Skip to main content

Reading the data

How PricedCare reads hospital price files: methodology, limits and what we leave out

Where our data comes from, exactly what we do with it, the rules we apply before publishing a comparison, and the things a price on this site cannot tell you.

By Martin S. Nielsen · 4 min read · updated 2026-09-19 · not medical advice

This page is the long version of the note that appears next to every table on the site. It explains where each number comes from, what we do to it, and what we deliberately do not do.

Sources

  • The hospital list. CMS's Hospital General Information dataset: every Medicare-certified hospital in the US with its CMS Certification Number (CCN), address, type and ownership. About 5,400 hospitals. We treat acute-care, critical-access, children's and rural emergency hospitals as active; psychiatric, long-term and federal (VA, military) hospitals are kept in the list but not compared, because they have no shoppable prices.
  • File locations. Each hospital's cms-hpt.txt, a text file the rule requires at the root of its website, listing the URL of its machine-readable file. Where that is missing or blocked, historical file locations from public datasets.
  • The files themselves. The hospital's machine-readable file in the CMS template, CSV (tall or wide layout) or JSON, in version 2 or 3. Files in spreadsheets or in hospitals' own layouts are recorded as unsupported.
  • Geography. Census core-based statistical areas for metro areas, ZIP centroids and place centroids for search, county centroids where a hospital's mailing ZIP is a post-office box. Hospital coordinates from the Census geocoder or, where it has no match, from OpenStreetMap.
  • Codes. The shoppable services list, using CPT, HCPCS and MS-DRG codes. The names on this site are our own plain-English descriptions; the AMA's CPT descriptions are copyrighted and not reproduced.

What we do with a file

  1. Download it, streaming, whatever its size, and read the header for the hospital name, the file date and the hospital's NPIs.
  2. Read every row. Keep rows whose code is on our list, and skip rows that carry a modifier or a drug unit, since those are different prices from the bare code.
  3. Store, per row: gross charge, discounted cash price, each payer-specific negotiated dollar amount with the payer and plan names, and the de-identified minimum and maximum. Percentages and algorithms are not stored, because they are not prices.
  4. Check who the file belongs to. A health system's cms-hpt.txt typically lists the files for all of its hospitals under one website, so the file's own NPIs and name are matched against our list; a file that belongs to a sister hospital is moved there, and a file that matches no hospital we know is set aside rather than attributed by guesswork.
  5. Delete the raw file. Only the extracted rows are kept, with the file's URL and date.

We re-read every file monthly. Hospitals must update at least yearly; many update quarterly.

Rules before we publish a comparison

  • One number per hospital per code. Where a hospital lists the same code on several chargemaster lines, we use the lowest cash line and say how many lines there were. We never average lines, because an average is not a number the hospital published.
  • Amounts under $1 are placeholders, not prices, and are ignored.
  • A procedure-and-metro page is published when at least three hospitals in the metro area list a cash price for the code. Below that, the page exists for people who follow a link but is kept out of search engines, and it says so.
  • Hospitals with no readable price are listed on every page where they would otherwise appear, with the reason we found: no website on record, website blocks automated reading, no file found, unsupported format, code not in file.
  • Every price shows its file date, and a file older than 13 months is marked.

What a price on this site cannot tell you

  • The total. A hospital's price is its facility fee. Physician, anesthesia, pathology and radiology fees are usually separate bills from separate companies.
  • Your price. With insurance, your cost is a share of the negotiated rate for your plan, after your deductible. Without insurance, the cash price is the starting point and financial assistance may be available below it.
  • What will be done. A diagnostic colonoscopy that finds a polyp is billed under a different code. The price is for the code, not the visit.
  • Quality. Nothing on this site says which hospital is better. The published price and the quality of care are, as far as anyone has measured, unrelated.
  • Facilities that are not hospitals. Ambulatory surgery centers, imaging centers and independent laboratories are not covered by the rule and are not in our data. They are often cheaper.

Corrections

If a number on this site does not match the hospital's file, or a hospital is missing or misplaced, write to [email protected] with the page and, if you have it, the file. We correct against the file, not against a phone quote, because the file is the hospital's own published statement.

Related guides