HEART+

HEART+ is the culmination of work that Jerry Williams (originator of HEART) and I (Julie Bell) have been doing over the last 15 years to check and update the methodology. The printed manual (Part 1) contains all the information needed to use HEART+. In the back of the manual is a QR code to Part 2. Part 2 contains technical information about the data in Part 1, advice on how to combine EPCs, and guidance on how to calculate the Assessed Proportion of Affect (APoA). An additional document provides the complete reference list. In total, these documents are approximately 150 pages of information about HEART+, its application and how it was developed.

The GTTs are now underpinned by more data sources. The EPCs have been modified to account for new evidence and 5 new EPCs have been added. Each EPC has been articulated to inform the APoA decisions. The calculation equation has been modified to ensure it cannot go above 1 (an artefact of the original method).

Importantly, Jerry and I have addressed many of the issues that have been raised since its inception in the 80s. This is the first time that there has been a publicly available manual. I believe it’s worth the wait!

Copies of HEART+ have been provided for the British Library; The Bodleian Library, Oxford; The University Library, Cambridge; The National Library of Scotland; The Library of Trinity College, Dublin; The National Library of Wales, Aberystwyth. 

If you would like a personal copy of the manual, they are available at HEART+ for £95 (plus p&p).

For those who would prefer not to buy the manual the key data points are provided in the following tables. 

The nominal failure probabilities for each Generic Task Type as of 2023

*Calculation required:

EPC34:  x1.1 for first half hour; then x1.05 for each hour thereafter

EPC35: Chronic sleep loss (over several nights): 12 hours lost   × 1.1; for every 24 hours lost   × 1.2  OR Acute sleep loss (in one night): for every hour lost above 2 hours, x1.04

EPC37:     x1.2 for each additional person

EPC38:     x1.16 for every 10 years over the age of 25

EPC40:     × 2.4 Task performed between midnight to 3 a.m.

               × 2.0 Task performed 3 a.m. to 6 a.m.

               × 1.3 Task performed around 2-3 p.m. 

EPC41:   x1.45 for each % point between 15% and 11% oxygen

EPC43:   × 2.2 per 1000 days’ non-use

The HEART+ calculation is used to generate the Human Error Potential (HEP)

Each EPC is calculated as follows:

[(EPC Affect – 1) x Assessed Proportion of Affect (PoA)] + 1)

The following is a series of LinkedIn posts about some of the EPCs that are represented in HEART+

EPC 3 – A low noise to signal ratio

For the practitioner / risk assessor this is about whether the information you want people to detect stands out against the background of either audible or visual interference.

It’s about sensation, the ability to detect a signal, closely linked with perception, the ability to interpret the information our senses provide. The ability to interpret the information we sense would be impossible without the brain’s ability to collect, store, transform and apply knowledge based on our learning and experiences. But for practical purposes, this is about the ability of the eyes and ears to detect information because as brilliant as they are, they have limitations. The ability to detect warnings, alerts, alarms, is essential for managing safety. Does the signal you want people to react to stand out from background noise and other signals? If it’s an alarm, is it louder, different in pitch or both from other alarms? When designing systems, we need to optimise how information is presented, so people don’t have to work at their limits.

The science behind the design of alarms, warning lights and alerts has led to the development of guidance such as EMMUA 191 and 201. In HEART, a low signal to noise ratio is quantified to increase error potential by up to a factor of 10.

EPC8 – Channel Capacity Overload (increases the risk of human error x8)

Current thinking in Psychology has moved from thinking about information processing as pots of resource that are shared by a Central Executive function, to thinking in terms of Threaded Cognition.

Threaded cognition proposes that streams of thought can be viewed as processing threads, each carried by a serial cognitive procedure. Say an alarm goes off. This starts a cognitive procedure which cues for resource in a “greedy, polite” way. Greedy because it will use all the available resource. Polite because it will give up resources when no longer needed. If you are doing two tasks at once, then you will have two separate threads active. Multiple threads can be active at once, and as long as they don’t require the same resources, there will be no interference. However, if they do need the same resources, only one thread can be processed at a time. Other threads will have to wait, and performance will deteriorate as a result. The resources being cued might include sensory processing, attention, working memory, psychomotor control and so on.

