openFDA Device Registration & Listing API Data

It is the regulatory census of who participates in the US medical device market, maintained because the law requires it rather than because anyone volunteered. Establishments that produce, prepare, propagate, compound or process devices brought to market in the United States must register with the FDA annually, and most must additionally list the devices they commercially distribute. The result is a structured answer to a question companies otherwise spend weeks on: who makes what, where, under whose ownership, and in what role.

Each record is a facility-to-product pairing. The facility side carries the registration number, FEI number, operating status, expiry year, proprietary and dba names, full address with ISO country code, an owner-operator block, and - for foreign establishments - a designated US agent. An establishment_type field separates the roles: manufacture, contract manufacture for another party, contract sterilization, repack or relabel, reprocessing of single-use devices, initial importing, foreign private-label distribution.

The product side lists each device under its three-letter product code from 21 CFR Parts 862-892, plus brand names and model numbers, premarket references where applicable, and an annotation layer that resolves each code into device name, risk class and medical specialty. At research time the dataset held 333,804 records, which makes it one of the largest machine-readable maps of medical device production capacity anywhere.

One real record, a Canadian contract manufacturer of burn dressings, flattened from its nested form into the shape a delivery arrives in:

registration.registration_number=9680091   fei_number=3002806830
registration.name=DERMA SCIENCES CANADA, INC (dba DERMA SCIENCES, INC.)
city=TORONTO Ontario   iso_country_code=CA
status_code=1 (Active)   reg_expiry_date_year=2026
establishment_type=Manufacture Medical Device for Another Party (Contract Manufacturer)

proprietary_name=Burnfree 4 Oz and 2 Oz
products.product_code=MGQ   products.openfda.device_class=U

Read it as one facility-product pair: a Toronto site, owned through an owner-operator chain, manufacturing wound and burn hydrogel dressings for another party, classified out of product code MGQ. Multiply that pattern across 333,804 records and you can pivot the whole device industry by country, role, specialty and risk class without leaving SQL.

Twenty documented fields split across two layers: the facility layer (registration, FEI, status, address, ownership, US agent) that locates every record in corporate and geographic space, and the product layer (brand names, product codes, premarket references, risk class) that carries the device. The dictionary below is the verified core; deeper columns confirmed during sample preparation fold out on request rather than being promised blind.

Geography: worldwide by construction - the registration obligation reaches any establishment, American or foreign, that markets devices in the United States. Foreign facilities appear with their home-country ISO codes and named US agents, so the dataset doubles as a directory of offshore manufacturing capacity serving the US market.

Temporal: documented coverage runs from 2007 to the present. Registrations live on an annual cycle - each carries an expiry year ending December 31 - so the dataset captures both the current active universe and the trail of lapsed registrations, which is longitudinal signal most people throw away when they deduplicate too eagerly.

Granularity: one record per registered establishment with nested arrays per listed device product. A contract sterilizer listing forty device categories yields one facility record with forty product entries, so counts of establishments and counts of listings are different numbers serving different questions.

Set against the wider Datadory catalog - 1,744 datasets, average quality score 7.81 - this slice scores 9/10: authoritative scope, stable schema, and a join key (FEI number) that connects it to the rest of the FDA device family.

API, files, or your warehouse. Daily, weekly, or hourly.

Pick the channel your team already works in: live lookups for point queries against a specific facility or product code, flat files sized for overnight warehouse loads, or a direct pipe into Snowflake, BigQuery or Redshift. Cadence is yours to set - and to change when your sourcing reviews change.

Every delivery ships with the full field dictionary, sample rows for validation, and a schema that holds steady between refreshes.

• Supply-chain and sourcing teams treat the registry as supplier-discovery infrastructure: pull every active contract manufacturer or contract sterilizer in a product-code family and a country, then work the list down to a qualified-alternates shortlist before a disruption forces the question.
• Competitive intelligence and market-sizing teams count establishments by role, specialty and geography. Facility counts per product code are a defensible proxy for how crowded a device category really is, and the owner-operator block reveals which firms hold several facilities.
• Regulatory and quality affairs teams monitor their own supplier base: registration status, expiry year, and whether a listed device carries premarket references, all queryable across hundreds of suppliers at once instead of spreadsheet by spreadsheet.
• Sales and growth teams build territory plans from facility density by region and establishment type - a contract-reprocessing pitch lands very differently than a hospital group-purchasing one, and the data tells you which doors exist where.

More on how these teams apply the slice lives on the data scientists, developers & builders and journalists & academics pages.

Three things, all solved by taking it through Datadory instead.

Registrations lapse annually, so the active universe moves. Total record count and currently-active establishment count diverge constantly; naive counts double-count sites whose registrations expired mid-cycle. We deduplicate on FEI number and timestamp the snapshot so a “number of active manufacturers” figure survives audit.

Listing dates are flagged unreliable by the agency itself. The creation date on a product listing cannot be trusted as a definitive launch date, so we treat it as indicative only and say so wherever it surfaces in analysis.

Nesting and annotation gaps need flattening decisions. Products arrive as arrays inside each facility record, and the enriched device-name/class/specialty join is present on some entries and absent on others. We flatten to one row per facility-product pair, preserve the raw nesting alongside, and fold the fields we cannot verify into the additional-fields-on-request tier rather than promising them blind.