How we source, compute, and verify everything on this site
Every number on this site is computed from public CMS data using the published Medicare payment rules, versioned so it can be reproduced later, and cited back to the exact file it came from. This page explains how.
Where the data comes from Copy link
All rate data is ingested from the files CMS publishes for each quarterly Physician Fee Schedule release: the PPRRVU file (RVUs, status indicators, policy indicators, and the conversion factor), the GPCI file (the three geographic indexes per payment locality), the ZIP-code-to-locality crosswalk, per-locality anesthesia conversion factors, OPPS site-of-service caps, and the NCCI edit files. We archive the raw source files as received, so every ingested value can be traced to the row it came from. Every release we hold, with official CMS download links, checksums, and a plain-English guide to each file's columns, is listed in the source-file directory.
One limitation: ZIP-to-locality resolution uses the crosswalk release in effect for the release being priced when we hold one that old, and otherwise falls back to the newest crosswalk we hold. For old claims in areas where locality boundaries have since changed, that fallback can resolve a ZIP to today's locality rather than the one in force then—select the locality directly when that distinction matters.
How rates are computed Copy link
Rates are computed with Medicare's published formula—(work RVU × work GPCI + practice expense RVU × PE GPCI + malpractice RVU × MP GPCI) × conversion factor—using the facility or non-facility practice expense value for the requested setting. Whether a code pays at all is decided by its status indicator, never by whether RVU values happen to be present: a blank RVU is treated as missing, not as zero, and carrier-priced codes are reported as having no national amount rather than a made-up figure. Sequestration is applied only as an optional, final step—never baked into the underlying rate. The full walkthrough is in How Medicare pays.
Which pricing rules we apply—and which we don't Copy link
A computed amount is only as trustworthy as the list of rules that produced it, so we publish
the full list here: every payment rule we apply, every one we apply with a documented
simplification, and every one we deliberately don't. When we don't model a rule, the affected
line is priced at its full amount with a reason saying so, never a guessed discount. API
integrations get the same list from GET /v1/pricing-rules.
Formula & payability
-
PFS payment formula Applied
Amount = (work RVU × work GPCI + practice-expense RVU × PE GPCI + malpractice RVU × MP GPCI) × conversion factor, with the practice-expense RVU matched to the requested setting: facility or office (non-facility).
-
Status-driven payability Applied
Whether a code pays is decided by its status indicator, never by its dollar value: A, R, and T pay; B, I, N, P, M, X, and E are not separately payable; C is carrier-priced with no national amount.
-
Blank RVUs are never zero Applied
A blank RVU (carrier-priced and similar rows) is reported as not payable, with a reason. No page or API response ever converts a blank into $0.
-
Qualifying APM participant conversion factor Applied
When the request sets the qpp flag, the qualifying-participant conversion factor for that release prices the line instead of the standard one.
-
Anesthesia payment Applied
Status-J codes price as (base units + time units) × the locality's anesthesia conversion factor. Base units come from the ingested CMS file; time comes from the request.
Limit: Without a locality and anesthesia time, only the anesthesia conversion factor is shown—a full amount is never estimated from missing inputs.
Geography & setting
-
ZIP to payment locality Applied
ZIPs resolve through the CMS ZIP-to-carrier/locality crosswalk in force for the requested period. A ZIP that spans localities or is missing from the crosswalk is reported as such, not guessed.
Caps & adjustments
-
OPPS payment cap Applied
When CMS publishes an OPPS cap for the code, modifier, and locality, the computed amount is capped at CMS's precomputed price for the requested setting, before sequestration.
-
Sequestration Applied
The sequestration reduction is an explicit request flag applied after the formula and any cap—never baked into a rate. The payment calculator applies it to the Medicare 80% share, the way it appears on a remittance.
Modifiers
-
Professional / technical split (-26, -TC) Partial
For codes whose PC/TC indicator is 1, the -26 and -TC amounts come from CMS's own modifier-specific RVU rows—looked up, not computed as a percentage.
Limit: Component-only indicator values (2, 3, 6) are simplified: those codes price from their base rows, and the split validation only opens the -26/-TC combination for indicator 1.
-
Bilateral surgery (-50) Partial
Indicator 1 pays 150% of the base amount; indicator 2 pays 100% because the RVUs already assume a bilateral procedure; indicator 3 pays both sides in full.
Limit: Indicator 0 currently reports no bilateral payment, where CMS still pays the single-side amount—a documented simplification.
