Accidents happen, a fact that not many can argue, whether its aviation, nuclear, maritime, and so forth. A topic that is more likely argued is accident analysis and causation, choosing an analysis model can be seen as choosing a camera lenses that gives a different outlook of the causation of the occurrence. Each lens has it positives and negatives, the difficulty comes in choosing the right lens (or group of lenses) for the right occurrence.
Now, using such lenses may be more difficult to implement as a practitioner, many organizations may not be able to fund the needed training, some analysis models require extensive time and resources to use, and others may not fit the overall objective of the organization’s investigation.
On the other hand, using analysis models has its benefits, it aids in minimizing the biases that result from an investigator’s previous experience. Structured models can help guide less experienced investigator’s, they are good communication tools, and aid in seeing the different relationships within the system (Underwood, 2013).
The objective of this paper is to review three different schools of analysis models but before the writer would like to give a simple scenario called “who hit the cat” in which these models can be implemented.
Who Hit the Cat?
The scenario begins with a managing director who arrives at work late, sits at the desk and finds contracts that were due to be processed and signed the day before. The manager grabs the contracts and throws them at the responsible secretary, in return the secretary goes to the human resources department, throws the contract and shouts “my boss is angry with me because you won’t do your job”. The human resources manager angrily goes to the frontline worker and says, “you will get these papers signed and processed or I will find some else who can!”. The frontline worker works additional hours overtime and only arrives home after a grueling eighteen-hour shift. While at home the frontline worker’s daughter asks if she can play to which the father angrily replies, “beat it”. The daughter runs to her room upset slamming the door open, not knowing their cat was behind the door.
Answering the question “who hit the cat” is a simple task (the girl), but as an investigator the question usually evolves around the premise of what caused this occurrence and how to keep it from happening again. The following segment will discuss three different schools of models and how they may aid an investigator in answering such questions.
Sequential Model: Such models follow a cause and effect philosophy, they see accidents as a sequence of events, a row of dominoes that fall one by one all of which was initiated by a root cause (the original domino) if which removed the accident could have been prevented. Sequential models have shown success in explaining occurrences related to mechanical/physical failures and occurrences in relatively simpler systems (think of the cat scenario explained above). Another benefit of sequential models is that they seem to do a good job of communicating consequences of events within proximity of the occurrence.
Sequential models are not without their drawbacks, the purely cause and effect nature of the model tends to handicap their ability at analyzing organization factors (Underwood & Waterson, 2013). It is the writer’s opinion that this model would be the most suitable at explaining the cat scenario (due to the simple nature of the problem).
Figure 1: Heinrich's domino model of accident causation
Epidemiological Model: Such models see accidents as a combination of latent failures (organizational influences for example) and active failures (actions or failure at the sharp end of the occurrence). Think of accidents as a disease, the latent failure are pathogens that stay dormant but when combined with active failures the consequences materialize. Such models do a better job (when compared to sequential) of explaining the relationship between factors at the blunt and sharp end of an accident. The Swiss Cheese Model (SCM) and the Human Factor Analysis and Classification System (HFACS) are examples of epidemiological models (Underwood & Waterson, 2013).
A drawback of such models is it still faces some of the issues of the cause and effect philosophy and some researchers argue that many epidemiological models do not really analyze accident but categorize them (Wienen et al., 2017). Using such a model on the cat scenario would require further effort but may help solves issues done at the organizational and supervisory level of the occurrence and would categorize the different failures (might put the failure done by the girl, cat, and worker in one category and the managers in another).
Figure 2: Reason's Swiss Cheese Model
Systemic Models: Unlike the previous two schools of models, systemic models do not follow a cause and effect methodology. Instead systemic models see accident as an unexpected behavior of the system resulting from the relationships of the different system components acting in ways that are rational given their local context. In such models, accidents are not a “root cause” problem but a control problem.
Example of such models are the Systems Theoretic Analysis Model and Processes model (STAMP) and Functional Resonance Analysis Method (FRAM) (Underwood & Waterson, 2012).
Although, such models may seem as the best choice on paper, they tend to be time and resource intensive and require a lot of domain knowledge and sometimes the benefits of using such a model may be hard to justify (think of justifying using such model in the cat scenario). This reason might help explain why such models have not gained widespread usage in the practitioner world (Underwood & Waterson, 2012).
Figure 3: Hollnagel's Functional Resonance Analysis Method (FRAM)
In the end, the writer does not believe in the notion of a bad analysis model but does believe some models (or group of models) may better suited for certain situations. Take the cat scenario for example, a simple problem with a minor consequence (when compared to accident in the aviation and nuclear industry) which would seem to favor a sequential method and would be hard to justify using a systemic one. Once an investigator or practitioner is familiar with the qualities of a model they can be valuable tools within their arsenal.
Underwood, Peter & Waterson, Patrick. (2012). A Review of Systemic Accident Analysis Models.
Contemporary Ergonomics and Human Factors 2012. 10.1201/b11933-82.
Underwood, P., & Waterson, P.. (2013). Accident analysis models and methods: guidance for safety professionals. Loughborough University. https://hdl.handle.net/2134/13865
Wienen, Hans & Bukhsh, Faiza & Vriezekolk, Eelco & Wieringa, Roel. (2017). Accident\Analysis Methods and Models - a Systematic Literature Review. 10.13140/RG.2.2.11592.62721.