NPPES remains the authoritative source for MIPS, cost attribution, and ACO enrollment today and shouldn't be replaced. The new CMS National Provider Directory (NPD) is a FHIR-based API layer that adds what NPPES can't explicit provider-organization affiliations, near real-time updates, and digital endpoints. For Healthmonix and Prism, the play is to keep NPPES operational while ingesting NPD alongside it, unlocking a "Provider Network Intelligence" layer.

NPPES Current standard
National Plan and Provider Enumeration System
Purpose
Assigns and maintains National Provider Identifiers (NPIs) — the identity registry for US healthcare providers
Established
1996 (HIPAA Administrative Simplification)
Access model
Monthly full replacement file + weekly incremental flat files (CSV). V2 format as of March 2026.
Data format
CSV flat files, downloadable from CMS
Update frequency
Monthly (full) + weekly (incremental)
Affiliation data
Weak — TIN/NPI relationships only, no explicit org affiliations
FHIR support
None natively — requires external mapping
Data entry
Self-reported by providers — no automated validation
Used for
MIPS eligibility, claims attribution, cost measures, credentialing, payer enrollment
Known issues
20% of data changes annually; CMS found only 28% match vs payer directories; branched/corrected 5,000+ times across industry
CMS National Provider Directory (NPD) New · 2026
directory.cms.gov · v0.5.2 released Apr 15, 2026
Purpose
API-driven, FHIR-based provider network layer — discovery, affiliations, and digital endpoints
Established
2025–2026 (active development; open source on GitHub)
Access model
FHIR API + bulk NDJSON files (compressed with zstd). Not designed for spreadsheet tools.
Data format
NDJSON (newline-delimited JSON), FHIR R4 resources
Update frequency
Near real-time via API; bulk file releases on cadence
Affiliation data
Explicit org relationships — who works with whom, group membership, network participation
FHIR support
Native FHIR R4 API — supports TEFCA, Direct messaging, EHR-to-EHR connectivity
Data entry
Combines CMS internal sources (NPPES, PECOS, claims) + industry partner data via ETL pipelines
Used for
Provider discovery, digital endpoint routing, directory services, interoperability. NOT yet used for MIPS, attribution, or cost.
Status
Pre-authoritative — not yet embedded in CMS scoring, eligibility, or payment programs
Side-by-side comparison
| Dimension | NPPES | NPD (directory.cms.gov) |
|---|---|---|
| Primary question answered | "Who are you?" — provider identity | "How do I find and connect to you?" — provider discovery |
| Architecture | Static flat file export | Live FHIR API + bulk NDJSON |
| Affiliation data | Weak — inferred from billing/TIN | Explicit org relationships, group membership |
| Data freshness | Monthly at best; 20% of records stale at any time | Near real-time via API |
| Interoperability | None — requires custom mapping | Native FHIR R4; TEFCA-ready digital endpoints |
| Authoritative for CMS programs | Yes — MIPS, PECOS, cost, attribution | Not yet — discovery/directory use only |
| Data quality | Self-reported; only 28% match vs payer directories | Combines multiple CMS sources — aims to improve accuracy |
| Accessibility for analysts | CSV — works in Excel, SQL, pandas | NDJSON — requires parsing; not spreadsheet-friendly |
| MIPS eligibility | Yes — current source of truth | No current use |
| Cost attribution / TIN-NPI | Yes — used today | Not yet |
| MVP / ASM attribution | Indirect via claims + NPPES | Future potential — affiliation data could reshape attribution |
| ACO / MSSP alignment | Current pipeline | NPD affiliation data could strengthen ACO network mapping |
| Replace NPPES? | No — not today. NPD is a new layer, not a replacement. | |
The core problem NPD is solving
Why NPPES alone isn't enough
CMS itself has at least five systems managing provider information. NPPES is the de facto standard but has been branched and corrected over 5,000 times across the industry — every payer, health system, and vendor maintains their own silo of fixes. The cost of not having a single source of truth is estimated at $2.76B per year. NPD is CMS's attempt to fix this at the infrastructure level.
What NPD adds that NPPES cannot
Explicit provider-to-organization affiliations, FHIR digital endpoints for EHR-to-EHR data exchange, and a queryable API that answers real-time questions — not just "who is this provider" but "who do they work with, where do they practice, and how do systems connect to them." These are the building blocks for dynamic network accountability.
What this means for Healthmonix and Prism
Today — no operational change required
Continue using NPPES for all current pipelines: MIPS eligibility, TIN/NPI attribution, cost measures, provider lookup. NPD is not authoritative for any of these today. Do not replace NPPES.
Near term — start ingesting alongside NPPES
Begin parsing NPD NDJSON alongside NPPES. Map NPI ↔ organization relationships from NPD affiliation data. Validate differences vs current attribution assumptions — especially for ACO and multi-site clients.
Strategic — Provider Network Intelligence in Prism
NPD enables a "Provider Network Intelligence" layer: real affiliations (not inferred), stronger MSSP/ACO alignment, ASM episode ownership clarity, TEAM bundle coordination. First mover advantage before competitors recognize this shift.
Positioning statement
"While others rely on static NPI data, Healthmonix is building on CMS's next-generation provider network infrastructure — so our attribution logic reflects how care is actually organized, not how it was billed."
Data format: what you're working with
NPPES — what you get today
Monthly CSV (V2 as of March 2026). Fields include NPI, entity type, legal name, taxonomy codes, practice location, mailing address, enumeration date, deactivation status. Weekly incremental file for new/updated/deactivated NPIs. No affiliation fields. No FHIR endpoints. Works in Excel, SQL, pandas with no transformation.
NPD — what you get now
Bulk NDJSON files compressed with zstd — not intended for Excel. Each line is a FHIR resource (Practitioner, Organization, OrganizationAffiliation, PractitionerRole, Endpoint). Requires a JSON parser and ETL pipeline to flatten into relational tables. minilabs has already unpacked this into SQL-query able format.
Bottom line
NPPES remains the operational source of truth for every CMS program today — MIPS eligibility, cost attribution, TIN/NPI relationships, ACO enrollment. Do not change current pipelines.
NPD is infrastructure, not a dataset. It is CMS signaling where provider accountability is heading: from static identifiers toward dynamic, affiliation-aware networks. The practices and platforms that understand this shift early will have a structural advantage as TEAM, ASM, and next-gen ACO models evolve.
For Prism the opportunity is specific: NPD affiliation data is the missing layer between "who submitted this claim" and "who is actually responsible for this patient's care across the network." That gap is exactly where attribution logic breaks down in episode-based cost models — and NPD starts to close it.
NPD is infrastructure, not a dataset. It is CMS signaling where provider accountability is heading: from static identifiers toward dynamic, affiliation-aware networks. The practices and platforms that understand this shift early will have a structural advantage as TEAM, ASM, and next-gen ACO models evolve.
For Prism the opportunity is specific: NPD affiliation data is the missing layer between "who submitted this claim" and "who is actually responsible for this patient's care across the network." That gap is exactly where attribution logic breaks down in episode-based cost models — and NPD starts to close it.
Healthmonix internal analysis · directory.cms.gov · github.com/CMS-Enterprise/npd · May 2026
