openFDA: Medical Device Datasets, API Endpoints, and Research Practice

openFDA is an open-access, Elasticsearch-powered data platform and Application Programming Interface (API) operated by the United States Food and Drug Administration (FDA) that serves structured, machine-readable public health records. For investigators and regulatory specialists in biomedical engineering, openFDA indexes premarket regulatory submissions, device classifications, post-market surveillance reports, and recall actions in developer-friendly JSON format.

Medical Safety and Regulatory Disclaimer: openFDA data is provided for research, regulatory informatics, and public informational purposes. The FDA explicitly includes a disclaimer across all endpoints stating that openFDA results must not be relied upon to make decisions regarding clinical medical care and should be assumed unvalidated for direct medical diagnosis.

What It Is

Implemented under the White House 2013 Executive Order “Making Open and Machine Readable the New Default for Government Information” and maintained under the 2019 “Open, Public, Electronic, and Necessary (OPEN) Government Data Act,” openFDA translates complex, legacy agency databases into standardized REST APIs. The platform enables academic investigators, software developers, and research administrators to query multi-decade agency archives without writing bespoke web scrapers or filing Freedom of Information Act (FOIA) requests.

Within the medical device lifecycle, openFDA acts as a high-throughput gateway for post-market surveillance, competitive landscape mapping, and predicate research. Whether examining whether a prototype qualifies under the statutory definition of what is a medical device, verifying product codes, or analyzing clearance trends, openFDA provides programmatic access to records from the Center for Devices and Radiological Health (CDRH).

Who Runs It and What It Covers

openFDA was launched in June 2014 by the FDA Office of Informatics and Technology Innovation (initiated by Dr. Taha Kass-Hout with Presidential Innovation Fellows in July 2013). The infrastructure is hosted in a public cloud environment and operates as a distributed Elasticsearch cluster, providing both live query endpoints and bulk downloadable datasets packaged as compressed JSON files (.json.zip).

The platform partitions agency data into major product nouns: drugs, devices, and foods. In biomedical engineering research, the openfda device endpoints cover several core regulatory and post-market tracking datasets:

• 510(k) Premarket Notifications (/device/510k.json): Sourced from CDRH clearance records, including applicant names, cleared dates, device trade names, 510(k) numbers (K-numbers), and statement/summary indicators. This endpoint complements modern structured submission pathways such as the eSTAR FDA submission template.
• Premarket Approvals (/device/pma.json): Premarket Approval review decisions for Class III medical devices, containing supplement numbers, advisory committee decisions, approval dates, and regulatory tracking records.
• Medical Device Classification (/device/classification.json): Product codes (three-letter codes such as FPA or DXZ), medical specialty panels, device classes (Class I, Class II, or Class III), and Code of Federal Regulations (CFR) regulation numbers.
• Medical Device Reports / Adverse Events (/device/event.json): De-identified post-market safety records from the Manufacturer and User Facility Device Experience (MAUDE) database. These records capture suspected device malfunctions, injuries, and deaths submitted under mandatory Safe Medical Devices Act reporting duties.
• Recall Enforcement Reports (/device/enforcement.json and /device/recall.json): Product recall actions categorized by Class I, Class II, or Class III recall designations, describing product defects, reason for recall, and distribution volumes.
• Establishment Registration and Listing (/device/registrationlisting.json): Manufacturing facilities registered with the FDA, foreign exporters, and registered proprietary names.
• Unique Device Identifier (/device/udi.json): Public Global Unique Device Identification Database (AccessGUDID) records containing device identification attributes, package models, MRI safety status, and latex content.
• COVID-19 Serology (/device/covid19serology.json): Independent performance validation data for serological test kits evaluated by FDA and NIH partners.

Every response adheres to a strict two-part top-level JSON structure: a meta block (containing the agency disclaimer, license links, last_updated date string, and total matching records) and a results array containing the payload objects.

How to Search or Use the openFDA API

Interacting with the openfda api involves querying HTTPS endpoints with standard URL parameters. The five core query parameters defined in official openFDA documentation govern all interactions:

1. Determine Endpoint and Base URL: Select the noun and dataset corresponding to your research objective. All requests must use secure HTTPS routed to https://api.fda.gov (for example, https://api.fda.gov/device/510k.json).
2. Acquire and Pass an API Key: While openFDA permits unauthenticated queries for casual exploration, researchers conducting automated extraction should register for a free API key. Supply the key as a query parameter (?api_key=YOUR_KEY) or via HTTP Basic Authentication headers (Authorization: Basic base64(YOUR_KEY:)).
3. Construct the Search Filter: Use the search parameter to construct Lucene-style search filters on specific record fields. For example, querying search=device_name:"dialyzer"+AND+decision_date:[2020-01-01+TO+2025-12-31] restricts results to dialyzers cleared within that 5-year window. If no field is specified, openFDA searches across all indexed text fields.
4. Set Record Limits: Specify the number of matching records returned per response using limit. The maximum allowable value for limit is 1,000 records per HTTP request (the default is 1 record if omitted).
5. Paginate Results Using Skip: Use the skip parameter in combination with limit to page through results (for example, limit=100&skip=100 fetches page 2). Note that openFDA limits the maximum skip value to 25,000 records.
6. Execute Aggregate Counts: Use the count parameter to calculate frequency distributions rather than retrieving raw documents. For example, passing count=openfda.device_class.exact on the classification endpoint returns an aggregated summary of devices across regulatory classes. The API returns up to 1,000 unique values.
7. Apply Sort Order: Use the sort parameter with the :asc or :desc modifier (such as sort=decision_date:desc) to order results chronologically or alphabetically.

