openFDA Unique Device Identifier (GUDID) API

Virtually every model of medical device marketed in the United States, described in one consistent frame. Under the FDA's Unique Device Identification rule, labelers must submit key device identification for each model carrying a UDI, and those submissions accumulate in the Global Unique Device Identification Database - GUDID. openFDA, the agency's machine-readable layer over its own holdings, exposes the corpus, and at the August 2026 research pass it reported 5,083,948 device identifier records.

A UDI decomposes into two parts, and knowing the split is most of understanding the data. The device identifier (DI) is the fixed, model-specific catalog number - the part GUDID stores in depth. The production identifiers - lot number, serial number, expiration and manufacture dates - vary unit by unit, so GUDID instead records which of them a model's UDI carries, as explicit per-record flags.

Inside Datadory's health-care-supplies slice this is the master file the other records hang off: clearances say who entered, recalls say what went wrong, and this database says what is actually on the shelf, model by model, flag by flag. Get a sample of this dataset cut to your categories.

Three real device master rows captured during verification, laid out flat exactly as a delivery lands:

# openFDA Unique Device Identifier (GUDID) API - device master rows captured during research

# examination glove, in active distribution
brand_name        : MICROFLEX
version_or_model  : XC-310-XL
company_name      : Ansell Healthcare Product
device_description: MICROFLEX XCEED XC-310 SIZE XL, Nitrile Powder-Free Examination Glove
primary_di        : 00769799310140
is_rx             : false
commercial_distribution_status : In Commercial Distribution

# spinal implant, classified and flagged
brand_name        : Fortilink-L IBF System
version_or_model  : 65-L-2660-13-12L
company_name      : Pioneer Surgical Technology, Inc.
device_description: Interbody Fusion Device
product_codes     : ODP - Intervertebral fusion device with bone graft, cervical (Class 2)
                    MAX - Intervertebral fusion device with bone graft, lumbar (Class 2)
gmdn_term         : 60762 Polymeric spinal interbody fusion cage
implantable       : true     is_single_use : true     mri_safety : MR Conditional

# orthopaedic instrument tip, tied to its clearance
brand_name        : AL Disc Prep
version_or_model  : 106.7083
company_name      : CTL Medical Corporation
device_description: Removable Wedge Distractor Tip, H13mm
primary_di        : 00841732166448
premarket_submissions : K192863
gmdn_term         : 44759 General internal orthopaedic fixation system implantation kit
is_sterile        : false    sterilization_methods : Moist Heat or Steam Sterilization

Read the three together and the database's personality shows. The glove row is pure current-state commerce: a live DI, an OTC-flagged consumable sitting in commercial distribution today. The fusion-cage row shows classification done by someone else - two product codes resolving to Class 2 regulation numbers, a GMDN term with a full definition attached, and the three flags a spinal registry cares about (implantable, single-use, MR Conditional) stated outright. The distractor-tip row demonstrates the paper trail: a primary DI tied back to clearance K192863, plus a sterile:false reading paired with the sterilization method the labeler expects a processor to apply. Multiply these shapes across five million rows and you have a queryable US device shelf.

Thirty-nine documented fields define a record, grouped four ways: identity (brand, catalog number, model number, company, DUNS, the identifiers block), classification (GMDN terms and enriched FDA product codes), state (distribution status and end date, MRI safety statement, sterilization attributes, sizes, storage) and labeling flags - roughly twenty booleans covering prescription versus over-the-counter, single use, kits, combination products, latex labeling and which production identifiers the UDI carries.

Definitions below were verified against the published field reference during the August 2026 research pass, a standard met by 85.7% of the 1,744 datasets Datadory catalogs.

• Geography: devices distributed in the United States, with labelers both domestic and foreign - which makes the corpus double as a directory of offshore manufacturing serving the US market, keyed by company name and DUNS number.
• Temporal: filings begin with the 2013 UDI rule phase-in and run to the present. Revisions are first-class citizens: every record carries a public version number, date and status trio, so a model's labeling history is a small time series rather than a single overwritten row.
• Granularity: one record per device identifier - the model version - which is the level procurement catalogs and safety registries actually join at. Counts of models and counts of packages are different numbers here by construction.

