TL;DR
The openFDA device adverse event database holds 25,711,469 reports. 62.1% are malfunctions, 36.4% are injuries, and 0.9% are deaths. Annual reporting grew 85% between 2020 and 2025. The number that matters for products liability is not the headline: it is the 9,346,364 injury reports, and the fact that 65.3% of all reports attach to Class II devices, the class most manufacturers treat as moderate risk.
Data study · pulled 13 August 2026
25.7 million device adverse event reports. Most of the exposure is in Class II.
MAUDE is the FDA database of medical device reports: adverse events submitted by manufacturers, importers, and user facilities. It is public, it is enormous, and it is the closest thing the device industry has to a running record of what actually goes wrong in the field.
Every figure on this page was counted directly from the openFDA device event API on 13 August 2026. Nothing is modelled or sampled. The methodology is at the bottom so any figure can be reproduced.
What is in the database
25,711,469 reports, by event type.
The insurance read
Class II is where the volume is.
Class II devices account for 16,784,939 reports, or 65.3% of the total. Class III, the class everyone treats as the serious one, accounts for 7,762,446, or 30.2%. Class I accounts for 963,288.
That matters because device classification frequently drives how a manufacturer thinks about its own risk. A Class II clearance is often read internally as evidence of moderate exposure, and products liability limits get sized accordingly. The reporting record does not support that reading. The overwhelming majority of what goes wrong in the field, including a large share of injury reports, involves Class II devices.
The practical implication is that classification is a regulatory pathway, not a liability forecast. Sizing a products tower off device class alone underprices the most common case.
Trend
Annual reporting grew 85% from 2020 to 2025.
2026 is excluded from the trend because the year is incomplete at the time of the pull. Report volume reflects reporting activity as well as underlying events: connected devices that transmit data automatically generate reports at a rate manual processes never did, so growth is not a clean proxy for deterioration in device safety.
Concentration
The top device types are all worn, implanted, or attached.
Dental implants, continuous glucose monitors, and infusion pumps dominate the list. What they share is dwell time: they stay with the patient for months or years rather than being used once and set down.
Long dwell time is the single most useful signal in this dataset for an insurance program, because it determines when a claim can surface. A device that remains in a patient for a decade can generate a claim a decade after sale, which is why occurrence form matters for products, and why on claims-made lines the retroactive date has to be preserved through every renewal and every change of carrier.
Methodology
How these numbers were produced.
All counts come from the openFDA device adverse event endpoint, open.fda.gov/apis/device/event, queried on 13 August 2026. openFDA reported the dataset as last updated 5 August 2026.
Totals and event-type counts come from the count=event_type.exact aggregation. Class counts come from count=device.openfda.device_class. Annual figures come from count=date_received, summed by calendar year of receipt. Device types come from count=device.generic_name.exact.
Two honest limitations. Device class is only populated where openFDA can match the report to a classification record, so class counts do not sum exactly to the total. And a MAUDE report is an allegation, not an adjudicated finding: submission does not establish that the device caused the event. Percentages are computed from the counts shown and rounded to one decimal place.
Two ways to start
Sizing a products tower off device class alone underprices the common case.
Explore on your own
Reviewing a contract or sizing a first program?
Run your sponsor MSA, GPO supplier agreement, or hospital purchase contract through the free tools. No login, no contact, results in about fifteen minutes. Built for founders and earlier-stage operators who want to understand the requirement first.
Talk to a specialist
Scaled operation, a live contract, or a complex program?
Multi-entity structures, high-limit towers, GPO and hospital contract portfolios, transactional and exit diligence. When the program is load-bearing, a specialist reads it clause by clause and rebuilds it to pass. End-of-business-day response.
Program review