MAUDE Adverse Event Reports for AI Medical Devices — What the Data Shows

The FDA Manufacturer and User Facility Device Experience (MAUDE) database is the post-market signal system for medical devices. When a cleared device malfunctions or is involved in an adverse event, that event is reported to MAUDE. For AI medical devices, MAUDE is an underutilized intelligence source that tells you something FDA's pre-market data can't: how AI devices actually perform in the real world.

What MAUDE Reports Tell You About AI Devices

MAUDE reports for AI devices typically describe one of several categories of failure:

False positive failures — the AI flagged a finding that wasn't there. In high-volume screening AI, false positive rates that seem acceptable in validation studies can generate significant adverse event volumes in real-world use. A sensitivity-optimized AI running on thousands of cases per day will generate a large absolute number of false positives even at high specificity.

False negative failures — the AI missed a finding that was present. These are more clinically serious and more likely to generate MDR reports because they can directly lead to delayed diagnosis or treatment.

Workflow integration failures — the AI output was presented in a way that led to clinical errors, not because the AI was wrong, but because the human-AI interface created confusion about the confidence level, the indication for follow-up, or the nature of the AI's output.

Distribution shift events — the AI performed well on the population it was validated on but degraded when deployed to patients with different demographics, comorbidities, or imaging characteristics than the training set.

Which AI Device Categories Have the Most Reports

Based on AIFDA Intel's analysis of MAUDE reports linked to AI-enabled cleared devices:

Radiology AI — highest absolute volume of adverse event reports, which is expected given that radiology AI has the most cleared devices and the highest deployment volume. Per-device event rates are not necessarily higher than other categories.

Cardiac monitoring AI — notable for reports involving algorithm-triggered alerts that led to unnecessary interventions. The high-stakes, time-sensitive clinical context of cardiac monitoring makes AI failures more likely to generate reportable events.

Autonomous diagnostic AI — the devices with the strictest special controls also generate the most scrutinized adverse event reports. FDA pays close attention to MAUDE reports for autonomous AI.

MAUDE as Competitive Intelligence

MAUDE reports for competitor devices tell you something 510(k) submissions can't: actual real-world performance issues. A competitor device with a high volume of adverse event reports may indicate:

• Real-world performance that doesn't match cleared claims — an opportunity for differentiation
• Distribution shift issues in specific patient populations — populations your device could serve better
• Workflow integration problems — areas where UX and human factors design matter as much as algorithm performance
• Emerging FDA scrutiny — high MAUDE volumes for a device often precede increased regulatory attention

MDR Compliance for AI Device Companies

If you have a cleared AI device, your MDR (Medical Device Report) obligations are the same as any other cleared device — with some AI-specific nuances. You must report device malfunctions that could cause or contribute to serious injury if they were to recur. For AI devices, this includes:

• Algorithm failures that produced clearly erroneous outputs in clinical use
• Performance degradation that is reproducible and not within the device's validated range
• Events where the AI output contributed to a clinical decision that led to patient harm

The challenge for AI companies is that algorithm outputs are often probabilistic and their contribution to clinical outcomes is difficult to isolate. Building clear event investigation and MDR decision procedures before deployment — not after your first event — is essential.

Using MAUDE in Pre-Market Planning

MAUDE data is valuable before you submit, not just after you clear. Before finalizing your validation study design, review MAUDE reports for cleared devices in your category. They'll tell you:

• What failure modes FDA is already seeing in real-world use — and therefore what your submission should address
• What patient populations have generated the most adverse events — and whether your validation dataset adequately represents them
• What performance thresholds have proven insufficient in practice — and whether your target specs need adjustment