Scale check: 5,083,948 records makes this the largest table in the health-care-supplies slice. On Datadory's 0-10 rubric the record scores 9, against a 7.81 average across all 1,744 cataloged datasets - held up by verified field definitions and shipped sample rows.

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

You choose the channel and the cadence; the thirty-nine-field dictionary above travels unchanged across all three. Name the brands, GMDN terms, product codes or distribution states you care about and the slice lands shaped to that scope - nested identifier, nomenclature and submission blocks flattened into columns your warehouse loads without a parsing project. Teams running catalog synchronization take scheduled refreshes so new model versions become observable history; teams answering a one-time sourcing question take a single shaped extract. Either way the keys stay identical, so each pull diffs against the last instead of starting over.

• Procurement and catalog synchronization - join the primary DI onto internal SKUs so purchasing systems carry standardized brand, model, GMDN and sterilization attributes instead of free-typed lookalikes.
• Supply-chain continuity screening - commercial_distribution_status and its end date separate the live shelf from quietly discontinued models before a disruption finds out for you.
• Competitive shelf monitoring - watch which rivals register new model versions, and which categories they stop filing for - movement visible months before press releases confirm it.
• Category sizing by nomenclature - count distinct models per GMDN term and per product code as an assortment-density proxy that survives audit better than vendor deck arithmetic.
• Identity resolution for safety work - bridge adverse-event and recall records onto model-level attributes so harm reports resolve to the exact configurations involved.
• Feature engineering - twenty labeled booleans per model give classifiers clean categorical inputs nobody had to hand-label.

Developers and data-product builders get the strongest fit: identifier-keyed lookups over a five-million-row corpus, stable enough to wire straight into sourcing and compliance tools - see developers builders use cases. Sales and growth teams segment prospects by labeler portfolio breadth and implantable flags before outreach (competitive intel product teams read the same columns for rival moves). Market researchers and consultants count models per GMDN term as assortment-density evidence for category sizing. Data scientists and ML engineers resolve device identities across sources on DI keys and GMDN terms (data scientists use cases). Investors and quant researchers read SKU breadth per labeler as a coarse diligence signal on device makers' actual market presence.

Within health care supplies, this record owns current-state product identity - what is on the market right now, model by model. The neighbors own different jobs. openFDA Device 510(k) Clearances documents all 175,814 premarket notifications since 1976, the who-entered-and-on-what-predicate ledger; its premarket submission numbers are exactly what this database's submission block points back to. openFDA Device Recalls captures 58,999 corrective actions since November 2002, the consequence layer. Across the slice boundary, openFDA Device Adverse Events (MAUDE) holds 25.7 million harm reports, while ClinicalTrials.gov API v2 - Studies Database tracks the evidence pipeline; the direct pairing sits in our dedicated comparison.

The standard stack runs in lifecycle order: clearances establish the device, trials establish the evidence, this master file establishes the shelf, recalls and adverse events establish what went wrong afterward. Only this leg is measured in millions of model-level rows.

Because the raw holdings are built for regulatory filing, not analysis, and the gap shows up in three places. Nested structures - identifiers, GMDN terms, product codes, submissions - arrive as arrays inside each record, so we flatten them into joinable columns while preserving the original nesting alongside. Labeler revisions accumulate silently upstream, so we keep successive captures, turning a mutable database into observable history you can diff. And the nomenclature itself moves: FDA announced in December 2024 that its Preferred Term (PT) codes will be phased out of GUDID in favor of GMDN terms, so deliveries document which coding each record set carries rather than letting the transition surface as a silent schema break.

Start with a sample scoped to your brands, GMDN terms or product codes; move to recurring delivery when the joins prove out. Browse the rest of the industry on the health care supplies data hub, the best health-care-supplies datasets ranking, or the full catalog. The GUDID glossary entry and unique device identifier primer unpack the terminology underneath.