-
Assistant at surgery (-80, -81, -82, -AS) Applied
Paid at the configured assistant percentages (physician and non-physician) when the assistant-surgery indicator permits payment. Indicator 0 prices at the same percentage and carries the documentation condition in the result, since CMS gates it on documented medical necessity rather than barring it. Indicator 1's statutory bar is honored with a reason and no amount.
-
Co-surgeons (-62) Applied
Each co-surgeon prices at the configured co-surgery percentage when the indicator permits it; the documentation-required case carries that condition in the result.
-
Team surgery (-66) Applied
CMS prices team surgery by individual report, so no fixed amount exists. Permitted lines return no amount with that reason rather than a guess.
-
Modifiers without a fixed percentage Not applied
Payment modifiers that carry no fixed fee-schedule percentage (reduced, discontinued, or unusual services, among others) are not priced. The line keeps its base amount and the response says the modifier was not applied.
Multiple procedures (MPPR)
-
Standard multiple-procedure reduction (indicator 2) Applied
Indicator-2 lines on a claim are ranked by fee-schedule amount: 100% for the first, 50% for the second through fifth. CMS suspends the sixth and later for by-report review, so those lines carry no invented amount—only the documented floor.
-
Pre-1996 multiple-procedure schedule (indicator 1) Not applied
Indicator 1 selects a different, older reduction schedule than the standard one. It appears on zero rows in every release we have ingested, so those lines bill at full amount with a reason rather than under the wrong schedule.
-
Endoscopy-family reduction (indicator 3) Partial
Endoscopy families sharing a base code (the ENDO BASE field) collapse into one service—the highest-paid member pays in full, each subsequent member pays its own amount minus the base code's amount—and the combined family total then joins the standard indicator-2 ranking. The base code's own fee-schedule amount is resolved even when it wasn't separately billed.
Limit: Requires the claim's indicator-3 lines to carry a resolvable ENDO BASE code; a line whose base amount can't be resolved bills at full amount with a reason instead of guessing.
-
Diagnostic imaging technical/professional-component reduction (indicator 4) Partial
The technical (TC) and professional (PC) components rank and reduce independently, each in its own pool, at the percentage in effect on the date of service: TC 50% off since 2010-07-01 (25% off before), PC 5% off since 2017-01-01 (25% off 2012-2016). Reduces the pre-OPPS-cap amount, then re-applies the cap—so a top-ranked line whose cap binds is still capped correctly.
Limit: Whether CMS imposes any procedure-count cap on this reduction (the way indicator 2 suspends the sixth-and-later to by-report) is not confirmed well enough for us to apply one, so the reduction has no cap here. A line missing both TC and PC dollar amounts bills at full amount with a reason.
-
Therapy practice-expense reduction (indicator 5) Partial
Only the practice-expense (PE) dollar amount is ranked and reduced—work and malpractice are always paid in full. Ranking runs per unit as well as per procedure (a 2-unit line contributes 2 ranking slots), at 50% off since 2013-04-01 (ATRA §633); the pre-2013-04-01 window also requires the claim type (professional or institutional) to select the correct 20%/25% split.
Limit: Same unconfirmed by-report-cap caveat as indicators 4/6/7. A line missing a PE dollar amount, or dated in the pre-2013-04-01 window without a claim type, bills at full amount with a reason.
-
Diagnostic cardiovascular technical-component reduction (indicator 6) Partial
The technical component ranks and reduces on its own (25% off since 2013-01-01); the professional component, if billed, is always paid in full—CMS never reduces it for this indicator.
Limit: Same unconfirmed by-report-cap caveat as indicators 4/5/7. A line missing a TC dollar amount bills at full amount with a reason.
-
Diagnostic ophthalmology technical-component reduction (indicator 7) Partial
The technical component ranks and reduces on its own (20% off since 2013-01-01—CMS proposed 25%, finalized 20% after public comment); the professional component, if billed, is always paid in full.
Limit: Same unconfirmed by-report-cap caveat as indicators 4/5/6. A line missing a TC dollar amount bills at full amount with a reason.
-
Multi-unit lines in MPPR ranking Partial
Indicator 5 (therapy) ranks per unit as well as per procedure—a multi-unit line contributes one ranking slot per unit, all sharing that line's per-unit practice-expense amount.
Limit: Every other indicator (2/3/4/6/7) still ranks by procedure only; a line's units feed the medically-unlikely-edit check but do not multiply into those rankings.
Claim-level edits
-
NCCI procedure-to-procedure edits Applied
PTP pairs are checked against the NCCI release covering the date of service, honoring each edit's effective window and modifier indicator, including the bypass modifiers on the correct line. A bundled line keeps its computed amount—the flag explains why a payer would deny it.
