The problem with root cause analysis

This brief paper talks about some challenges with RCA methods. For an article in “the problem with” series, this one is pretty mild and somewhat favourable towards RCA.

Since RCA refers to a range of approaches and tools, this looks at more general issues and no particular tool. The authors note upfront that RCA does appear to have value (at least in healthcare) but has been applied without enough attention to what makes it work within a particular context and without customising the specifics for its purpose.

However, RCA approaches have “consistently failed to deliver benefits on the scale or quality needed”, some reasons of which are now covered.

1. The unhealthy quest for ‘the’ root cause

The name itself is said to be a problem as it may, even inadvertently, suggest that a single root cause or a small number of causes can be found. This can promote “a flawed reductionist view” (p417).

Such a perception can result in “simple linear narratives” that end up replacing more complex and potentially helpful “multiple and interacting contributions to how events really unfold” (p417).

They note that this effect may be exacerbated by some RCA techniques which favour a temporal narrative (e.g. timelines or 5 Whys) over a wider systems view.

2. Questionable quality of RCA investigations

A number of issues impact the quality of learning from RCAs. One is an overarching time pressure to resolve incidents quickly. However, reconstructing situations to learn from them involves time to collate various forms of information from people, documents, observations, IT etc.

Another factor they note is problematic in healthcare is that investigations are mostly undertaken by local teams and not necessarily aided by people with proficiency in systems thinking, human factors, cognitive interviewing and the like indicative of high-risk industries.

3. Political hijack

Here the authors nicely state that “Constrained by strict timelines, and skewed by hindsight bias … and lack of independence from the organisation where the event took place, RCAs … often end up a compromise between ‘depth of data and accuracy of the investigation” (p418).

The overarching time pressure can lead to goal displacement where the production of the investigation report is seen as the end product rather than effective double loop learning. Further, the investigation reports fail to account for the rich interpersonal interactions and discussions or the hierarchical tensions and power [although this isn’t restricted to RCA].

In alignment with other research about how factors other than accident causal factors decide on the recommendations [e.g. Lundberg, Rollenhagen & Hollnagel], it’s said this is a problem with investigations where investigations may end their analyses once they reach “mutual convenience” or where things get too difficult (organisational factors). This also isn’t restricted just to RCAs, though.

4. Poorly designed or implemented risk controls

Not much needs to be said here as it’s familiar to most, but evidence points to the endemic issue of RCAs corrective actions focusing on ‘weaker’ solutions—many administrative—like reminders, toolboxes and training, rather than effective organisational and work redesign and engineering etc.

Further, it’s noted that little resources is invested in following up on whether improvement actions were implemented or effective for their task. Research highlighted by the authors found between 45 – 70% of action plans hadn’t been properly implemented.

5. Poorly functioning feedback loops

Here it’s said that for effective learning from incidents, several conditions must be satisfied and most obviously, sharing the outcome of the incident with stakeholders. Learning is said not to always happen by itself but “purposeful intent is needed both to disseminate the findings … and ensure that the recommended actions made salient and actionable” (p418).

However, feedback mechanisms (in healthcare) are said to function poorly and contribute to the “disenchantment of staff … and frustrating the kind of double-loop learning … needed to secure change” (p418). Sounds a lot like the construction industry.

6. Disaggregated analysis focused on single organisations and incidents

Here it’s discussed that RCA approaches tend to favour analysing individual events in isolation “and within bounded organisations”. Learning from broadly and sharing learnings more broadly are said to be limited.

Over-committing to firefighting single incidents may “[frustrate] the organisation’s ability to assess its vulnerability to recurring events” (p418). It may further “lead to an unwarranted commitment of resources to averting specific very rare events rather than addressing the conditions that allowed the event to occur” (p418-19).

7. Confusion about blame

Here they discuss the challenges around just culture approaches and determining accountability. One point they raise and not related to RCA itself is that some just culture tools used in investigations act like “prescriptive algorithms and decision tools (such as culpability tree) to objectify culpability. Such ‘calculus-like logic’ … may imply that actions committed by staff are binary (either acceptable or unacceptable) without appropriate appreciation of the messiness of the system in which the action occurred” (p419).

8. The problem of many hands

This point discusses the challenge of where many stakeholders are involved and resultingly “no individual is responsible either for that outcome or for fixing the problems that caused it” (p419).

They say that investigations may fail to assign responsibility to individual actors and rather absorb responsibility into the organisation. I can’t speak to healthcare but my experience in construction, oil & gas and other industries is the opposite: I find no shortage of individual responsibility and rather an abject lack of consideration for organisational factors.

In moving forward, the authors discuss some ways to improve the quality and impact of RCA investigations, which I won’t cover. However, they do note that “psychological and emotional readiness of patients and families involved in the investigative process needs to be considered” (p419).

Applying this logic to construction, the psychological and emotional readiness and wellbeing of workers involved in incidents need to be considered, as I think the shame, concern, stress and embarrassment that workers involved in incidents can be easily overlooked – as too their role in helping to repair the needs of stakeholders (as per restorative just culture approaches, e.g. Dekker).

Link in comments.

Authors: Mohammad Farhad Peerally, Susan Carr, Justin Waring, Mary Dixon-Woods, 2017, BMJ Qual Saf