The model explains how we multitask. The evidence shows that when people appear to be carrying out two tasks simultaneously, what they are actually doing is switching rapidly between tasks. The more dissimilar the tasks the easier this will be. However, if the tasks share characteristics, for instance both need visual processing resources, then the individual will only be able to do one at a time. One task usually suffers a greater detriment in performance.

If at the same time as the alarm goes off, an outside operative ‘radios in’, then the decision maker has two auditory processing tasks at once. If you add in a third channel of information, say someone else in the control room enquires about the alarm, then the likelihood that the operator would process all three signals appropriately is significantly reduced.

Channel capacity overload looks at how presenting information to the same modality (vision or hearing) through more than one channel affects human performance. Each piece of information will rely on the same information processing mechanisms and the cognitive threads will need to cue for processing, during which time important information can be lost from the system, reaction times will slow, and errors such as right action on the wrong object will become more likely.

EPC 9 – the need to unlearn a technique and apply one which requires the application of an opposing philosophy.

Research shows that learning changes the structure and function of the brain not just behaviour (or perhaps changes behaviour by changing the structure and function of the brain). This research includes studies of London cab drivers who acquire “The Knowledge” – information about c.25,000 streets within a 6-mile radius of Charing Cross Station and the location of landmarks within them (Eysenck and Keane, 2016). Adults who qualified in the knowledge were found to have a selective increase in brain matter (the little grey cells), whilst those who failed did not. This research has been replicated in people learning to juggle – 5% increase in grey and white matter; instrumental musical training in children – increase in the primary auditory area and primary motor area; and adults carrying out musical practice – more practice, larger cortex.

With practice, behaviours become more and more “automated”, requiring no conscious input. This allows us to carry out routine tasks whilst our limited cognitive resources think about something else. For example, once we have learnt to drive, we do not have to consciously think about where the gear stick is or how much to depress the gas pedal to accelerate gradually. We can focus on monitoring and interpreting the environment outside the car. In fact, many experienced drivers arrive at a familiar destination without any recollection of the drive that has brought them there. It seems that this learning is embodied in the brain.

This automaticity is an essential survival trait. We would not have survived as a species long if we had to use our limited information processing capacity to contract and extend muscles as we ran away from a sabre-toothed tiger. However, it can also result in problems. The initial behaviour has changed the shape of the brain. You have to reshape the brain – and that will take a lot of practice.

A colleague of mine at the HSE investigated an accident in which a maintenance mechanic suffered life-threatening injuries. He had employed the “right-tighty, lefty-loosey” rule as the isolated butterfly valves on the pipework. Unfortunately, one of the valves had been installed back to front and he had locked it in the open position.

When controls and safeguards work in the opposite way to expected, error will increase – in HEART, error is quantified as 6 times more likely.

Communication: If you have hybrid workers with onsite responsibilities, how do you promote effective communication?

Hybrid working is an attractive way of working for many. It brings benefits for focusing on desk-based work while easing the strain between work and home commitments. The quality of communication between staff working at home and those on site is something we’ve been thinking about recently. When we look back at the Longford explosion (1998) the remote working of engineers (moved from the Longford site to Melbourne) resulted in those engineers having oversight rather than a ‘hands on’ approach.

Within the major hazard sector, the importance of communication in critical activities such as shift handover is well established by Human factors/ergonomics – Shift handover (Human factors/ergonomics – Shift handover (hse.gov.uk)). How does the oncoming shift know about the site activities and understand the impact that could have on their work? Communication protocols for using radios and long-range discussions within and between maintenance teams Human factors/ergonomics – Safety critical communications (Human factors/ergonomics – Maintenance error (hse.gov.uk)), are essential to reduce misunderstandings. Communication is also essential in permit-to-work systems, so people know who is working where, what systems are offline, and what controls have been put in place to manage emergencies during that time.

If staff work at home, should similar standards be applied for communication between those at home and those on site? I think so.

Clarity is needed about what work is being done at home and how it applies to the site, how information will be communicated to be effective, promoting a shared understanding of what is happening on site, with particular importance where safety is implicated.

Within HEART+, communications issues are addressed in EPC16 – an impoverished quality of information conveyed by procedures and person/person interaction. Evidence shows that human error can be around 3 times more likely in situations where communication is impoverished.

EPC17 – Little or no independent checking or testing of output.

How useful are independent checks? Independence is a difficult condition to achieve for checking activities. Whilst double checks start off as a useful risk control measure, if few errors are detected, then checking may become cursory or non-existent. Work pressure and confidence in colleagues can result in checking tasks being devalued.

