NPPES-deactivated VA-state NPIs still billing public claims data
NPPES deactivation should mean the provider is no longer in practice. Billing strictly after deactivation is either a data quality problem (NPI reused or misattributed) or evidence of work being done under a closed identifier — both are state PI flags. **Result: 3 of 1,495 VA-state NPPES-deactivated NPIs billed at least one public claims source strictly after their NPPES deactivation date.** Top: SHAHID, MUHAMMAD (deactivated 2015) — Medicaid \$57K + Part B \$178K + Part D \$32K, all three sources, all post-deactivation.
Headline
195 of 78,995 NPPES-deactivated NPIs (across all states, per NDH practice-state retention; NPPES nulls practice state on deactivation) billed at least one public claims source STRICTLY AFTER their NPPES deactivation date. Post-deactivation totals: Medicaid $13,709,307, Medicare Part B $2,841,502 (CY 2023), Medicare Part D $9,044,086 drug cost (CY 2023). 62 of 195 matched NPIs appear in MULTIPLE post-deactivation billing sources, which is a stronger signal than any single source. Match rule: claim month/year > NPPES deactivation month/year. Each match is either a data-quality problem (NPI reused or misattributed) or evidence of work being done under a closed identifier — both are state PI triage flags.
195 / 79.0K = 0.25%
unit: count
Null hypothesis
Zero NPPES-deactivated VA-state NPIs appear in Medicaid Provider Spending, Medicare Part B, or Medicare Part D for the calendar month/year strictly following the NPPES deactivation date.
Denominator
1,495 VA-state NPPES-deactivated NPIs = NDH practitioner._state='VA' ∩ NPPES npi_deactivation_date IS NOT NULL ∩ npi_reactivation_date IS NULL. NPPES nulls practice address on deactivation, so VA-state attribution comes from the NDH side (the H10 "NPPES-deactivated but still listed in NDH" signal extended to its claims-side consequence).
Data source
NPPES (BigQuery bigquery-public-data.nppes.npi_optimized) × HHS Medicaid Provider Spending + Medicare Part B + Medicare Part D. Match rule: claim month/year > NPPES deactivation month/year (strict post-deactivation). See `analysis/claims_sources/nppes_deactivation_join.py`.
Notes
Per-state CSV at /api/v1/states/<state>/h31-deactivated-paid.csv. Match rule for Medicaid: CLAIM_FROM_MONTH > deactivation month/year (month-level precision). For Medicare Part B / Part D: deactivation_year < 2023 (year-level precision since the source files are CY 2023 aggregates). Multi-source matches are the strongest signal — same NPI billing across Medicaid + Medicare after NPPES deactivation is direct evidence of work being done under a closed identifier. State PI triage priority order: (1) multi-source matches, (2) Part D opioid prescribers, (3) high-dollar single-source matches.