Behavioral-health PECOS taxonomy misalignment (highest-recoupment cohort)
Subset of H37 narrowed to behavioral-health NUCC codes — counselors, psychologists, LCSWs, marriage/family therapists, addiction counselors, mental health counselors. Behavioral-health wrong-taxonomy is the highest-recoupment-risk category under the 2026 verification rules: payer rejection is automatic (not a warning), and the rejection covers the entire period the wrong code was in place. This finding surfaces the at-risk cohort per state so PI offices, provider organizations, and individual clinicians can verify before CMS does.
What this means
Behavioral-health providers
Your taxonomy code drives every payer's service-coverage logic. A wrong code that has been in your PECOS record for years can trigger recoupment over that same multi-year window once a payer audit catches it. Run AINPI's lookup, check yours.
Behavioral-health group practices
Roster-wide PECOS audit before the 2026 enforcement window opens. The cost of fixing a CMS-855I refile is trivial; the cost of recoupment on a year of misclassified claims is not.
State Medicaid PI offices
The behavioral-health subset is the highest-priority slice within H37 for SMD-letter Element 4 (other comprehensive measures). The cohort is small enough per state to triage manually.
Null hypothesis
Zero NPIs with a behavioral-health NPPES taxonomy slot have a PECOS PROVIDER_TYPE that does not resolve to a behavioral-health NUCC family via the CMS Medicare ↔ NUCC crosswalk.
Denominator
Subset of H37 where any NPPES taxonomy slot matches the NUCC behavioral-health family (codes 101Y* counselor, 103T* psychologist, 1041C* clinical social worker, 106H* marriage and family therapist, 106E* mental health counselor, 106S* addiction counselor, plus related specializations).
Data source
H37 output filtered by the NUCC behavioral-health subtree. Per-state CSV at `/api/v1/states/<state>/h38-behavioral-health-pecos-mismatch.csv`. See `analysis/h38_behavioral_health_pecos.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.