Provenance model · every value should explain where it came from

RxAtlas is built for teams that need to inspect source context before they automate around a record. The API, MCP tools, and snapshots carry source identifiers, source URLs, and field-level notes. Sources move and retire pages, so exactly what each claim lets you verify is laid out in what you can verify.

Provenance is evidence context, not a warranty. Source terms, update cadence, and coverage limits still apply; review /docs/sources for source and licence notes, and /docs/data-dictionary for the full field catalog.

What provenance means here

Each exported record under /v1/drugs can combine values from multiple upstream records. Provenance records keep those values auditable by naming the field, source, source class, source record URL, and whether the value is currently effective.

What you can verify

Public claims come in three tiers, and each tier says exactly what it lets you check. Nothing on this page asks you to take a link on faith.

1. Source ownership — always verifiable

Every provenance record names the source and source class behind the value. That attribution is part of the delivered record itself and does not depend on any external page staying up.

2. Release receipts — always verifiable

The release version and snapshot fingerprint tie everything you receive to one exact, immutable delivery. If a question comes up later, the receipt identifies the data you evaluated, whatever the live web does.

3. Resolvable evidence routes — checked every release

Provenance also keeps each record URL exactly as the source published it. Public sources move and retire pages, so rather than implying every deep link resolves, a bounded sample is probed each release and the result is stated below. A field whose sampled link does not resolve is displayed as UNVERIFIED rather than as resolved; resolving ones are displayed as resolved only while the check says so.

Link check 2026-09-23: 40 of 48 sampled evidence links resolved. Treat deep links as best-effort pointers; the source, source class, and release receipt are the verifiable evidence. 4 advertised fields are displayed as UNVERIFIED because a sampled link behind them did not resolve: nadac_per_unit, pricing_unit, nadac_as_of_date, nadac_effective_date.

source_record_url / field-level provenance shapejson

{
  "field": "field_name",
  "source": "cms-sdud",
  "source_class": "government | registry | manufacturer | aggregator | derived",
  "source_record_url": "https://data.medicaid.gov/dataset/158a1baa-5506-400a-8ec3-97756f0b0536",
  "effective": true,
  "superseded_at": null
}

How to inspect it

Start with a narrow API call, inspect the nested provenance payload, then compare source codes against the source registry. Snapshots preserve the same model in export-friendly tables and ship licence notes alongside the data files.

provenance inspectionbash

curl -s "https://api.rxatlas.dev/v1/drugs?ndc=75834-258&limit=5" \
  -H "Authorization: Bearer $RXATLAS_API_KEY"

curl -s "https://api.rxatlas.dev/v1/sources" \
  -H "Authorization: Bearer $RXATLAS_API_KEY"

The manifest example Resolve a product NDC starts at /v1/drugs?ndc=75834-258&limit=5: Resolve an FDA product NDC through authenticated production access.

Fields with source context

These field catalog entries are the fields most likely to carry provenance notes or audit-specific scope. The complete list is in the data dictionary.

idRecord identity

RxAtlas canonical drug-product id.

Provenance: Generated by the canonical product resolver and stable across snapshot exports.

primary_sourceRecord identity

Source code for the first source that created the canonical record.

Provenance: Set from the source registry code that first seeded the canonical product.

ndc_listRecord identity

Distinct NDCs observed across source records for this canonical product.

Provenance: Aggregated from openFDA NDC package/product identifiers attached to the canonical product.

RESOLVEDevidence link: 8 sampled openfda-ndc links resolved at the 2026-09-23 check

source_idRecord identity

First-seen source identifier used to seed the canonical record, such as an NDC, RxCUI, SPL set id, or application number.

Provenance: Copied from the source identifier used during initial canonicalization.

source_record_urlRecord identity

URL for the latest source record that contributed to the canonical row.

Provenance: Taken from the source connector or source registry homepage when record URLs are not per-row.

rxcuiClassification

RxNorm concept identifier when the product has an unambiguous redistributable mapped RxCUI.

Provenance: Limited to redistributable RXNORM/MTHSPL mappings.

RESOLVEDevidence link: 11 sampled openfda-ndc/rxnorm links resolved at the 2026-09-23 check

ingredient_identifiersProduct form

UNII ingredient associations with openFDA UNII substance-name synonyms and independent NDC/UNII source versions.

Provenance: The product-to-UNII relation comes from the openFDA NDC bulk row; names come from the separately versioned openFDA UNII bulk feed.

_provenanceProvenance

Per-field active provenance entries with source, source URL, confidence, and fetched timestamp. Specialized UNII and shortage sidecars carry their own source/version fields.

_sourcesProvenance

All resale-cleared sources actively contributing either ordinary field provenance or a specialized sidecar to this product.

Source catalog

Source descriptions explain coverage and redistribution posture before a buyer opens the full source table.

CMS SDUD

Commercial-use cleared

State and national aggregate Medicaid utilization facts at utilization-type, exact NDC11, year, and quarter grain; suppressed measures remain null and reimbursement is not represented as price or coverage.

DailyMed

Commercial-use cleared

Structured Product Label titles, versions, and publication dates joined to the SPL set identifiers carried by FDA product records.

Drugs@FDA

Commercial-use cleared

FDA application, product-number, submission, and document-metadata records delivered as separate auxiliary tables without claiming product-number-to-NDC applicability.

FDA IID

Commercial-use cleared

Quarterly FDA inactive-ingredient potency and maximum-daily-exposure facts delivered at ingredient/route/dosage-form grain; CAS Registry Numbers are excluded and product applicability is not asserted.

Reviewer caveats

Review these limitations alongside the field and source notes before production use.

RxAtlas is reference data, not medical advice, prescribing advice or reimbursement advice.
Buyers remain responsible for validating source terms and regulated use in their own workflow.

Related evaluation guides

Connect this diligence evidence to the relevant integration decision.

Drug Product Data API