FDA's Root Causes and Use Error

The way FDA analyzes recalls and applies its list of root-cause possibilities needs a dose of transparency.

Since 2016, the FDA Recalls Database has had a pull-down menu of the FDA-determined root cause of each recall. There are 43 items on the list, including "other," "pending," "under investigation by firm," and "undetermined/unknown." This list appears to have hierarchical issues in that some items (e.g., device design) seem to encompass other items (e.g., component design/selection and software design change). I have not been able to find any explanation of these 43 terms or the methodology by which FDA selects them, and my inquiries to FDA in this regard went unanswered despite an autoreply from DICE that I would get a response in one to two days.

Recalls have been tagged with these terms on the individual recall pages long before they appeared in the pull-down search menu. It might also be noted that entering these same terms in the "reason for recall" box gives a different set of results. Using the "simple search" option also yields recalls that are different from those found using the menu.

One of the terms in the menu is “use error,” defined elsewhere by FDA as referring to “to a situation in which the outcome of device use was different than intended, but not due to malfunction of the device.” An alternative FDA definition is “user action or lack of action that was different from that expected by the manufacturer and caused a result that (1) was different from the result expected by the user and (2) was not caused solely by device failure and (3) did or could result in harm.” Use error is generally distinguished from “user error” in that user error implies the conclusion that the user is to blame, while use error recognizes that the cause of the error could be the design of the device or the system as well as the environment of use. (Use error was first introduced in 1995 in an MD+DI editorial.) Under either definition, one would not expect a problem that is clearly the fault of the manufacturer to be identified as a use error, as is sometimes the case in the recall database. Further, causation clues can be found in what it took to correct the problem.

For example, in a lab system software recall, the problem was identified as mis-posting of results, and the correction was a “mandatory hotfix.” "Hotfix" is a great term, since it implies a quick fix that has not been thoroughly tested. But how is this situation a use error? In another example, a bracket was found to be failing at a welded joint. Without identifying anything that the user was doing to cause this failure, it would appear to be strictly a design and/or manufacturing problem. Yet, the fix was for the user to inspect all devices that include the fixture for damage before every use. Apparently then, the user could have noticed that there was an impending weld failure before using the device, and if they now inspect the device before every use, they could presumably recognize a failing weld in a timely manner. Henceforth, it will therefore be the user’s fault if a broken bracket injures someone. In a third example, it was noted that an ID dot was misplaced on an orthodontic device, which might guide the user to misplace it in the patient. The recall action was to return all unused product. Once again, how is it a use error if the user follows the marking on the device? Why isn’t it a manufacturing error? One reason might be that there is no “manufacturing error” on FDA’s list.

I also noted a classic human factors situation in a use error recall in which a component could be connected to a patient temperature cable instead of the patient temperature probe. A longstanding human factors design principle is that things shouldn’t be able to be readily connected to two different cables when only one such connection is correct. The “fix” in the recall was to tell the users in a safety alert to, in effect, be more careful. This is an example of the least effective third way to deal with a hazard—the first being to eliminate the hazard and the second being to guard against the hazard.

The various terms that reflect no actual answer have some long outstanding examples. The earliest “pending” is from 2008, while the earliest “under investigation by firm” is from 2011. However, these are both for “terminated” recalls, and a footnote tells us that “per FDA policy, recall cause determinations are subject to modification up to the point of termination of the recall,” apparently meaning that once the recall is over, what caused it in the first place is no longer worth reporting even if determined.

The primary reason to understand the causes of recalls is to use this knowledge to prevent recurrence as causation information provides feedback into design, however slow and imperfectly this may occur. Additional reasons may be to assign blame in the litigation domain and to satisfy regulatory requirements, a mission that sometimes exists unto itself. If FDA-determined reasons are to be part of this process, then how FDA determined its list of reasons, and how FDA analyzes recalls and applies its 43 possibilities, needs a dose of transparency.

William A. Hyman

William A. Hyman is a professor emeritus in the department of biomedical engineering at Texas A&M University and adjunct professor of biomedical engineering at the Cooper Union. Reach him at [email protected].

500 characters remaining
After following up with DICE, and some supervisory personnel, I did get a partial answer about the list of FDA root causes with respect to the meaning of each term. I was told they are self explanatory.