Priming can also undermine the accuracy of checks. Checkers are much less likely to spot an error if they are told the “right” answer in advance. For example, saying “this is 5mg of morphine, can you check?” will prime the checker to perceive the quantity as 5mg. The question activates the neural pathways associated with 5mg of morphine. Once activated, it becomes easier to activate again. Asking “tell me how much morphine is here” is more likely to result in catching an error.

Independent checks may be a weak form of control measure, but in the real world, it’s sometimes necessary and beneficial to include them. In pharmacies and busy wards, double checks on dispensed medications are the only control measure to set against high task demands. Independent checks are also used in pharma as part of quality control, often supported by training to stress the importance of these.

There is fairly consistent evidence that – so long as the check is truly independent – these checks can reduce errors by a significant amount. This may mean the person checking not just the results, but the thinking and processes behind it. Asked to check a system is isolated? Don’t rely on what you are told, walk the line yourself. In HEART, the lack of checking is estimated to increase the likelihood of error by 3.

Distraction: Think you can work with distractions?

I don’t think any of us would be surprised to find that distractions can have a negative impact for individuals. In HEART+, we have quantified the impact of distractions on human reliability and the results might be equally disturbing for workers in safety-critical roles.

Distraction is where your attention is drawn away from the thing you need to focus on, to something else, leading to possible confusion and an increased likelihood of task failure. The impact of switching attention between two tasks, and with a task being interrupted has been investigated. The findings from both types of study showed that distractions make you approximately 4 times more likely to make a task error.

Keeping track of a situation, knowing where you are in a task procedure, knowing if you locked the front door, are all impacted by distractions. There are a multitude of studies looking at the impact of distractions on human performance in different settings from cockpits to operating theatres. 

EPC 24: A need for absolute judgements that are beyond the capabilities or experience of an operator make it 1.8 times more likely that an error will occur.

The picture illustrating this post is a screenshot from the online retailer Amazon of some post-its I recently bought. Post-its are the favourite tool of social scientists and an essential part of the Human Factors toolkit as far as I’m concerned. I’m not picky though – as long as they are big enough to write on, and they stick to the wall (and don’t take the plaster off), I’m happy.

The size of the parcel from amazon made me uneasy when these particular post-it notes arrived – because they are in fact miniature post-its. One quarter of the size of your standard square post-it. Why? Why? Who would want tiny post-its?

I’m not alone in making errors when purchasing things online. Stories abound of people being surprised when dolls-house furniture turns up instead of the chest of drawers or sofa they were expecting.

In this case, the seller could reasonably claim that they did include size information quite clearly in the blurb about the product. However, many people struggle with imagining size from a quoted figure, and this is made much more difficult when an absolute judgement is required. That is, a judgement made without any reference points for comparison. Had these post-its been shown next to a reference object – a ruler, a pencil, a normal post-it – I would have made a relative judgement and I don’t think I’d have got it wrong.

When it’s critical that someone makes the right judgement call, like opening a valve just enough but not too much, make sure they have a reference point to support that judgement and they are much more likely to get it right.

EPC 40. Time of Day

Roh et al (2023) collected data on computer usage as a proxy for productivity, in a study of 789 office-based workers over a two-year period. One of their key findings was that worker’s productivity varied by time of day and decreased in the afternoons. There was also a significant decrease on Friday afternoons – probably compounding time-of-day effects with tiredness at the end of the working week and the anticipation of the weekend. Of note for error analysis, is that the number of typos dramatically increased in the afternoon.

One of the ‘newly’ quantified error producing conditions in HEART+ is “Time of Day”. It’s about the effect of mental alertness on reliability at particular times of day. While it is well established that people are most compromised between around 00:00 and 03:00, they also suffer a deterioration in performance between 03:00 and 06:00, and between 14:00 and 15:00. The Roh et al findings provide further evidence that the scheduling of work is important, particularly for safety-critical tasks or any activity needing optimal performance from the worker. HEART+, predicts that a task performed around 14:00 and 15:00 is 1.3 times more likely to fail than at times of peak alertness. While this risk is relatively low compared to a task performed between midnight and 03:00 which is 2.4 times more likely to fail, it might be important.

Finally – this is a link to a short piece that I (Julie) wrote about quantification, enjoy!

Why I love quantification – Human Factors Expertise Ltd (hf-expert.co.uk)