In addition to device informatics, investigators studying combined product lines or clinical trials can leverage openFDA pharmaceutical datasets alongside the FDA NDC Directory.

Strengths and Limits

Understanding both the analytical power and structural boundaries of openFDA is necessary for sound research practice and regulatory verification.

Key Strengths

• High-Speed Programmatic Access: Bypasses CAPTCHA-protected, form-driven legacy FDA web forms, enabling automated data collection scripts in Python, R, and bash.
• Standardized Harmonization: Normalizes heterogeneous regulatory filings into unified JSON schemas with consistent field naming conventions.
• Open Data Licensing: Public domain data licensed without proprietary royalty fees, facilitating reproducible research across academic and clinical institutions.
• Zero PII Exposure: Built-in privacy scrubbing removes patient names and sensitive identifiers from adverse event filings before indexing.

Technical and Analytical Limitations

• The 25,000 Skip Ceiling: Queries that match more than 25,000 records cannot be fully paginated using the REST API alone. Datasets exceeding this threshold require downloading openFDA bulk archive zip files (.json.zip) or scoping search criteria into narrower date partitions.
• Rate Limitations: Unauthenticated queries are constrained to 240 requests per minute and 1,000 requests per day per IP address. Registered API keys raise the daily ceiling to 120,000 requests per day (with the same 240 requests/minute burst threshold). Exceeding limits returns HTTP 429 Too Many Requests.
• Post-Market Reporting Biases: Adverse events reported in the MAUDE database (/device/event.json) are passive surveillance submissions. Submission of an adverse event report does not confirm that the device caused the adverse outcome, and reporting rates fluctuate based on clinical publicity and regulatory recalls.
• Unvalidated Clinical Status: Data is provided as-is from manufacturer submissions; the FDA explicitly disclaims data accuracy for direct diagnostic or therapeutic decision-making. Verification against engineering design histories and medical device design controls remains mandatory for device developers.

openFDA vs. Direct FDA Portals

Biomedical researchers choose between openFDA and primary agency web portals depending on whether tasks require manual inspection or computational analysis.

Feature	openFDA API	CDRH Legacy Search (510k/PMA/MAUDE)	AccessGUDID Web Portal	Bulk FDA Download Files
Interface Type	RESTful JSON API & Elasticsearch	HTML form search portal	HTML search & simple lookup	Raw flat text, XML, or JSON files
Automated Querying	High (standardized HTTP GET requests)	Low (requires web scraping, session state)	Low (primarily interactive browser lookup)	Batch processing only (no live API queries)
Pagination Limits	Max 1,000 per page; max 25,000 skip	Page-by-page HTML tables	Browser-driven display (500 max)	Complete dataset (no pagination limits)
Data Harmonization	High (standardized field names and dates)	Variable (dependent on legacy database age)	High (standardized UDI specifications)	Raw relational tables (requires custom ETL)
Rate Limits	240 req/min; 120,000 req/day with key	Informal IP throttling / bot blocks	Standard web browsing constraints	One-time bulk download (no rate limits)
Best Research Use	Systematic literature and clearance trends	Single-device regulatory audits	Specific packaging and barcode verification	Full multi-year machine learning pipelines

Frequently Asked Questions

Is openFDA free to use for academic and clinical research?

Yes. openFDA is a free, public service funded by the United States Government. Obtaining an API key requires only an email address, carries no usage fees, and provides an allocation of up to 120,000 requests per day.

How does openFDA handle medical device adverse event reports?

openFDA indexes medical device reports from the CDRH MAUDE system via /device/event.json. The platform automatically strips personally identifiable information (PII) while preserving device brand names, manufacturer details, event types (malfunction, injury, or death), and reporter narratives.

What should researchers do if a query returns more than 25,000 records?

Because openFDA enforces a maximum skip parameter value of 25,000, queries returning larger result sets cannot be retrieved through continuous API pagination. Investigators should either partition the search into chronological sub-windows using date range filters or download the full dataset using openFDA raw download archives.

How should openFDA datasets be cited in scientific manuscripts?

When reporting findings derived from openFDA, authors should document the exact query URL, the endpoint queried, the access date, and the specific last_updated timestamp returned in the API response’s meta block. This ensures reproducibility as datasets receive weekly or monthly updates from the FDA.

Sources

• U.S. Food and Drug Administration. (2026, November 8). About openFDA: History, Architecture, and OPEN Government Data Act Mandate. openFDA Portal. https://open.fda.gov/about/
• U.S. Food and Drug Administration. (2026, November 8). openFDA API Documentation and General Concepts. openFDA Portal. https://open.fda.gov/apis/
• U.S. Food and Drug Administration. (2026, November 8). openFDA Authentication and Standard Rate Limits. openFDA Portal. https://open.fda.gov/apis/authentication/
• U.S. Food and Drug Administration. (2026, November 8). openFDA Query Parameters: Search, Count, Sort, Limit, and Skip. openFDA Portal. https://open.fda.gov/apis/query-parameters/
• U.S. Food and Drug Administration. (2026, November 8). Medical Device API Endpoints: 510(k), PMA, Classification, and MAUDE Event Records. openFDA Portal. https://open.fda.gov/apis/device/