The Safety Database Everyone Cites and Almost No One Reads Correctly
- Aug 20
- 2 min read
Updated: 4 days ago
Ask how safe a surgical robot is, and most answers point to the same source: a report count pulled from a database called MAUDE.
That number is real; what it's usually assumed to mean is not.
What MAUDE Actually Is
MAUDE, the Manufacturer and User Facility Device Experience database, is the FDA's public collection of medical device reports on adverse events.
It holds the last ten years of reports, updated monthly, filed by manufacturers and hospitals when a device caused or may have contributed to a death, serious injury, or malfunction, plus voluntary reports from clinicians and patients.
What the FDA Itself Says You Can't Do With It
The FDA states directly that the incidence, prevalence, or cause of an event cannot be determined from this data alone, because of underreporting, inaccuracies, lack of verification that the device actually caused the event, and no information about how often the device is even used.
A report can contain the word death without being classified as one, it only counts that way if the person filing it believed the device contributed.
Report counts can't be turned into rates, and they can't be fairly compared across two different devices without knowing how many procedures each one has actually performed.
Why This Matters More for AI-Enabled Devices
Very few AI-enabled surgical devices currently have any MAUDE entries at all, and the database has no dedicated field to flag when a failure involved the AI component specifically.
That means even where reports do exist, there's often no way to tell an algorithm-driven error apart from a purely mechanical one.
What To Do With That
Treat any MAUDE search as a floor, not a rate, a lower bound on what's actually happened rather than a complete picture.
A clean search is not the same as a clean safety record; it may simply mean nobody filed a report.
The FDA's own language is the sentence worth remembering: this data isn't meant to evaluate rates of adverse events or to compare how often they occur across different devices.
