Federally excluded NPIs whose post-exclusion Medicare Part B billing is telehealth-dominant
Subset of H40 with the post-exclusion HCPCS mix filtered to telehealth-specific procedure codes. Federally-excluded providers whose post-exclusion billing concentrates in telehealth-delivery codes is a known fraud pattern — virtual practice lets the excluded provider continue without a physical office that would attract local oversight, and most telehealth platforms maintain their own LEIE/SAM screening that should have caught the exclusion. Cohort is small but each row carries unusually high prosecutorial weight because the "I didn't know I was excluded" defense is weaker for a provider actively credentialed through a telehealth platform. Per-state CSVs at `/api/v1/states/<state>/h42-excluded-telehealth-dominant.csv`. **Methodology note**: the published Medicare Physician & Other Practitioners by Provider AND Service file aggregates `Place_Of_Srvc` to F (Facility) / O (Office) at file build, so claim-level POS 02 / POS 10 codes are not directly recoverable. H42 therefore operationalizes "telehealth" via the HCPCS code list (G2010, G2012, G2061-G2063, 99421-99423, 99441-99443, G0425-G0427, G3002-G3003), which exist primarily or exclusively as telehealth-delivery procedures.
Headline
**Null hypothesis supported.** Zero federally-excluded NPIs in CY 2023 show ≥80% of post-exclusion Medicare Part B services billed under telehealth-specific HCPCS codes. Sensitivity sidecar (60–79%): 0 NPIs; high-confidence (≥95%): 0 NPIs. Two readings are consistent with this result: (a) federal exclusion screening is in fact catching telehealth-specific Part B billing pre-payment, or (b) the post-exclusion cohort billing Part B is too small for the dominant-share threshold to register at all. Use H40 (per-claim recoupment unit) for the headline cohort instead; H42 is intended as a sharpened sub-test.
unit: count
What this means
State Medicaid PI offices
High-prosecutorial-weight cohort — the recoupment math is per-claim and the credentialing-failure trail makes the platform itself a secondary inquiry target. Pair with H40's per-HCPCS detail for the recoupment letter's evidence section.
Regulators
Telehealth-only post-exclusion is a defined pattern — § 455.436 screening was meant to catch exactly this. Where the pattern recurs across platforms, the screening-control failure is systemic, not provider-specific.
Provider organizations
If you operate a telehealth network, the LEIE/SAM screening cadence and the date your credentialing system last refreshed it are the two numbers H42 inverts. Use the cohort to backtest your screening pipeline against AINPI's exclusion register.
Null hypothesis
Zero LEIE/SAM-excluded NPIs have post-exclusion Medicare Part B billing where ≥80% of services bill under the published telehealth-specific HCPCS code list (see methodology note in summary).
Denominator
Federally-excluded cohort (~8,619 NPIs nationally, active LEIE or SAM, score ≥ 1.5), joined against the by-Provider-AND-Service file (H40's match table), filtered to NPIs whose post-exclusion HCPCS distribution puts ≥80% of services under telehealth-specific codes. Strict-post-exclusion temporal filter same as H40 (per-NPI leie_excldate / sam_active_date at (NPI, year) grain). The 80% threshold is publishable; sensitivity at 60% / 95% as sidecar.
Data source
CMS Medicare Physician & Other Practitioners by Provider AND Service (same as H40). Telehealth-HCPCS list applied after H40's join; service-share threshold gates inclusion. Implementation: `analysis/h42_excluded_telehealth_dominant.py` (to be added), depends on H40's compute stage. Reference for the telehealth-HCPCS list: CMS Telehealth Services List (`cms.gov/medicare/coverage/telehealth/list-services`).
Notes
H42 is a pure filter on H40's per-state CSV output — no separate source file read. Telehealth-HCPCS list is from the CMS Telehealth Services List (cms.gov/medicare/coverage/telehealth/list-services). Excludes general E/M codes (99202-99215) that are telehealth-enabled but used predominantly in-person — the published file does not carry the 95/GT modifier needed to distinguish in-person from telehealth for those codes. Sensitivity bands at 60% and 95% published as sidecar so readers can pick their own falsification threshold.