PECOS PROVIDER_TYPE vs NPPES NUCC taxonomy disagreement
CMS designated PECOS as the authoritative source for Medicare enrollment under the 2026 verification rules. State Medicaid systems must demonstrate alignment with PECOS, and the window between "discrepancy found" and "enrollment affected" tightens. This finding cross-references each NPI's PECOS PROVIDER_TYPE_CD (from the CMS Public Provider Enrollment Extract) against its NPPES NUCC primary taxonomy via the CMS Medicare ↔ NUCC crosswalk. Mismatches are the regulatorily significant signal: the provider is enrolled to bill one type of service but registered in NPPES for another.
What this means
State Medicaid CMOs
Under the 2026 verification rules your MMIS must align with PECOS. A non-matching taxonomy on a behavioral-health provider does not generate a warning — it generates a denial, then recoupment over the entire window the wrong code was in place. AINPI's per-state CSV gives your PI team the cohort to pre-emptively verify.
Individual providers
PECOS records do not update themselves. If you changed credentialing, switched specialties, or moved from W2 to private practice and did not refile a CMS-855B/I, your PECOS taxonomy is probably wrong. The fix is your responsibility; the cost of a wrong code now is a recoupment letter.
Provider organizations
Run AINPI's per-state list against your roster. Anyone with a PECOS-NPPES taxonomy mismatch is at active denial risk. The 2026 timelines compress the window between flag and enforcement — preemptive cleanup wins.
Null hypothesis
Every Medicare-enrolled NPI in PPEF has a PROVIDER_TYPE that maps cleanly through the CMS Medicare ↔ NUCC taxonomy crosswalk to the NPI's NPPES primary taxonomy. No mismatches at scale.
Denominator
2,470,908 individual NPIs in the PPEF (2026-04-01 release) with non-null PECOS_ASCT_CNTL_ID + NPI. Each row's PROVIDER_TYPE_CD is mapped through the CMS Medicare ↔ NUCC crosswalk (same crosswalk H10–H13 use); the resolved NUCC code is compared to all 15 NPPES taxonomy slots with switch-aware logic (primary only, then any slot).
Data source
PPEF (`frontend/data/cms-claims/PPEF_Enrollment_Extract_2026.04.01.csv`) × NPPES `cms_npd.practitioner` (BigQuery) × CMS Medicare Provider and Supplier Taxonomy Crosswalk (Oct 2025, fetched fresh per run). Streams the PPEF once, joins via NPI, applies the crosswalk in Python. See `analysis/h37_pecos_taxonomy.py` (pre-registered, not yet shipped).
Pre-registered — results not yet published.
This finding is listed here before results drop. That is the project's trust contract: the null hypothesis and the computation are public first, and numbers follow. Methodology: /methodology.