Limit: When no NCCI release covers the period, the check is reported as absent with a warning—never as "no edits".
-
Medically unlikely edits (MUE) Applied
Line units are checked against the MUE limit for the code, honoring the adjudication indicator that decides whether units aggregate per line or per day.
-
Global surgical periods Not applied
Each code's global-period indicator is ingested and shown on its code page.
Limit: Claim checks do not yet enforce it: no warning when a service falls inside another procedure's global period, and post-operative-period modifiers are not adjudicated. Enforcing this across claims needs visit history a single-claim check does not have.
Versioning: past releases are never edited Copy link
Each quarterly release (year plus A/B/C/D quarter) is stored as its own dataset, and a published release is never edited in place. When CMS revises a file, we ingest the revision as a new version alongside the old one. That's what makes historical lookups trustworthy: a rate retrieved for Q2 2024 always answers with Q2 2024 data, and every page and API response states which release produced its numbers.
Parsing without guessing Copy link
CMS shifts column order, header names, and title rows between releases, so our import locates the header row and maps columns by name against a maintained alias table—never by fixed position. If a required column can't be found, the import stops with an error instead of guessing. Wrong-but-plausible data is worse than no data.
How the explanatory content is written and verified Copy link
The glossary and guides are original prose written from primary CMS sources, and each page lists the sources it draws on, with links. Specific factual claims are tracked individually: each claim is tied to its source, re-verified against the source's actual text, and its verification resets automatically whenever either the claim or the source URL changes. We check source links for broken pages on a schedule. Reference pages show the date their content last materially changed—real revision dates, never decorative ones.
Before we publish a claim, it is cross-checked against independent sources, weighed by authority. CMS's own files, manuals, and Federal Register rules count most. Specialty-society and device-maker documentation comes next—corroborated wherever a claim reflects well on its source. Coding-reference sites count as one vote each, and newsletters or forum posts are leads to go verify, never sources to publish from. A claim needs agreement from independent sources—different organizations, not republications of the same text—ideally including a primary one. Where the underlying data can be downloaded, we parse it ourselves rather than trust a summary: checking claimed NCCI edit pairs against CMS's actual edit files has caught codes that secondary sources said carried an edit but don't. And when credible sources disagree, the claim is held for review rather than resolved by picking one—the same rule as the rates: wrong-but-plausible is worse than absent.
Payer negotiated-rate benchmarks Copy link
Commercial benchmarks come from the machine-readable files insurers publish under the federal Transparency in Coverage rule. We keep professional, fee-for-service prices stated in dollars, group them by code, county and specialty, and report the 25th, 50th and 75th percentiles across every provider, contract and price behind a group. Percentages of Medicare are computed against the Medicare amount for the same locality when the page is rendered, never stored. A negotiated rate describes the contract, not what a claim paid.
We publish a benchmark only when at least 10 providers from at least 5 separately contracting groups stand behind it (tax IDs that contract together, like a health system's, count as one group), and no single group accounts for more than 25% of it. This is the 5/25 test from Statement 6 of the 1996 DOJ/FTC Statements of Antitrust Enforcement Policy in Health Care, the safety zone for price surveys. The Justice Department withdrew those statements in February 2023 and the FTC in July 2023, so they are no longer a safe harbor; we keep the two numeric tests as our own publication rule. We count the group that negotiated the contract rather than each clinician, because one large group can put hundreds of clinicians behind a single negotiated rate. A group is usually one tax ID, but a health system often contracts through dozens of them, so tax IDs that appear together on the same negotiated rates across codes, or that share most of their clinicians, count as one group. A group's share is its share of the observations behind the benchmark, which is its weight in every percentile. Every published benchmark shows how many providers and contracting groups stand behind it, and a benchmark we can't check against this rule is not published.
Some markets pay on two schedules: one for independent practices and another, often much higher, for health systems and other large groups. There a single median describes neither. So we also compute every benchmark separately for groups under 100 clinicians in the county and for larger groups, under the same publication rule. Where both publish and their medians are at least one and a half times apart, the page shows the two side by side instead of the combined median.
What we deliberately don't show Copy link
Official CPT code descriptions are copyrighted by the American Medical Association, so public pages carry our own original descriptions written from primary sources rather than AMA descriptor text. HCPCS Level II descriptions are maintained by CMS and shown freely. And where an answer requires more information—a ZIP that spans two localities, a carrier-priced code, a blank RVU—we say so instead of inventing a number.
Corrections Copy link
If you find a number or a claim you believe is wrong, we want to know: [email protected]. Corrections to explanatory content update the page's revision date; corrections to rate data arrive as new release revisions, preserving what was previously published.