Understanding the human factors contribution to railway accidents and incidents in Australia

One of several studies which have applied the HFACS (Human Factors Analysis and Classification System) methodology to understanding rail incidents.

What I like about this study is that even though the HFACS methodology has a natural inclination towards mapping human-related action/inaction (error and violation, or variations of performance variability, whatever you want to call it), it ended up strongly supporting a systems & organisational view.

40 Australian rail safety investigation reports were mapped against HFACS.

Note a couple of limitations/caveats. Investigation reports, and also secondary analyses with HFACS and any other method, aren’t just 1:1 reports of objective truths and facts that existed in the world. They’re heavily socially constructed, based on the existing assumptions, norms and worldviews of the investigators, the operating environment and also the methodology employed.

That’s not to say that there aren’t objective findings – a valve fails, a bolt shears, a permit wasn’t completed, there’s an exchange of energy, but their role and relevance in the accident trajectory, how they were understood and/or missed by people, and even where you decide to start and stop the investigation are influenced by a range of subjective, inter-subjective and objective factors.

On the latter, what-you-look-for-is-what-you-find (i.e. see Lundberg’s work). Therefore, what’s interesting about these types of studies isn’t just what percentage of events are categorised as error or equipment failure and the like, but also how the investigators came to those conclusions and how the methodology and their own worldviews directed them to those conclusions.

But, I digress. Below highlights the HFAC framework:

Results

Out of the 40 events, 21 were derailment of rolling stock, 10 collision, 6 safe working breach/irregularity and 3 involved shunting. The authors constructed 330 contributing factors.

In short, they highlight that:

·????????Nearly half of all incidents resulted from equipment failure – most the product of inadequate maintenance or monitoring programs

·????????For human-related associations, slips of attention (skill-based variability) was most associated, and usually resulted from decreased alertness and physical fatigue

·????????Inadequate equipment design (like driver safety systems) was frequently identified as an organisational influence

·????????“Nearly all incidents were associated with at least one organisational influence, suggesting that improvements to resource management, organisational climate and organisational processes are critical for Australian accident and incident reduction” (p1750)

A breakdown of associated factors are shown below:

Based on this secondary analysis of existing accident reports, 43% of the incidents had their primary or proximal factors residing in equipment failure. The authors note that “very few unsafe acts, preconditions for unsafe acts or supervisory factors were associated with these events” (p. 1753).

The physical environment was the most common non-organisational influence associated with incidents. Here high ambient temperature was the most widespread problem.

All incidents linked to derailments had equipment failure as a proximal factor, and all were associated with inadequate equipment or equipment in poor condition rather than operator action.

In all but four of the incidents “an organisational oversight was identified”, such as inadequate monitoring or checking of equipment/resource. Here they note that while some reports also identified an “unsafe act”, the “reports stressed that this unsafe act would not have been significant if equipment had been maintained to operational standards” (p1754) [*** ding ding ding, context and local rationality matter.]

23 incidents were seen to be triggered by “errors of frontline personnel”.

Under the “unsafe acts” section, they note that more errors resulted than violations [** I dislike that word in this context, but whatever]. Skill-based variability was most common, mostly via attention failure (15 of 20 events).

Seven decision errors emerged – seen mostly as misapplied procedural steps. However, they note that “on several occasions a poor decision was made because the employee possessed inadequate knowledge of the system or operational procedures” (p.1754, emphasis added).

This suggests that in most cases, people didn’t intend or volitionally override an existing process or control.

For “preconditions for unsafe acts”, the condition of the operator was the most commonly associated category. They say that “nearly all incidents were associated with an adverse mental state” (p1754) which resulted in an incorrect expectation or assumption. For instance, many drivers were said to have approached a red signal with the expectation that the signal would be green. This was amplified by distraction and decreased alertness.

Of the violations, seven were routine rather than exceptional. Procedural departures from operating procedures “was a regular occurrence, often going unnoticed or even tolerated by authority” (p1754; these types of workarounds are often good insights on how dysfunctional our systems can be and how difficult safe and reliable work can be for operators.)

Ten incidents involved an adverse physiological state, with physical fatigue being the most frequently mentioned. A small number listed personnel factors, like inadequate teamwork or communication.

Ten incidents had environmental factors implicated, with the only recurring issues being vegetation near the track obscuring the sighting of a signal and poor equipment design.

“Unsafe supervision” was rarely associated in the reports. They argue that this could be due to:

a) a high standard of supervision

b) a tendency for accident investigations to focus on underlying organisational failures

c) a smaller role played by supervisors in railway operation incidents

d) a tendency for supervisors to underplay their role in the incident given that they have a major role in the investigations

Wrapping things up, they highlight that organisational influences contributed to all but one event. 40% of organisational factors related to resource management, and equipment/facility resources associated across all levels of the incident hierarchy.

Nearly half of organisational influences belong to “organisational process”, including poor procedures, worker scheduling, and issues of oversight like risk management and safety management systems.

Overall, this paper based on its secondary analysis of accident reports [** noting the limitations of both the original reports and then analysing them a second time] highlighted that most incidents involved organisational factors. Where human performance variability was implicated, it largely resulted from issues of human physiology or sensemaking (attention, fatigue, confusion) or poor design/equipment.

Link in comments.

Authors: Baysari, M. T., McIntosh, A. S., & Wilson, J. R. (2008). Accident Analysis & Prevention, 40(5), 1750-1757.