乌鸦传媒

A significant number of medical device adverse events that resulted in patient deaths were miscategorized in the FDA's Manufacturer and User Facility Device Experience (MAUDE) database, according to an analysis using a natural language processing algorithm.

Flagging terms commonly associated with death, the algorithm identified 290,141 reports in which a serious injury or death was reported; 52.1% of these events were reported as deaths, and 47.9% were classified as either malfunction, injury, or missing (the report was not put in any category), reported Christina Lalani, MD, of the University of California San Francisco, and colleagues.

From their overall sample, they found that 23% of reports with a death were not placed in the death category, amounting to roughly 31,552 reports filed from Dec. 31, 1991, to April 30, 2020, they noted in .

The physicians and manufacturers who report adverse events with medical devices must choose whether to classify the event as a malfunction, injury, death, or "other." Per , the reporter is required to categorize an adverse event as an official death if the cause of death is unknown, or if the device "may have caused or contributed to a death."

In the current analysis, about 56% of reports involving patient deaths were for class III devices -- the highest-risk medical devices -- and 40.1% were for class II devices, Lalani and team wrote.

The three most common product codes among the adverse event reports were for ventricular assist bypass devices (38,708 reports), dialysate concentrate for hemodialysis (25,261 reports), and transcervical contraceptive tubal occlusion devices (14,387 reports).

The natural language processing algorithm used in this study was implemented to scan through reports and identify terms such as "patient died," "patient expired," "could not be resuscitated," and "time of death." Of the 70 natural language processing terms in total that were associated with a death, 62 (88.6%) were found among miscategorized adverse event reports involving a patient death. And, of all 62, there were 17 terms that had an estimated percentage of 100%, meaning that "every time that term was used, the patient had died, even though the reporter had not classified the report as death," the team wrote.

Only 18 terms had sample sizes large enough for researchers to calculate confidence intervals; among them, the words "death" or "deaths" were linked to 12% of adverse event reports in which a patient died, but were classified as malfunction, other, or missing -- the highest rate of all the analyzed terms.

Lalani and colleagues acknowledged that one of their study's major limitations was that they only included reports that contained at least one term associated with death, as opposed to all the reports from the MAUDE database. This, they noted, will likely lead to the underestimation of the actual number of deaths that were improperly categorized.

"The classification chosen by the reporter is vital, as the FDA must review all adverse events reported as deaths, which is not the case for other reporting categories," the authors wrote. Improving the accuracy of these reports is crucial, since the frequency of patient deaths is what can prompt the FDA to pursue investigations into the device's safety, they added.

Notably, almost all of the reports analyzed in this study -- 95.9% -- were submitted by manufacturers, pointing to an inherent conflict of interest, the authors pointed out.

"It may not be in their interest to facilitate identification of serious problems with their own devices in a timely manner," they wrote. "There have been multiple instances of delays by manufacturers in reporting serious malfunctions and deaths that were associated with medical devices, as well as complete failures to report."

Therefore, it's likely that a significant number of patients have been unknowingly treated with devices that were later revealed to be dangerous, Lalani and colleagues noted. For example, they referenced the reporting failures that occurred from 2002 to 2013, when 32,000 women reported adverse events associated with the permanent birth control device Essure; the FDA, however, only received 1,023 adverse event reports from the manufacturer during that 11-year period.

"For these reasons, physicians, hospitals, and patients should submit reports directly to the FDA instead of or in addition to reporting through the manufacturer," they concluded.

In response to this story, a spokesperson from the FDA told 乌鸦传媒 via email that the presence of the word "death" -- or any other related terms that the authors used in their analysis -- does not automatically mean that the event should be classified as a death in the Medical Device Reporting (MDR) database.

"Many of the MDRs in the JAMA Internal Medicine study that were characterized as having been misclassified were for Class III devices, which often are used in sick individuals who may be at high risk of death from their underlying illness," the spokesperson wrote.

The FDA has reached out to the research team for further review.

Comment