Methodology

1. Source acquisition

The pipeline downloads the CMS NPPES Data Dissemination Monthly Version 2 archive from the official NPI Files page and records the release label, release date, download timestamp, byte count, SHA-256 checksum, and archive inventory. This site was built from NPPES Data Dissemination August 2026 V2 (release date August 10, 2026). Only the main provider file is processed; the other-name, practice-location, endpoint, and weekly files are not used.

2. Fields read

From each row the pipeline reads the NPI, entity type, name components or organization legal business name, credential text, enumeration date, last update date, deactivation and reactivation dates, up to fifteen taxonomy codes with their primary switch, and the primary practice location fields (street lines, city, state, ZIP code, country, and telephone). No other columns are read.

3. Eligibility

A record is eligible when all of the following hold:

• the NPI has ten ASCII digits and passes the NPI check-digit rule;
• the entity type is 1 (individual) or 2 (organization);
• the record is active: no deactivation date, or a reactivation date after the last deactivation;
• a display name can be formed from the recorded name fields;
• the practice location is in one of the 50 states or the District of Columbia with a United States country code, a city name without digits, and a five-digit ZIP code;
• at least one taxonomy code is present;
• dates parse as calendar dates.

In NPPES Data Dissemination August 2026 V2, 9,302,551 of the records in the main file were eligible: 7,362,368 individuals and 1,940,183 organizations. Directory counts on state and city pages use this eligible set.

4. Published fields

Profiles publish the display name, NPI, entity type, active status, primary and additional taxonomy codes, practice street, city, state, ZIP code, and practice telephone, the enumeration and last-update dates, and the source release. The same field set applies to individuals and organizations. Mailing addresses, fax numbers, authorized officials, license numbers, endpoints, email addresses, employer identification numbers, and other tax identifiers are never published. Taxonomy codes appear with their official display names, grouping, classification, and specialization from the NUCC Health Care Provider Taxonomy code set version 26.1; definitions are not reproduced.

5. Selection of cities and profiles

Eligible records are grouped by state and normalized city. City groups with fewer than ten eligible records are excluded. Within each state and DC, up to five cities with the most eligible records are selected first; the remaining city slots are filled from the national ranking by eligible count, giving exactly 1,937 city pages.

Profiles are chosen only from selected cities. Each record receives a completeness score that rewards only the presence of publishable facts (name, location, ZIP code, taxonomy code, credential text, street, telephone, parseable dates). Every selected city reserves its five highest-scoring records, alternating entity type where both exist; the remaining slots are allocated across states in proportion to eligible counts and filled with the highest-scoring records, with a maximum of 50 profiles per city. Ties are broken by a stable SHA-256 hash of the release identifier and NPI. The result is exactly 9,000 individual and 9,000 organization profiles, 18,000 in total, plus 51 state and 1,937 city pages and 12 fixed pages: 20,000 pages.

The score does not measure quality, popularity, licensure, patient volume, or affiliation. Being published here, or not, carries no meaning about a provider.

6. Count definitions

"Eligible active NPPES records" counts every eligible record in the full national file for the scope shown (nation, state, or city). "Profiles currently published by this experimental website" counts only the pages on this site. The published set is a selected subset; no page claims to list all providers in an area.

7. Validation before publication

Automated checks confirm that projected rows equal eligible plus rejected rows, that every published document matches the URL manifest, that no prohibited field appears in any output, that every page has one heading, a unique title and self-canonical address, valid structured data, and working internal links, and that every page is reachable from the home page. The build fails if any check fails.

8. Refresh

The site reflects a single monthly snapshot. CMS updates NPPES continuously and publishes new files monthly and weekly; changes appear here only when a later file is processed. For the current record consult the NPPES NPI Registry.