Multi-enrollment NPIs with conflicting state addresses
PPEF has ~2.98M rows but only ~2.47M individual NPIs — so ~500K NPIs have multiple enrollments. Many are legitimate (multi-state practice, hospital + private practice, etc.) but a subset have CONFLICTING state addresses that signal stale records: partnership move never refiled, retirement never updated, group-practice split where one half kept the legacy enrollment. Under the 2026 verification rules, stale PECOS practice addresses are a flag. This finding surfaces the conflicting-state cohort per state.
What this means
State Medicaid PI offices
The conflicting-state cohort is a triage layer: any NPI in your state with a competing out-of-state PECOS record is either telehealth (verify), stale (clean up), or fraudulent (refer). Per-state CSV gives your team the working list.
Individual providers
If you moved practice states and did not close the old enrollment, your PECOS has two competing records. The new authoritative-source rule means the wrong one might win during a verification check. File a CMS-855I to close the old enrollment.
Null hypothesis
Every NPI with multiple PPEF enrollment records has consistent state metadata across those records. Multi-state listings only occur with consistent metadata (e.g., explicit telehealth practice).
Denominator
NPIs in PPEF with ≥2 ENRLMT_ID records and ≥2 distinct STATE_CD values across those records. Per-state attribution: each state where the NPI has any enrollment record.
Data source
PPEF (`frontend/data/cms-claims/PPEF_Enrollment_Extract_2026.04.01.csv`). Single pass, GROUP BY NPI, COUNT(DISTINCT STATE_CD) > 1. Per-state CSV at `/api/v1/states/<state>/h39-pecos-multi-state.csv`. See `analysis/h39_pecos_multi_enrollment.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.