Fred Manuele has a lot of respect for the work of Erik Hollnagel and Sidney Dekker. And why not? These two have been part of the movement in safety to stretch all our thinking when it comes to incident investigations.

I have been taking you on a journey through Manuele’s article which appeared in the May 2016 issue of Professional Safety magazine that drilled deep into the two most vital questions to be answered when investigating an incident. We’ve established that it’s one thing to find out the reveal the who, what, when and where of any incident.

[Image courtesy of Flickr user Jan Tlk via a Creative Commons license]

[Image courtesy of Flickr user Jan Tlk via a Creative Commons license]

The really tough part is the how and why of an incident, and it’s important to understand that the answers to those questions should not end up in a single root-cause answer, as Hollnagel and Dekker have written in the past, and Manuele looks to unpack in his piece.

Where Hollnagel and Dekker ultimately agree in their writings is that the idea of finding a single root cause for any incident is at best counter productive and at worst completely unproductive. The key is not to find a single cause  but instead to take a broader look at the context in which the incident occurred and look at all the factors that went into the incident.

Hollnagel and Dekker, in short, said that any incident models that attempt to find a single cause should be disregarded or used as only one tool and not the be-all, end-all of an incident investigation.

The two basically have been pushing for some organization (ASSE, OSHA, NIOSH, etc.) to get behind an incident model and promote it as the baseline model for all organizations to use in their respective incident investigations. We’ll briefly look at three types of models and add comments by Hollnagel and Dekker about each.

  1. Sequential: A sequential incident model looks at a sequence of events, in order, that result in the incident occurring. Hollnagel and Dekker see these are great in theory, but very few incidents fit into a two-dimensional, linear sequence; more often there are events that occur parallel to each other (as in, a third dimension) that a sequential model cannot properly take into account.  The danger with these is the causal effect that is found, which could result in the investigation ending too soon. Noting system deficiencies is necessary, but often is found after a cause is found, which was supposed to be the initial point of the investigation.
  2. Epidemiological: This kind of model is very dynamic,which is almost opposite of a sequential model. It takes into account all dimensions of an incident,  which includes causal factors that were active at the time of the incident and those that were building up over time and came to a head in conjunction with other factors at the point of the incident.
  3. Systems: These are the models that Hollnagel and Dekker seem to favor more than others, because these kinds of models take more of a 10,000-foot view of an incident and all the factors that led to it within the context of the overall system (or systems that may be working in concert or exclusive with each other). In these models, there is an assumed (and correct) interrelationship between all systems and the various factors that go into the creation of an incident. These kinds of models are the most “macro” of the models; these are the best ones, according to Hollnagel and Dekker, for fully assessing the context of an incident and thus holding the ability to find all the various factors hat go into any incident. Those of us veterans know that an incident is actually a “perfect storm” of several causal factors that converge at the same point. We think in a systems model.

Next time, we’ll go deeper in Manuele’s article and discuss root causes and root-cause analysis. And no, we’re not quite done yet after that!