Two federal lab registries, and what each cannot see

Two files, two legal objects

The CLIA file’s unit is the establishment: a certified facility, keyed by a CLIA certificate number, with a facility name, an address, a certificate type, and program status codes. It exists because CLIA regulates the place where testing happens.

The billing file’s unit is the billing entity: an NPI that submitted Medicare Part B claims, with the name it bills under and annual payment totals. It exists because Medicare pays claims, and claims are filed by NPIs.

One laboratory business routinely spans both, under different names, and neither file carries the other’s key. The CLIA file has no NPI column. The billing file has no CLIA number column. We found no CMS-published crosswalk from a CLIA certificate number to a billing NPI, and the join we attempted without one is measured below.

Question by question

Does this facility hold an active CLIA certificate, and of what type? Only the CLIA file answers. It carries the certificate type (compliance, waiver, accreditation, PPM, registration), program termination and eligibility codes, and participation dates. The billing file says nothing about certification.

Did this entity bill Medicare, and for how much? Only the billing file answers. For CY2024 it carries each NPI’s total Medicare payment, beneficiary count, service count, and number of billed codes. The CLIA file carries no dollar field at all.

Is this laboratory independent, or hospital-owned, or part of a chain? Neither file answers. We measured this directly: in the CLIA clinical laboratory file, the provider category and subtype codes are constant across the entire file, and the ownership control code assigns hospital-named facilities the same values as plainly freestanding ones (computed 2026-08-21). The only workable public signal is the facility name itself, which is a keyword heuristic, not an ownership record.

Was this laboratory an applicable laboratory under PAMA? Neither file, and no join of the two. The definition at 42 CFR 414.502 (eCFR text as of August 20, 2026, read 2026-08-25) requires that the entity, among other conditions, in a quoted phrase this site returns to constantly: “In a data collection period, receives more than 50 percent of its Medicare revenues” from the CLFS or the Physician Fee Schedule. That is a ratio over one NPI’s revenue composition. The billing file publishes payment totals, never composition. The CLIA file publishes no revenue at all. The ratio is invisible from public data, which is why the status is self-determined.

The two populations, counted

Both counts computed 2026-08-21.

From the CLIA Provider of Services file (Q2 2026 release): 17,682 clinical laboratory rows, of which 1,188 are active and eligible by program code, of which 1,035 remain after removing 133 hospital-keyword names and 20 named chains.

From the Medicare billing file (CY2024 payments): 2,743 NPIs typed “Clinical Laboratory”, of which 2,274 remain after removing 50 hospital-keyword names and 419 named-chain rows, of which 1,729 had CY2024 payments at or above $12,500 and 1,572 at or above $25,000. The $12,500 figure is PAMA’s low expenditure threshold from 42 CFR 414.502, but note the unit mismatch: the regulation defines it over a six month collection period and the file is annual, so the $25,000 line is the conservative doubled-bar reading and the $12,500 line overstates who would clear a six month bar.

So one registry says about one thousand independent active establishments. The other says about seventeen hundred independent above-threshold billers. Neither number is “the labs”, and the two populations cannot be reconciled by the obvious method, which is the next finding.

The join that returned zero

We took the 60 highest-signal independent billers from the Medicare file and looked for each one in the CLIA file by normalized name plus state, with containment matching, not just exact equality. Zero of sixty matched (recorded 2026-08-24).

The reason is the two-objects point above, now measured: the name an entity bills Medicare under is a billing name, often a legal or holding entity; the name on its CLIA certificate is a facility name. For sixty real, currently billing laboratories, those names did not overlap even loosely. Practical consequences, in order of cost:

1. You cannot enrich the billing file with certificate data, or the reverse, by name. Any dataset claiming to have done so at scale should be asked which key it used.
2. A lab checking its own situation needs both of its identities from its own records: its billing NPI for anything about Medicare money, its CLIA number for anything about certification. Each registry will refuse to answer for the other.
3. Counting “laboratories” produces different numbers depending on which registry you ask, and both are defensible. Any single count quoted without naming its registry is underspecified.

Method

Sources: CMS CLIA Provider of Services file, clinical laboratories, Q2 2026 release, and CMS Medicare Physician and Other Practitioners by Provider, CY2024 payments, both fetched via the data.cms.gov API and processed on 2026-08-21. All counts are row counts after stated filters; exclusion counts are printed, not folded in. The hospital and chain filters are name-keyword heuristics and both files’ counts inherit that limit. The 0 of 60 join was run 2026-08-24 with normalization and containment; the regulation quoted is 42 CFR 414.502, eCFR text as of August 20, 2026, read 2026-08-25.

What a single NPI’s public billing looks like against the one threshold that is public is what the checker at pamawatch.com/check shows, and that boundary, proven against suggested against unknowable, is the point of this whole note.