Methodology

Quality ratings

The headline star rating on each acute-care hospital profile is the official CMS Overall Hospital Quality Star Rating, published quarterly by the Centers for Medicare & Medicaid Services. We display this rating verbatim — we do not adjust or re-weight it.

Specialty facilities (children's hospitals, psychiatric hospitals, inpatient rehabilitation facilities, long-term acute care hospitals, and Veterans Affairs medical centers) are not assigned CMS star ratings. Their profiles show measures from the quality program appropriate to the facility type.

Price transparency grades

Each hospital is graded A–F on what CareRanks could find and parse from its machine-readable price file (MRF):

• A — we parsed cash or negotiated prices for 25 or more procedure categories
• B — we parsed cash or negotiated prices for fewer than 25 categories
• C — we parsed the file, but it contains only list (gross) charges
• D — we found an MRF, but could not extract usable prices from it
• F — we could not find a machine-readable price file
• N/A — the facility type is not required to publish one

Grades measure our ability to find and read each hospital's file — they are not a legal compliance determination, which only CMS makes. Files are re-checked monthly, and hospitals can request a re-crawl at any time. Grade criteria are independent from quality ratings — a hospital can be highly rated on quality and poorly rated on transparency, or vice versa.

Procedure-level pricing

Procedure pages aggregate published gross charges, cash prices, and negotiated rates from each hospital's MRF. Where data is incomplete or absent, we say so explicitly rather than guess.

Hospital price files also contain values that are statistically implausible as prices for the procedure listed — $0 or $1 placeholder rows, amounts orders of magnitude outside what every reporting hospital publishes (a common signature of cents recorded as dollars, or of per-visit fragments landing in a total-price column), and sentinel maximum values. We omit such values from display and from every comparison, median, and transparency grade rather than present them as prices. Omission is per value — if a hospital's gross charge for a procedure fails the checks but its cash price is credible, we show the cash price. The checks are statistical fences computed for each procedure and price type across all reporting hospitals, deliberately wide: a value must sit roughly an order of magnitude outside the middle of the reported distribution before it is omitted. They are a versioned part of our editorial methodology, and under the current version they affect about 1.8% of the values we parse. An omitted value may be a placeholder, a unit or data-entry mismatch, a bundled or non-standard amount, or occasionally a genuine but extreme price — we do not adjudicate which, and omission is not a finding that a hospital erred. We never alter or clamp a published value: the hospital's file still says what it says; we simply decline to display values that fail our documented plausibility checks.

How nearby comparisons work

Profile pages may include a "How this facility compares nearby" section. Its comparison group is built the same way for every facility, from CMS data alone:

• Like compares with like. Facilities are compared only within their own vertical and care class — acute-care and critical access hospitals together; psychiatric, rehabilitation, long-term acute care, children's, and VA facilities each only against their own kind; nursing homes with nursing homes; dialysis facilities with dialysis facilities.
• Geography is real and stated. We search 25, then 50, then 100 miles (measured between ZIP-code centers) and stop at the first distance with at least 5 comparable facilities reporting shared measures. The section always says which distance applied. Home health agencies don't serve a single point, so their group is agencies whose CMS service area covers at least half of the same ZIP codes instead.
• A 5-peer floor. Facilities without at least 5 comparable neighbors within 100 miles get no comparison section at all — we never stretch to a statewide group and call it "nearby". A missing comparison section reflects geography and CMS data availability only; it is never a statement about a facility's quality.
• Groups cap at the 8 nearest peers. Superlatives ("highest", "only") and the "best nearby" column appear only when the group is at that full size of 8.
• Medians, not verdicts. Each row shows the facility's value next to the group median. Star medians are shown to one decimal; CMS category ratings (better / average / worse) are compared on their ordinal scale, and when an even-sized group splits between two categories we report the stronger one, so the bar a facility is measured against is never understated. There is no composite score and no ranking.
• Suppression is disclosed. A measure appears only when at least 3 group members report it, and every row says "reported by N of M nearby peers".
• Every section is stamped with the CMS data period and the comparison-method version that produced it.

Comparisons are computed from CMS data only. Claiming a profile or paying CareRanks for anything can never change a comparison group, the values shown, or the sentences generated — the same firewall that applies to grades and ratings.

Updates

Quality data comes from CMS Care Compare and is refreshed when CMS publishes — quarterly for most measures, annually for some programs. Price data is parsed from hospital MRFs; an automated monthly job checks every known price file for updates and re-validates the full dataset before anything is published. Each data-bearing page shows a "data as of" date drawn from the pipeline's own refresh log.

Related: About CareRanks · Data sources