A GUDID device record answers a narrow question: which device does this Device Identifier point to, and what identification attributes did the labeler submit for it? The FDA describes GUDID as a reference catalog for devices with a unique device identifier, holding only the DI portion of the UDI as the key to device information, with flags indicating which production identifiers appear in the UDI [6]. Reading a record correctly means treating it as labeler-supplied identification data subject to business rules, not as an independent verification of performance, clearance status or clinical suitability. The practical skill is separating fields that anchor identity from fields that merely describe packaging or labeling context.

Start with the record's structure and status, not the description

Every DI record in the GUDID web interface is organized into three expandable sections: Device Information, Device Status and Device Characteristics [4]. That structure is the fastest orientation for a reviewer opening an unfamiliar record, because it groups identity attributes separately from commercial distribution status and from physical characteristics such as single-use or MRI safety designations. Before reading any individual value, confirm which section you are in; a field that looks like a product claim is often a labeling attribute recorded for identification purposes.

Status fields deserve equal attention. A Published DI record has been submitted to GUDID, and after the grace period completes, device information becomes available for search and retrieval by the public; during the grace period the record may be edited, and once the grace period completes the record has limited editing [4]. A record you retrieve publicly may therefore reflect a submission that has already moved past its most editable phase. Treat publication status as a timing signal about data maturity, not as a quality endorsement.

Distinguish the DI from the production identifiers it flags

The single most common misreading is expecting a GUDID record to identify a specific manufactured unit. GUDID contains only the DI, which is the primary key for looking up device information, while the UDI itself is the DI plus the PI [3]. The database does not include production identifiers, but it does carry PI flags showing which PI attributes are in the UDI [6]. A record stating that lot or expiration date is present tells you the label carries that information; it does not tell you the lot or date of the unit in your hand.

This distinction shapes how you use the record in practice. If your question concerns a specific shipment, the GUDID record establishes the device family and the fact that lot or serial tracking exists, and the physical label supplies the instance-level values. The AccessGUDID Device Lookup API makes the same separation visible: supplying a full UDI returns parsed PI information such as lot number and expiration date in response headers, while a DI-only lookup returns the device record [5]. Choosing the right query parameter is therefore a substantive decision, not a formatting preference.

Practical detail: reading the PI flags correctly

When a record shows that lot, serial number, expiration date or manufacturing date is present, read that as a statement about the UDI structure on the label, not as a value for any unit. The FDA states that GUDID does not include production identifiers but contains PI flags indicating which PI attributes are in the UDI [6]. If you need instance-level values, the AccessGUDID Device Lookup API returns parsed PI fields such as lot number, serial number and normalized expiration date in response headers when a full UDI is supplied, and those headers appear only for a UDI query [5]. A DI-only lookup will not produce them.

Read optional fields as signals of labeler practice

Not every field carries the same evidentiary weight. The AHRMM best practices guide notes that GUDID contains both required and optional data elements, populated and updated by the device labeler [7]. Device Description is listed there as optional, yet the same guide recommends manufacturers always populate it because it is a critical field for providers distinguishing similar devices and for patient documentation in electronic health records [7]. A sparse description is a documented possibility rather than an anomaly, and its absence limits what you can conclude from the record alone.

Version or model number presents a parallel case. The AHRMM guide records that where a device lacks a version or model number, the manufacturer catalog number should be populated in both the catalog and version/model fields [7]. When you see the same string in both positions, that pattern may reflect the documented convention rather than a data error. Reading optional and conditionally populated fields against published conventions prevents false conclusions about labeler carelessness, and it keeps your review focused on what the record can actually support.

Account for timing, versioning and the public view

Public GUDID data is not a live mirror of the submission system. Device identification information is made available through AccessGUDID and OpenFDA, and what appears there is the most recent data submitted that has completed the grace period after initial publication [2]. The AHRMM guide puts a related figure on the refresh: GUDID takes seven days to update manufacturer data [7]. Neither statement promises real-time synchronization, so a discrepancy between a labeler's internal record and the public view may be a timing artifact rather than a data conflict.

Version metadata helps you reason about change over time. AccessGUDID lookup responses include fields such as public version status, public version number, public version date and device publish date alongside the record status [5]. A record showing an update status with a later version date tells you the labeler revised the submission; it does not tell you what changed or why. For any review where the revision history matters, the documented route is the labeler's own GUDID account view, where edits to unpublished and published DI records are logged with edit date, edit time and the name of the editing user [3].

Know what the record cannot establish

A GUDID record is identification data, and the surrounding documentation is explicit about its limits. The GUDID User Manual states that FDA technical documents, including that manual, do not establish legally enforceable responsibilities, and that "should" means something is suggested or recommended but not required [2]. That sentence matters when a record field looks incomplete: a recommendation in supporting documentation is not a submission requirement, and its absence is not by itself evidence of noncompliance. Separately, the GUDID guidance notes that certain attributes are not released to the public, citing examples such as company physical address and FDA listing number, with the full list in the Data Elements Reference Table [3].

Known data-quality concerns also bear on interpretation. The AHRMM guide records that hospitals, software providers and consultants want to rely on GUDID identification data for standardizing device data, but that known data quality issues within the database impede integration into third-party applications, ERP and EHR systems [7]. That is a statement about integration experience, not a measured error rate, and it does not establish how widespread any specific defect is. A defensible reading treats each field as labeler-asserted, checks it against the physical label where the guidance calls for consistency, and records what remains unverified rather than filling gaps by inference.

A worked review sequence

As a suggested workflow rather than a tested procedure, consider a reviewer assessing an unfamiliar record for an implantable device. Begin with the three-section layout to locate identity attributes, then confirm record status and publication timing before relying on any value [4]. Next, check the PI flags to learn whether lot, serial, expiration or manufacturing date appear in the UDI, remembering that GUDID carries flags rather than the PI values themselves [6]. Then read the description and model fields against the conventions for optional and substituted fields, and note where the record is silent [7].

Close by documenting the retrieval date and the version metadata returned with the record, since the public view reflects data that has completed the grace period and may lag the labeler's submission [2][5]. Where a field is missing or ambiguous, the appropriate next step is a query to the labeler rather than an inference from adjacent fields. This sequence produces a review that states what the record supports, what it leaves open, and which questions require the labeler's own account history to resolve.