Understand the distinction: NPI Type 1 (Entity Type Code 1 in NPPES) is issued to individual human health care providers; NPI Type 2 (Entity Type Code 2) is issued to organizations, groups, and other non-individual entities.
In X12 837P professional claims, populate the billing provider loop (Loop 2010AA) with the group/organization NPI Type 2 and the rendering provider loop (Loop 2310B) with the individual provider NPI Type 1 — both are required when a provider bills under a group.
In credentialing systems, credential providers at the Type 1 (individual) NPI level; payer network enrollment and privileging decisions are made per individual provider. The Type 2 NPI is used for group enrollment and billing routing.
When building provider directories, create separate records for Type 1 and Type 2 NPIs and link them via organizational affiliation — FHIR models this as Practitioner (Type 1) linked to Organization (Type 2) via PractitionerRole.
Validate both NPI types in NPPES before use in claims: query the NPPES API with number=NPI_VALUE and confirm enumeration_type matches expected entity type (NPI-1 or NPI-2).
For sole proprietors, note that they may bill under their Type 1 NPI alone or obtain a Type 2 NPI for the practice; some payers require a Type 2 NPI for all group billing regardless of practice size — check payer-specific requirements.
Known gotchas
Submitting a Type 2 NPI in the rendering provider field of an 837 claim (where a Type 1 is expected) causes claim rejections or incorrect attribution at most payers — always validate NPI type before populating claim loops.
A sole proprietor's Type 1 NPI and their practice's Type 2 NPI are different numbers; using them interchangeably in credentialing or claims creates mismatches that trigger payer audits.
NPPES allows a Type 1 provider to be associated with multiple Type 2 organizations, and a Type 2 organization to have multiple Type 1 members — your data model must support this many-to-many relationship rather than enforcing a one-to-one link.
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp