NPPES Provider Golden Record

About the Dataset

Every team that builds on NPPES eventually builds the same thing — a cleanup layer to make the raw registry usable. The Python scripts, the address scrubbers, the dedup logic, the exclusion cross-checks, the FHIR mapping work. It's the project nobody budgeted for and everybody ends up owning. We built it once, so your team doesn't have to.

The NPPES Provider Golden Record is the cleaned, deduplicated, enriched version of the Centers for Medicare & Medicaid Services (CMS) National Plan and Provider Enumeration System (NPPES) — refreshed weekly within days of each CMS release.

We validate every practice address against USPS postal records, standardize it to deliverable form, and score each with a confidence rating — so claim denials from bad ZIPs and typos stop at our pipeline, not yours. We surface the duplicate NPI pairs — candidate matches across millions of providers — and resolve them with tiered-confidence scoring, so the same physician doesn't show up three times in your network map. We cross-reference every NPI against CMS PECOS for Medicare enrollment, specialty, and group practice PAC ID — and flag OIG/LEIE exclusions and deactivated NPIs before they reach your billing team.

We fill missing NUCC taxonomy gaps, attach hospital CCN and group practice affiliations from CMS Care Compare, and enrich every address with census tract, CBSA metro, HPSA shortage scores, MUA designations, congressional and state-legislative districts, IANA timezone, and a native Snowflake geo_point GEOGRAPHY column for ST_DISTANCE and ST_DWITHIN spatial queries. And we publish pre-built FHIR R4.0.1 Practitioner, Organization, Location, and PractitionerRole resources — so your next TEFCA, QHIN, HIE, or EHR integration project starts on day thirty instead of day zero.

One weekly-refreshed source of truth for revenue cycle, credentialing, network planning, compliance, and TEFCA-ready interop.

Get Full Access | Snowflake Marketplace | Free Trial

Upgrade Path for CMS NPPES Provider Dataset Subscribers: This is the premium counterpart to the raw CMS NPPES Provider Dataset. The Golden Record delivers every NPI in NPPES after address validation, census/ACS enrichment, PECOS enrollment joins, OIG exclusion checks, duplicate clustering, and FHIR mapping — consolidating work that teams typically distribute across internal data quality, interoperability, and compliance pipelines into a single analytics-ready dataset refreshed weekly on the CMS release cadence.

Backwards-compatible by design: every table name, column, and join path from the CMS NPPES Provider Dataset is preserved in the Golden Record. Existing subscribers can swap the database name in their queries — PROVIDERS, PROVIDERS_ADDRESSES, PROVIDERS_IDENTIFIERS, PROVIDERS_LICENSES, PROVIDERS_TAXONOMIES, and TAXONOMIES keep their original shape, so existing SQL runs unchanged. New enrichment columns and tables are layered alongside.

What You Get

Addresses validated & geocoded	Every practice address USPS-standardized, confidence-scored, and paired with a native Snowflake GEOGRAPHY point for ST_DISTANCE / ST_DWITHIN spatial queries
Duplicate NPIs resolved	Tiered-confidence clustering collapses multi-NPI providers into unique real-world entities, with the canonical and all alternate NPIs preserved
Medicare enrollment joined	PECOS enrollment status, specialty, enrollment state, and group practice PAC ID attached to every NPI
Compliance risks flagged	OIG/LEIE exclusions, deactivated NPIs, and critical data-quality issues surfaced as clean booleans
Taxonomy gaps filled	Missing NUCC primary taxonomy codes inferred and flagged, so specialty classification is complete across every record
Hospital & group affiliations appended	CMS Care Compare hospital CCN affiliations, group practice, medical school, graduation year
Geographic & demographic context enriched	Census tract, CBSA metro, HPSA shortage scores, MUA designations, congressional and state legislative districts, IANA timezone, ACS demographics and economics
Interop resources pre-built	FHIR R4.0.1 Practitioner, Organization, Location, and PractitionerRole resources ready to drop into TEFCA, QHIN, HIE, or EHR pipelines
Coverage	Every US state, DC, and territories
Refresh cadence	Weekly — our pipeline monitors for new CMS NPPES releases and loads the validated, enriched data within days of publication, automatically

Consumer tables (11 in the share):

• Provider master: PROVIDERS, PROVIDERS_ADDRESSES, PROVIDERS_IDENTIFIERS, PROVIDERS_LICENSES, PROVIDERS_TAXONOMIES, TAXONOMIES

• Enrichment outputs: PROVIDERS_ENTITIES, PROVIDERS_FHIR, PROVIDERS_DUPLICATES

• Metadata: PRODUCT_INFO, DATA_DICTIONARY

Overview

The NPPES Provider Golden Record provides analytics-ready access to every US healthcare provider with enrichment that isn't available in the raw CMS feed.

Provider master:

• PROVIDERS – One row per NPI. Identity, lifecycle, secondary-name fields, taxonomy enrichment, OIG exclusion, CMS Care Compare hospital/group affiliations, PECOS Medicare enrollment, duplicate cluster flags, quality signals

• PROVIDERS_ADDRESSES – One row per (NPI, address_type). Both Practice_Location and Business_Mailing rows per provider. Practice rows carry USPS-validated address, geocoded lat/long + native Snowflake geo_point GEOGRAPHY, census tract/block/place codes, ACS demographics, HPSA and MUA shortage designations, congressional + state-legislative districts, IANA timezone — mailing rows are raw NPPES passthrough

• PROVIDERS_IDENTIFIERS – Other provider identifiers (PECOS, Medicaid, prior NPIs, state license numbers per NPPES OTHER_PROVIDER_IDENTIFIER_*). Join PROVIDER_ID = PROVIDERS.ID

• PROVIDERS_LICENSES – State medical licenses (state + license number)

• PROVIDERS_TAXONOMIES – Per-NPI taxonomy rows with the raw NPPES PRIMARY flag (sparsely populated ~15% of rows — for dense primary-taxonomy signal use PROVIDERS.INFERRED_PRIMARY_TAXONOMY_CODE)

• TAXONOMIES – NUCC taxonomy reference (~880 codes with grouping, classification, specialization, definition)

Enrichment outputs:

• PROVIDERS_ENTITIES – One row per unique real-world provider with multi-NPI duplicate clusters pre-collapsed and critical rollups preserved (any_medicare_enrolled, any_oig_excluded) — query this for accurate unique-provider counts

• PROVIDERS_FHIR – FHIR R4.0.1 Practitioner (or Organization), PractitionerRole for Medicare enrollment, and Location resources stored as VARIANT, ready to drop into TEFCA, QHIN, EHR, or HIE pipelines

• PROVIDERS_DUPLICATES – Bridge table of suspected duplicate NPI pairs with similarity scores, match type, confidence tier, and recommended surviving NPI — forensic view behind PROVIDERS_ENTITIES

Metadata Tables

Every Dataplex data product includes these standard metadata tables:

Table	Purpose
PRODUCT_INFO	Single-row product snapshot — current provider count, address count, enrichment coverage, last refresh
DATA_DICTIONARY	Column descriptions for every consumer-facing view

Entity Relationship Diagram

NPPES Provider Golden Record Entity Relationship Diagram

The primary join key is npi for every consumer-facing table. PROVIDERS.ID (VARCHAR UUID) is the FK target for PROVIDERS_IDENTIFIERS.PROVIDER_ID, PROVIDERS_LICENSES.PROVIDER_ID, and PROVIDERS_TAXONOMIES.PROVIDER_ID. Address enrichment joins via NPI: PROVIDERS_ADDRESSES is one row per (NPI, address_type) so each provider has up to two rows (Practice + Mailing).

Data Tables

PROVIDERS — the primary consumer table

One row per NPI. Identity, taxonomy, OIG exclusion, CMS Care Compare, PECOS Medicare enrollment, duplicate flags, and quality scoring. Join to PROVIDERS_ADDRESSES on npi for validated addresses and geo/demographic enrichment. Filter is_canonical = TRUE to dedupe a roster. The id column (VARCHAR UUID) is the FK target for PROVIDERS_IDENTIFIERS, PROVIDERS_LICENSES, and PROVIDERS_TAXONOMIES — join via PROVIDER_ID = PROVIDERS.ID.

Key Features:

• One row per NPI (9.5M+ US healthcare providers including active + deactivated)

• Validated identity and lifecycle fields including the PROVIDER_OTHER_* secondary-name family (maiden, former, DBA/AKA)

• PECOS Medicare enrollment joined per NPI — medicare_enrolled is strict TRUE/FALSE

• OIG/LEIE exclusion flagged per NPI with exclusion type and date

• CMS Care Compare hospital affiliations, group practice, medical school, graduation year

• Duplicate cluster labeling with a stable is_canonical filter for safe roster dedup

• Refreshed weekly on the CMS NPPES release cadence

Key Columns

This table highlights the columns used most often. PROVIDERS has 85+ columns in total including identity fields (id, 10× PROVIDER_OTHER_*). Query DWV.DATA_DICTIONARY WHERE table_name = 'PROVIDERS' for the complete, authoritative spec with types and descriptions — we publish the metadata to Snowflake so it always matches what's in the view.

Column	Type	Description
npi	NUMBER	National Provider Identifier — 10-digit CMS ID, primary key
address_id	VARCHAR	FK to PROVIDERS_ADDRESSES.address_id — join for address + enrichment
entity_type_code	NUMBER	1 = Individual, 2 = Organization
is_individual / is_organization	BOOLEAN	Convenience flags derived from entity_type_code
provider_organization_name	VARCHAR	Legal business name (organizations)
provider_last_name / first_name / middle_name	VARCHAR	Name parts (individuals)
provider_full_name	VARCHAR	Derived display name
provider_credential_text	VARCHAR	MD, DO, NP, etc.
provider_enumeration_date	DATE	When CMS first issued the NPI
is_active	BOOLEAN	Currently active (not deactivated)
npi_status	VARCHAR	active or deactivated
years_in_practice	NUMBER	Years since enumeration
primary_taxonomy_code	VARCHAR	NUCC taxonomy code (e.g., 207R00000X = Internal Medicine)
specialty_description	VARCHAR	Human-readable specialty from NUCC lookup
is_specialty_inferred	BOOLEAN	TRUE when taxonomy was inferred to fill a NPPES gap
is_excluded	BOOLEAN	On the OIG/LEIE exclusion list
exclusion_type_description	VARCHAR	OIG exclusion reason
exclusion_date	DATE	Date added to OIG exclusion list
medicare_enrolled	BOOLEAN	Enrolled in Medicare per PECOS (strict TRUE/FALSE)
medicare_specialty_description	VARCHAR	PECOS-reported Medicare specialty
medicare_enrollment_state	VARCHAR	State of PECOS enrollment
medicare_group_practice_pac_id	VARCHAR	PECOS group practice PAC ID
medical_school	VARCHAR	From CMS Care Compare
graduation_year	NUMBER	From CMS Care Compare
hospital_affiliation_count	NUMBER	Number of CCN affiliations (0–3)
group_practice_name	VARCHAR	Primary group practice from CMS Care Compare
is_canonical	BOOLEAN	TRUE for the canonical NPI in a duplicate cluster — filter for a deduplicated roster
is_duplicate_suspected	BOOLEAN	Participates in a high-confidence duplicate pair
duplicate_group_id	NUMBER	Cluster ID across all members (NULL if singleton)
duplicate_cluster_size	NUMBER	Count of NPIs in this cluster
has_quality_flags	BOOLEAN	Any active data quality flag
has_critical_quality_flag	BOOLEAN	Critical-severity flag (e.g., undeliverable address)
overall_quality_score	FLOAT	0.0–1.0 composite quality score
created_at	TIMESTAMP	When the source data was last loaded
updated_at	TIMESTAMP	When this record was last refreshed

PROVIDERS_ADDRESSES — validated addresses with full geo and demographic enrichment

One row per (NPI, address_type) — both Practice_Location and Business_Mailing rows per provider. Practice rows carry the full enrichment stack (USPS validation, Geocodio census/CBSA/HPSA/MUA, lat/lng/geo_point GEOGRAPHY, political districts, timezone, ACS demographics). Mailing rows are raw NPPES fields only with enrichment columns NULL.

Key Features:

• One row per (NPI, address_type) — ~19M rows total, ~9.5M practice + ~9.5M mailing

• Raw NPPES fields: street1, street2, city, state, zipcode, countrycode, telephone_number, fax_number, address_type, id (VARCHAR UUID), provider_id (VARCHAR UUID FK to PROVIDERS.ID)

• Practice rows carry USPS validation with address_validation_status indicating valid / corrected / undeliverable / not_found (see status values below)

• Geocoded lat/long coordinates and native geo_point GEOGRAPHY on practice rows

• Census tract, block, place FIPS codes

• ACS demographics and economics (population, median age, race/ethnicity, median household income, housing, education, veteran status)

• HRSA HPSA shortage scores (primary care, mental health, dental) and MUA designations

• Congressional and state legislative districts

• IANA timezone with UTC offset and DST observance

Filter WHERE address_type = 'Practice_Location' for enriched geo analytics. Filter WHERE address_type = 'Business_Mailing' for correspondence addresses.

Key Columns

This table highlights the columns used most often. PROVIDERS_ADDRESSES has ~60 columns in total. Query DWV.DATA_DICTIONARY WHERE table_name = 'PROVIDERS_ADDRESSES' for the complete, authoritative spec — we publish the metadata to Snowflake so it always matches what's in the view.

Column	Type	Description
id	VARCHAR(36)	Row PK
provider_id	VARCHAR(36)	FK to PROVIDERS.ID
npi	NUMBER	NPI — matches PROVIDERS.NPI
address_type	VARCHAR	Practice_Location or Business_Mailing
street1 / street2 / city / state / zipcode	VARCHAR	Address as filed with CMS
countrycode	VARCHAR	ISO country code (US)
telephone_number / fax_number	VARCHAR	Contact numbers as filed
validated_address_line1 / validated_city / validated_state / validated_zip / validated_zip4	VARCHAR	USPS-validated canonical address (practice rows only)
latitude / longitude	FLOAT	Geocoded coordinates (practice rows only)
geo_point	GEOGRAPHY	Native Snowflake spatial type — use with ST_DISTANCE, ST_DWITHIN, ST_CONTAINS (practice rows only)
address_validation_status	VARCHAR	Outcome of address verification (practice rows) — see status values below
address_confidence_score	FLOAT	0.0–1.0 validation confidence
state_fips / county_fips	VARCHAR	Census FIPS codes
county_name	VARCHAR	County display name
tract_code / block_code / block_group	VARCHAR	Census Bureau tract/block geography
metro_area_name / metro_area_code / metro_area_type	VARCHAR	CBSA name, code, metropolitan/micropolitan
combined_statistical_area_name	VARCHAR	CSA (where applicable)
congressional_district / congressional_district_name	VARCHAR	US House district
state_senate_district / state_house_district	VARCHAR	State legislative districts
timezone_name	VARCHAR	IANA timezone (e.g., America/New_York)
timezone_utc_offset	NUMBER	UTC offset in hours
total_population	NUMBER	ACS total population for the tract
median_household_income	NUMBER	ACS median household income
pct_bachelors_or_higher	FLOAT	ACS educational attainment
is_in_hpsa	BOOLEAN	County has an active HRSA HPSA designation
hpsa_primary_care_score / mental_health_score / dental_score	NUMBER	HPSA severity (0–26, higher = more severe shortage)
is_in_mua	BOOLEAN	HRSA Medically Underserved Area designation
is_rural	BOOLEAN	Not metropolitan
created_at / updated_at	TIMESTAMP	Lineage

address_validation_status values

What each outcome means when you're working with the data:

Value	What it means for you
valid	The address as filed with CMS is deliverable and standardized — no changes needed. Safe to use practice_* or validated_* columns interchangeably.
corrected	The filed address had issues (typos, non-canonical casing, missing ZIP+4). The validated_* columns contain the corrected, deliverable version — use those for mailing and claim submission, not the practice_* originals.
undeliverable	Address verification determined the location cannot receive mail (invalid apartment, deleted street, etc.). Flag these providers for outreach or directory correction before using them for claims or mailings.
not_found	The address could not be matched against authoritative address records — could be a new construction, an irregular PO box format, or a data-entry error. Verify before mailing or submitting claims.

PROVIDERS_ENTITIES — deduplicated provider roster

One row per unique real-world provider. Multi-NPI duplicate clusters collapsed into a single canonical entity with the full NPI list preserved in all_npis. Critical rollups (any_oig_excluded, any_medicare_enrolled, any_has_critical_quality_flag) aggregate across cluster members so compliance signals are never silently deduped away.

Key Features:

• Use this for accurate unique-provider counts (avoid inflating from multi-NPI clusters)

• all_npis ARRAY preserves every NPI in the cluster for roster remediation

• Any OIG exclusion or Medicare enrollment on any cluster member is surfaced

Key Columns

This table highlights the columns used most often. PROVIDERS_ENTITIES has 38 columns in total. Query DWV.DATA_DICTIONARY WHERE table_name = 'PROVIDERS_ENTITIES' for the complete spec.

Column	Type	Description
entity_id	NUMBER	Stable cluster key (duplicate_group_id for clusters, canonical_npi for singletons)
canonical_npi	NUMBER	Representative NPI for the entity
all_npis	ARRAY	Every NPI in this entity (length 1 for singletons)
npi_count	NUMBER	Cluster size
any_medicare_enrolled	BOOLEAN	TRUE if ANY cluster NPI is Medicare-enrolled per PECOS
any_oig_excluded	BOOLEAN	Compliance rollup — TRUE if ANY cluster NPI is on the OIG/LEIE list
any_has_quality_flags	BOOLEAN	TRUE if ANY cluster NPI has a data quality flag
any_has_critical_quality_flag	BOOLEAN	TRUE if ANY cluster NPI has a critical quality flag
entity_type_code	NUMBER	1 = Individual, 2 = Organization
provider_full_name	VARCHAR	Derived display name from canonical NPI
provider_organization_name / last/first/middle_name	VARCHAR	Name parts from canonical NPI
provider_credential_text	VARCHAR	Credentials from canonical NPI
primary_taxonomy_code / specialty_description	VARCHAR	NUCC taxonomy from canonical NPI
cms_specialty	VARCHAR	CMS Care Compare specialty from canonical NPI
is_active	BOOLEAN	Canonical NPI's active status
npi_status	VARCHAR	active or deactivated
is_excluded / exclusion_type_description / exclusion_date	BOOLEAN / VARCHAR / DATE	Canonical NPI's OIG status
medical_school / graduation_year / hospital_affiliation_count / group_practice_name	—	From CMS Care Compare
overall_quality_score	FLOAT	Composite score from canonical NPI
address_id	VARCHAR	FK to PROVIDERS_ADDRESSES (canonical NPI's practice address)
created_at / updated_at	TIMESTAMP	Lineage

PROVIDERS_FHIR — drop-in FHIR R4.0.1 resources

One row per NPI, 1:1 with PROVIDERS. Three FHIR R4.0.1 resources stored as VARIANT: a Practitioner or Organization (switched on entity_type_code), a PractitionerRole representing Medicare enrollment (NULL when not PECOS-enrolled), and a Location with coordinates. Ready to drop directly into TEFCA, QHIN, EHR, or HIE pipelines — no FHIR transformer required.

Key Features:

• US Core-aligned field conventions

• Vendor-scoped extensions for OIG/LEIE exclusion on Practitioner.extension

• meta.lastUpdated preserved from source

• Conditional PractitionerRole (Medicare enrollment only) mirrors real-world semantics

Column Reference

All 7 columns:

Column	Type	Description
npi	NUMBER	National Provider Identifier (1:1 with PROVIDERS.npi)
fhir_resource_type	VARCHAR	Practitioner (individual) or Organization (Type 2 NPI)
fhir_practitioner	VARIANT	FHIR R4.0.1 Practitioner or Organization resource as JSON
fhir_practitioner_role	VARIANT	FHIR R4.0.1 PractitionerRole for Medicare enrollment (NULL when not enrolled)
fhir_location	VARIANT	FHIR R4.0.1 Location resource with coordinates
created_at	TIMESTAMP	When the source data was last loaded
updated_at	TIMESTAMP	When this record was last refreshed

PROVIDERS_DUPLICATES — bridge table of suspected duplicate pairs

Tiered-confidence duplicate NPI pairs with similarity scores and recommended primary NPI. The forensic view behind PROVIDERS_ENTITIES — use this to audit dedup quality or build custom dedup logic.

Column Reference

All 10 columns:

Column	Type	Description
npi_1	NUMBER	First NPI in the pair
npi_2	NUMBER	Second NPI (always greater than npi_1)
similarity_score	FLOAT	0.0–1.0 weighted similarity score across name, address, phone, and taxonomy
confidence_tier	VARCHAR	Evidence strength for treating the two NPIs as the same entity — see tier values below
match_type	VARCHAR	Pattern of similarity that triggered the match — see match type values below
entity_class	VARCHAR	individual or organization
recommended_primary	NUMBER	Suggested survivor NPI (lower-numbered = earlier CMS registration)
detection_date	DATE	When the pair was first detected
created_at	TIMESTAMP	When the source data was last loaded
updated_at	TIMESTAMP	When this record was last refreshed

confidence_tier values

How much trust to place in a pair when collapsing a roster:

Value	What it means for you
high_confidence	Strong evidence the two NPIs are the same real-world provider (exact name + shared address or phone). Safe to use directly for roster deduplication — PROVIDERS_ENTITIES pre-collapses these clusters for you.
possible_match	Partial similarity that warrants a human review step before deduplication. Useful for surfacing dedup candidates in an MDM workflow, but don't merge blindly.

match_type values

The pattern of similarity that triggered the pair:

Value	What it means for you
exact_match	Name, address, and/or phone match character-for-character on both NPIs. Highest-confidence category.
probable_match	Strong fuzzy similarity (e.g., name + shared practice address or phone). Meaningful overlap but with minor variations.
fuzzy_variant	Name spelling variants sharing location or taxonomy signals — typically hyphen/space/punctuation differences, or common name aliases. Lower precision but surfaces legitimate name changes and data-entry inconsistencies.

PROVIDERS_IDENTIFIERS — other provider identifiers

Other provider identifiers per NPI — PECOS enrollment IDs, Medicaid provider IDs, prior NPIs, state license numbers per NPPES OTHER_PROVIDER_IDENTIFIER_* spec. One row per (provider, identifier). ~2.76M rows.

Key Features:

• ~2.76M identifier rows across every NPI that filed one

• Join PROVIDER_ID = PROVIDERS.ID

• Includes ALL NPIs (canonical + non-canonical) so historical claim joins don't silently drop deactivated providers

Column	Type	Description
id	VARCHAR(36)	Row PK
provider_id	VARCHAR(36)	FK to PROVIDERS.ID
identifier	VARCHAR	The other identifier value
type_code	VARCHAR	NPPES Other Provider Identifier Type Code
type_value	VARCHAR	Human-readable type label
state	VARCHAR	State code where the identifier was issued
issuer	VARCHAR	Issuing authority
created_at / updated_at	TIMESTAMP	Lineage

PROVIDERS_LICENSES — state medical licenses

State medical licenses per provider. One row per (provider, license). ~6.96M rows.

Key Features:

• ~6.96M license rows

• Join PROVIDER_ID = PROVIDERS.ID

Column	Type	Description
id	VARCHAR(36)	Row PK
provider_id	VARCHAR(36)	FK to PROVIDERS.ID
licenses_number	VARCHAR	License number as filed
state	VARCHAR	State code that issued the license
created_at / updated_at	TIMESTAMP	Lineage

PROVIDERS_TAXONOMIES — per-NPI taxonomy rows

Per-NPI taxonomy rows with the raw NPPES PRIMARY flag preserved. One row per (NPI, taxonomy) — up to 15 taxonomies per provider. ~9.94M rows. The HEATLHCARE_PROVIDER_TAXONOMY_CODE column name reflects the spelling carried over from the source NPPES schema.

The raw PRIMARY flag is sparsely populated (~15% of rows) per NPPES spec. For a dense inferred-primary signal, use PROVIDERS.INFERRED_PRIMARY_TAXONOMY_CODE on the main PROVIDERS table — we use a ROW_NUMBER fallback to pick the best taxonomy when the raw flag is missing.

Key Features:

• ~9.94M taxonomy rows

• Join PROVIDER_ID = PROVIDERS.ID

• Join HEATLHCARE_PROVIDER_TAXONOMY_CODE = TAXONOMIES.CODE to get classification/specialization names

Column	Type	Description
id	VARCHAR(36)	Row PK
provider_id	VARCHAR(36)	FK to PROVIDERS.ID
taxonomy_id	VARCHAR(36)	FK to TAXONOMIES.ID
heatlhcare_provider_taxonomy_code	VARCHAR	NUCC taxonomy code
primary	BOOLEAN	Raw NPPES primary flag — sparsely populated
created_at / updated_at	TIMESTAMP	Lineage

TAXONOMIES — NUCC taxonomy reference

NUCC (National Uniform Claim Committee) taxonomy code definitions. ~880 rows, updated quarterly. Join from PROVIDERS_TAXONOMIES.HEATLHCARE_PROVIDER_TAXONOMY_CODE or PROVIDERS.PRIMARY_TAXONOMY_CODE to TAXONOMIES.CODE for classification names.

Column	Type	Description
id	VARCHAR	Row PK
code	VARCHAR	10-char NUCC taxonomy code
grouping	VARCHAR	Top-level grouping (e.g., Allopathic & Osteopathic Physicians)
classification	VARCHAR	Mid-level classification (e.g., Internal Medicine)
specialization	VARCHAR	Specialization (e.g., Cardiovascular Disease)
definition	VARCHAR	NUCC's plain-language definition
notes	VARCHAR	Additional NUCC notes
display_name	VARCHAR	Pre-concatenated human-readable name
section	VARCHAR	NUCC spec section reference
created_at / updated_at	TIMESTAMP	Lineage

Data Quality

Standardization

Every address is USPS-validated and geocoded, with AI-driven normalization as a fallback for addresses USPS can't resolve. Taxonomy codes are filled in via inference when NPPES is missing a primary — is_specialty_inferred flags which records were enriched. NPI lifecycle fields (is_active, npi_status) are computed from deactivation and reactivation dates to handle NPPES's sparse status flagging.

Data Freshness

Check when data was last refreshed using PRODUCT_INFO:

SELECT product_name, provider_count, address_count,
       enrichment_coverage_pct, last_updated_at, version
FROM DWV.PRODUCT_INFO;

Getting Started

Platform Schema Reference

This dataset is available on Snowflake Marketplace. Queries use a schema-only reference — the database is set by the share context:

Platform	Schema	Example
Snowflake	DWV	DWV.PROVIDERS

Discover Available Data

Start with PRODUCT_INFO for a product-wide snapshot and DATA_DICTIONARY for column-level descriptions on every view.

Snowflake

-- Product snapshot: freshness and enrichment coverage
SELECT product_name, provider_count, address_count,
       enrichment_coverage_pct, duplicate_pairs_count,
       active_provider_count, last_updated_at, version
FROM DWV.PRODUCT_INFO;

Snowflake

-- Column-level data dictionary for every consumer-facing view
SELECT table_name, column_name, data_type, column_description
FROM DWV.DATA_DICTIONARY
WHERE table_name = 'PROVIDERS'
ORDER BY ordinal_position;

Foundational Queries

Specialty breakdown with Medicare enrollment rate

Snowflake

SELECT specialty_description,
       COUNT(*) AS total_providers,
       SUM(CASE WHEN is_individual THEN 1 ELSE 0 END) AS individuals,
       SUM(CASE WHEN is_organization THEN 1 ELSE 0 END) AS organizations,
       SUM(CASE WHEN medicare_enrolled THEN 1 ELSE 0 END) AS medicare_enrolled
FROM DWV.PROVIDERS
WHERE is_active AND is_canonical
GROUP BY specialty_description
ORDER BY total_providers DESC
LIMIT 20;

Revenue Cycle: Reduce Claim Denials from Bad NPI Data

See original CMS addresses next to USPS-corrected addresses side by side — the records that would have triggered denials on the raw NPPES feed.

Snowflake

SELECT p.npi, p.provider_full_name, p.specialty_description,
       a.practice_city AS original_city, a.practice_state AS original_state,
       a.validated_city, a.validated_state, a.validated_zip,
       a.address_validation_status
FROM DWV.PROVIDERS p
JOIN DWV.PROVIDERS_ADDRESSES a ON p.address_id = a.address_id
WHERE a.address_validation_status = 'corrected'
ORDER BY p.npi
LIMIT 25;

Spatial and Demographic Analytics

Market sizing: provider density by metro with ACS demographics

Snowflake

SELECT a.metro_area_name,
       COUNT(DISTINCT p.npi) AS providers,
       a.median_household_income,
       a.total_population,
       ROUND(COUNT(DISTINCT p.npi) * 10000.0
             / NULLIF(a.total_population, 0), 2) AS providers_per_10k,
       a.is_in_hpsa
FROM DWV.PROVIDERS p
JOIN DWV.PROVIDERS_ADDRESSES a ON p.address_id = a.address_id
WHERE p.is_active AND p.is_canonical
  AND a.metro_area_name IS NOT NULL
GROUP BY a.metro_area_name, a.median_household_income,
         a.total_population, a.is_in_hpsa
ORDER BY providers DESC
LIMIT 20;

Network adequacy: providers within 10 miles of an anchor NPI (native geo_point)

Anchor the search to a real provider's practice location and find every other provider within a radius. Swap anchor_npi for any NPI to re-target the search.

Snowflake

-- Anchor: find everyone within 10 miles of a reference provider's practice location.
WITH anchor AS (
    SELECT a.geo_point,
           p.provider_full_name || ' (' || a.validated_city || ', ' || a.validated_state || ')' AS anchor_label
    FROM DWV.PROVIDERS p
    JOIN DWV.PROVIDERS_ADDRESSES a ON p.address_id = a.address_id
    WHERE p.npi = 1003000126   -- anchor_npi: swap for any NPI
)
SELECT p.npi, p.provider_full_name, p.specialty_description,
       a.validated_city, a.validated_state,
       anchor.anchor_label,
       ROUND(ST_DISTANCE(a.geo_point, anchor.geo_point) / 1609.344, 1) AS distance_miles
FROM DWV.PROVIDERS p
JOIN DWV.PROVIDERS_ADDRESSES a ON p.address_id = a.address_id
CROSS JOIN anchor
WHERE p.is_active AND p.is_canonical
  AND p.npi != 1003000126
  AND ST_DWITHIN(a.geo_point, anchor.geo_point, 16093)   -- 16,093 meters = 10 miles
ORDER BY distance_miles
LIMIT 25;

TEFCA and FHIR-Ready Interop

Drop-in FHIR resources for EHR, HIE, and QHIN pipelines

Three FHIR R4.0.1 resources per NPI — a Practitioner or Organization, a PractitionerRole for Medicare enrollment, and a Location. Inspect the VARIANT in Snowsight's JSON viewer to verify conformance.

Snowflake

SELECT npi,
       fhir_practitioner,
       fhir_practitioner_role,
       fhir_location
FROM DWV.PROVIDERS_FHIR
LIMIT 10;

FHIR bulk export: one resource type per row for NDJSON streaming

Compose a FHIR bulk export bundle — UNION ALL the three resource types into a single NDJSON-shaped stream for QHIN or $export pipelines.

Snowflake

SELECT 'Practitioner'     AS resource_type, fhir_practitioner      AS resource
FROM DWV.PROVIDERS_FHIR WHERE fhir_practitioner IS NOT NULL
UNION ALL
SELECT 'PractitionerRole' AS resource_type, fhir_practitioner_role AS resource
FROM DWV.PROVIDERS_FHIR WHERE fhir_practitioner_role IS NOT NULL
UNION ALL
SELECT 'Location'         AS resource_type, fhir_location          AS resource
FROM DWV.PROVIDERS_FHIR WHERE fhir_location IS NOT NULL
LIMIT 30;

Medicare Network Planning: PECOS-Enrolled Providers in HPSA Areas

Cross-reference PECOS Medicare enrollment with HRSA HPSA shortage designations to surface in-network providers where they're most needed.

Snowflake

SELECT p.npi, p.provider_full_name, p.specialty_description,
       p.medicare_specialty_description, p.medicare_enrollment_state,
       a.validated_city, a.validated_state,
       a.hpsa_primary_care_score
FROM DWV.PROVIDERS p
JOIN DWV.PROVIDERS_ADDRESSES a ON p.address_id = a.address_id
WHERE p.is_active AND p.is_canonical
  AND p.medicare_enrolled
  AND a.is_in_hpsa
ORDER BY a.hpsa_primary_care_score DESC
LIMIT 25;

Roster Dedup Without Losing Signal

High-confidence duplicate NPI pairs with similarity scores

Tiered-confidence duplicate pairs with match type and recommended primary NPI — safe to use for roster collapse without losing historical claim lineage.

Snowflake

SELECT d.npi_1, d.npi_2, d.confidence_tier, d.similarity_score,
       d.match_type, d.entity_class,
       COALESCE(p1.provider_last_name, p1.provider_organization_name) AS name_1,
       COALESCE(p2.provider_last_name, p2.provider_organization_name) AS name_2,
       d.recommended_primary
FROM DWV.PROVIDERS_DUPLICATES d
JOIN DWV.PROVIDERS p1 ON d.npi_1 = p1.npi
JOIN DWV.PROVIDERS p2 ON d.npi_2 = p2.npi
WHERE d.confidence_tier = 'high_confidence'
ORDER BY d.similarity_score DESC
LIMIT 25;

HIPAA Compliance: OIG-Excluded Providers Still Active in NPPES

OIG/LEIE-flagged NPIs that are still showing as active in NPPES — pull this before billing or claim submission.

Snowflake

SELECT p.npi, p.provider_full_name, p.specialty_description,
       p.exclusion_type_description, p.exclusion_date,
       p.is_active, p.npi_status
FROM DWV.PROVIDERS p
WHERE p.is_excluded
ORDER BY p.exclusion_date DESC
LIMIT 25;

Who Uses This Data

Common Use Cases

• Revenue cycle: reduce first-pass claim denials — Replace raw NPPES with validated, USPS-corrected addresses, verified NPI lifecycle status, and PECOS-confirmed Medicare enrollment to cut denials caused by bad provider master data

• Biennial revalidation and HIPAA compliance — OIG/LEIE exclusion flags and revalidation-aware NPI status help meet the 2026 CMS biennial revalidation mandate and reduce exposure to HIPAA fines

• TEFCA, QHIN, and HIE interoperability — Query pre-built FHIR R4.0.1 Practitioner, Organization, PractitionerRole, and Location resources for drop-in TEFCA and bulk-export pipelines without building a FHIR transformer

• Provider network design for VBC and ACO programs — Build deduplicated provider networks with PROVIDERS_ENTITIES, cross-reference PECOS enrollment, and filter by HPSA shortage and congressional district

• Network adequacy modeling and market sizing — Use native geo_point GEOGRAPHY with ST_DWITHIN / ST_DISTANCE plus ACS demographics (median household income, population, education) and CBSA metro boundaries

• Provider roster master data management (MDM) — PROVIDERS_DUPLICATES and PROVIDERS_ENTITIES together provide a golden record of unique real-world providers with the signal (OIG, Medicare, quality) preserved across every multi-NPI cluster

• Underserved-area provider analytics — HRSA HPSA primary care, mental health, and dental shortage scores plus MUA/MUP designations layered on every provider for rural health, CMMI, and 340B analysis

Related Datasets & Research

This dataset pairs well with:

• CMS NPPES Provider Dataset — The raw source registry that this Golden Record is built on. Subscribe to the raw feed if you need the unprocessed relational tables (addresses, licenses, identifiers, taxonomies) for custom research; subscribe to the Golden Record if you want the same provider universe cleaned, deduped, enriched, and FHIR-ready

• HRSA Healthcare Resources Dataset — Source of the county-level HPSA, MUA, and AHRF workforce signals used here. Join to HRSA for workforce rate analysis by county

• CMS Data Feeds Dataset — Complete CMS data catalog including Care Compare and PECOS source feeds

• TEFCA / QHIN implementations — Drop-in FHIR R4.0.1 resources compatible with the ASTP TEFCA Common Agreement and the QHIN Directory schema

Frequently Asked Questions

How is the NPPES Provider Golden Record different from the raw CMS NPPES Provider Dataset? The raw dataset is the unprocessed NPI registry as CMS publishes it weekly — addresses with typos and outdated ZIPs, no deduplication, no Medicare enrollment, no OIG exclusion flags, no FHIR mapping, no geo enrichment. The Golden Record is that same universe after USPS address validation (around 23% of addresses get corrected), geocoding with census and ACS enrichment, PECOS Medicare enrollment joins, OIG/LEIE exclusion checks, CMS Care Compare hospital and group affiliations, NUCC taxonomy gap-filling, tiered-confidence duplicate detection, and FHIR R4.0.1 resource generation. Most NPPES consumers eventually write their own cleanup pipeline — this dataset replaces that pipeline.

How often is the data updated? Weekly, automatically. Our pipeline monitors CMS for each new NPPES release and loads the updated, validated, and enriched data within days of publication — no manual intervention needed. Coverage spans every active and deactivated NPI in the registry (9M+ providers).

What FHIR version is supported? FHIR R4.0.1. Every row in PROVIDERS_FHIR contains a US Core-aligned Practitioner or Organization resource, a PractitionerRole for Medicare enrollment (NULL when not PECOS-enrolled), and a Location resource. The resources use standard http://hl7.org/fhir/sid/us-npi identifier systems and can be dropped directly into a QHIN, EHR, or HIE pipeline.

Can I use the Snowflake GEOGRAPHY column for spatial queries? Yes. PROVIDERS_ADDRESSES.geo_point is a native Snowflake GEOGRAPHY type built from the geocoded lat/long. Use it with ST_DISTANCE, ST_DWITHIN, ST_CONTAINS, and other Snowflake spatial functions for proximity, radius, and polygon containment queries without converting from lat/long on every query.

How do I query "unique providers" without double-counting multi-NPI entities? Use PROVIDERS_ENTITIES. It pre-collapses multi-NPI duplicate clusters into one row per real-world entity while preserving any_oig_excluded and any_medicare_enrolled rollups so compliance signals aren't silently deduped away. Alternatively, filter PROVIDERS with is_canonical = TRUE for a one-row-per-entity view that keeps the full column set.

Are inactive or deactivated NPIs included? Yes. NPPES never recycles an NPI and historical claims still reference deactivated NPIs, so the dataset preserves them. Use is_active = TRUE and is_canonical = TRUE for a clean active-provider roster, or query without those filters for historical analysis.

What's the pricing and is there a trial? Full access is $800/month on Snowflake Marketplace. A 14-day free trial includes 1,500 NPIs that are consistent across every view — the trial set is pre-materialized for fast evaluation queries. Metadata views (PRODUCT_INFO, DATA_DICTIONARY) are always fully available.

Get Started

Ready to access the NPPES Provider Golden Record?

Questions? Contact our team for a walkthrough.

Platform	Action
Snowflake	Get on Snowflake Marketplace →
Free Trial	Start 14-Day Free Trial →

About Dataplex

Dataplex Consulting & Data Products delivers turnkey, analytics-ready data products that make complex public and commercial data easy to use across modern data platforms. Our data pipelines include automated quality checks and active monitoring to ensure timely, reliable, and well-structured data that is ready for downstream analytics, machine learning, and operational use.

In addition to data products, Dataplex provides data engineering and analytics consulting services to organizations of all sizes. We bring deep, hands-on experience supporting both early-stage companies and large enterprises, helping teams build scalable data platforms, improve data reliability, and become more data-driven.