Root Cause Analysis is a fine tool asked, constantly, to do work it was never built for. A serious event happens and someone says, reasonably enough, let us also use this to make sure the process is safe going forward. Or, stranger and increasingly common, a team reaches for RCA to assess a process where nothing has actually gone wrong yet, because RCA is the risk tool they know. Both moves blur a distinction that is worth keeping sharp, because each tool does one job well and neither does the other’s.
RCA looks backward. Its job is to understand a specific event that has already harmed or nearly harmed a specific patient: what happened, why, and how to prevent that particular pathway from recurring. It is event-driven, retrospective, and owed to the patient and often to the regulator. It is also, by nature, narrow. It examines the one pathway that failed, follows it to its contributing conditions, and produces a fix for that route. That is not a weakness. It is the job. Asking it to also survey every other way the process might fail is asking a coroner to run a health screening.
Proactive analysis looks forward. Its job is to examine a process while it is still working and find the failure modes that have not happened yet, across all the pathways, ranked by how much they matter. It needs no casualty and produces no account of any particular event. Failure Mode and Effects Analysis is the developed form of this, and it answers a different question entirely: not why did this happen, but where will this fail.
The two are not rivals, and the clean way to use them together is a sequence. When an event occurs, run the RCA, because the event is owed a proper accounting and the organisation needs to understand what actually happened. Then, on the redesigned process the RCA produces, which now has no track record precisely because it is new, run the proactive analysis to find the failure modes the redesign has quietly introduced. The RCA tells you why the wall fell. The proactive analysis checks the new wall before you lean on it.
The failure to keep the jobs separate is not academic. Run an RCA-shaped exercise on a process with no event and you get a hunt for a root cause that does not exist, which frustrates everyone and finds nothing. Run an FMEA instead of an RCA after a real event and you have skipped the accounting the patient was owed, in favour of a general survey that answers a question nobody asked. Match the tool to the question. Backward-looking harm gets the backward-looking tool. Forward-looking worry gets the forward-looking one. The mistake is almost always using a good tool for the wrong tense.
This article accompanies the book Before Harm Happens.