DeviceWatch now checks for MAUDE narratives the FDA publishes after the initial report and adds late text through the audit trail. Reports without published text are labelled.
The narrative adds context to an adverse event report: what happened, how the device was being used and what the reporter observed. That context may arrive after the coded fields.
Following the report as text arrives
DeviceWatch reads MAUDE through the FDA’s weekly bulk files and openFDA’s search API. Both ingestion paths now revisit reports with missing narratives.
The weekly import carries stored reports without text into the next join, so a narrative can be matched even when it arrives in a later file. The scheduled openFDA run also checks previously stored events for newly available text, in batches of 50 report keys.
These changes address a gap in how late narratives were collected. A backfill also checked the FDA’s yearly narrative file for text missing from previously stored reports.
Preserving the record
Late narratives are added as audited corrections, with the original row retained. If the FDA sends a report again without text, DeviceWatch preserves the narrative already stored. The late-narrative checks fill empty fields without replacing existing text.
That gives reviewers both the additional context and a record of when it became available in DeviceWatch.
Labelling reports without a narrative
When the FDA has not published a narrative, the event list, dashboard and event page now say “No FDA narrative”. The event PDF keeps the narrative heading and states that no narrative has been published.
The label distinguishes unavailable source text from analysis that has not yet run. It describes the current record without suggesting that a narrative will never arrive.
AI-generated text is also labelled separately from FDA source text. Alongside these changes, daily verification checks each organisation’s audit chain and flags failures.
DeviceWatch monitors MAUDE, recalls and 510(k) clearances for device makers. More at devicewatch.app.