current to the day. The FY2024 file is the most recent complete national set.
Not a quality measure. Cheap is not better and expensive is not better. Where CMS publishes a star rating we show it beside the cost so the two can be read together, but the dollar rating itself says nothing about care.
Cash prices are not everyone's price. The prices we publish are what a hospital lists for a self-paying patient. If you have insurance, your price comes from a contract we cannot see.
Cash price outliers are trimmed. Hospitals report billing units inconsistently, so raw minimums and maximums are meaningless. We publish 10th percentile, median and 90th percentile, and drop any hospital more than 25 times away from the median for a given item.

On CPT codes

Procedure-level pricing in the United States is expressed in CPT codes, which the American Medical Association owns and licenses. This site uses only CMS-owned cost-centre definitions, so it carries no AMA-licensed content. Going to true per-procedure pricing means either licensing CPT or restricting output to CMS-owned code sets.

What comes next

Payer-negotiated rates. The price archive we build on carries cash and list prices but not per-insurer negotiated rates. Those exist in the raw hospital files and are the single biggest remaining upgrade: they would turn the insurance section into a real accepted-plans list.
Marketplace issuer network files, for commercial plan acceptance per facility.
Surgery centre cost ratings. Still blocked. Surgery centres file no cost report, and Medicare's own ASC rate tables are behind an AMA licence. What we can say without either is that Medicare pays a surgery centre roughly 62 percent of what it pays a hospital outpatient department for the same work, which is a setting-level average, not a per-facility rating.

Found an error, or run a facility listed here? krishan@thinkscoopinc.com.