The openFDA device API uses the search parameter as the primary filter mechanism, following a consistent syntax of search=field:term appended to a device endpoint. Queries always begin with a base endpoint such as api.fda.gov/device/registrationlisting.json or api.fda.gov/device/pma.json, and the search parameter narrows results to records matching a specific field value [1][4]. Without a specified field, the API searches every field, which can return broad and less predictable results [2]. The API returns one matching record by default unless a limit is specified, and the maximum allowed limit for a single call is 1000 records [1][2]. Regulatory affairs teams should therefore plan filter combinations carefully rather than relying on a single broad query for complete datasets.

Search syntax and the core query parameters

The search parameter is the basic building block of openFDA queries, and every search follows a special syntax: search=field:term [1][2]. For example, a search for records with HQY in the products.product_code field is written as search=products.product_code:HQY [1]. The API documentation lists five query parameters: search selects matching field values, sort arranges results by a chosen field in ascending or descending order, count returns the number of unique values for a field across matched records, limit controls how many records are returned, and skip moves past a specified number of matching records before returning the next set [2]. Understanding these five parameters is essential for constructing repeatable queries.

When a field is omitted from the search parameter, the API looks for the term across all fields, which may create unintended matches and larger result sets than expected [2]. The API returns only one matching record for a search unless the limit parameter is used, and the largest allowed limit is 1000 records per call [1][4]. The skip parameter can be combined with limit to paginate results, but the largest allowed skip value is 25000 [2]. For result sets larger than that, openFDA documentation points to a separate Paging approach rather than continuing with the standard skip parameter [2]. Regulatory analysts should treat the 1000-record limit as a hard boundary that shapes how they divide investigations into smaller, verifiable queries.

Selecting the right device endpoint and fields

Device data is spread across multiple endpoints, including registration and listing, premarket approval, classification, and unique device identifier records [1][3][4][7]. Each endpoint has its own set of searchable fields, so a filter that works on one device endpoint may not exist on another [2]. The registration and listing endpoint contains information about establishment locations and the devices manufactured at those establishments, while the premarket approval endpoint supports searches such as decision_code:APPR for approved decisions [4][6]. Analysts must first identify which endpoint contains the needed data before selecting filters, because a term applied to the wrong endpoint will not produce meaningful regulatory intelligence.

The device classification endpoint lists about fifteen searchable fields, including device_class, device_name, product_code, regulation_number, review_code, medical_specialty, implant_flag, and life_sustain_support_flag [3]. These fields support queries that go beyond simple text matches, such as filtering by classification string or by a flag indicating whether a device is implantable [3]. The unique device identifier endpoint provides example queries for record-level filters, including public_device_record_key and public_version_status:Update [7]. Because the openFDA documentation organizes fields by endpoint, downloading the fields reference document for a specific dataset is a practical first step before drafting a filter strategy [3].

A recommended query construction sequence

Start with the base endpoint for the dataset you need, such as api.fda.gov/device/registrationlisting.json, then append search=field:term before adding limit, sort, count, or skip parameters as needed [1][2]. Always include limit=1 or a small limit such as limit=5 on the first exploratory call so you can confirm field names and term formats before running a larger request [1]. When checking for record completeness, use an existence filter such as _exists_:public_device_record_key on the appropriate endpoint [7]. If you need to page through more than 1000 records, use the 1000-record limit with skip or the documented paging method rather than attempting a single oversized request [2].

Field matching, existence checks, and harmonized fields

Filters can express more than exact equality. Example queries in the device UDI endpoint demonstrate an existence check using the syntax _exists_:public_device_record_key, which returns records that include a specific field [7]. This kind of filter is useful when the goal is to identify completeness of structured data rather than to find a single product. A second example uses public_version_status:Update to match records where the public version status is the exact term Update [7]. These two syntaxes, exact term matching and existence matching, give analysts different tools depending on whether they need a content filter or a data-coverage filter.

Harmonized openFDA fields add another layer to search filtering. Depending on the endpoint, searches can use either native record fields or harmonized openFDA fields added as annotations in a special openfda dictionary [5]. For device categories, the documentation shows harmonization across regulation_number, fei_number, device_name, device_class, medical_specialty_description, registration_number, k_number, and pma_number, though not all datasets use all of these fields [5]. Harmonization is not universal because it requires an exact match, so records with misspelled names or no matching identifier may lack openfda sections entirely [5]. As a result, searching a harmonized field will only return a subset of records from a given endpoint, and analysts should not assume that a harmonized filter covers the full dataset [5].

Step-by-step example: evaluating a registered product code

A regulatory affairs analyst trying to confirm which establishments registered a specific device could begin with the registration and listing endpoint. As a suggested workflow, the analyst first reviews the searchable field list for that endpoint, then starts with a narrow product code query, such as search=products.product_code:HQY, and adds limit=5 to inspect the first five matching records [1]. This first result set can reveal whether the field name and term match the analyst's expectation before a larger retrieval is attempted. The analyst can then add count or sort parameters if the next step is to understand how many unique values exist in a field across matched records [2].

If the initial query returns results that include harmonized fields, the analyst can cross-check whether the same product appears under other device identifiers, but only if those fields exist for that specific record [5]. The analyst should account for the fact that harmonized searches return a subset of records, so a missing harmonized field value does not mean a device is absent from the source data [5]. Before drawing conclusions, the analyst would also need to account for the registration and listing dataset's stated time period, which for this source runs from 2007 through a date in November 2026 and is updated monthly [6]. A query result reflects the point-in-time snapshot of that dataset and should not be treated as live FDA publication data.

Known limits and what a clean query cannot prove

The openFDA platform states that not all data has been validated for clinical or production use and that openFDA serves publicly available data without personally identifiable patient information [8]. The API returns JSON with two sections: meta contains metadata such as a disclaimer, licensing information, last-updated date, and total matching records, while results contains the matched records themselves [8]. Reading meta after a query is a practical way to verify that the total matching record count aligns with what a return of results actually shows. A large result count can signal that a filter is too broad or that pagination will be needed.

Every material statement drawn from openFDA device data must account for the possibility that the API may be rate-limited or restricted according to the Terms of Service [1]. The documentation also states that the API should not be relied on for medical care decisions [1][8]. For a B2B regulatory research workflow, this means openFDA device filters support dataset discovery, coverage checks, and preliminary matching, but they do not establish clinical appropriateness, legal compliance, or a device's current market status. Any downstream conclusion that depends on accuracy, completeness, or regulatory status should be validated against the underlying FDA source documents rather than the API response alone.