How much can or should we rely on accident or injury history when putting together risk assessments?
This question is a natural one and it is frequently asked, but there is surprisingly little guidance on it. It initially seems intuitive to say – “we’ve done it this way for years, never had an accident or injury.”. Unfortunately, that’s also often one of the first things that is said AFTER something has gone wrong as well. So what’s going on, and how much can or should we rely on history to predict the future?
When we put together a risk assessment, we need to consider two main things – 1) is the hazard coming to fruition foreseeable? and 2) are there any reasonably practicable controls?
Look out ahead!
The simple analogy I use is when thinking about historical data is, you can’t predict what’s in the road ahead by looking in the rear-view mirror – it could be that a truck in front has just done an emergency stop, or you’ve reached a roundabout!!
When history works – sort of
Risk assessments are about predicting what could happen, and looking for models that help you to do that. In some industries – primarily high hazard industries – the risks associated with particular arrangements are rated quantitatively, based on historical data. In my experience this is because a) situations have particularly high levels of similarity between sites, and consistency over time, b) the broader historical data is available (sometimes), and c) the high potential cost and disruption of interventions etc. demands a level of cost-benefit detail and analysis that means a quantitative approach is needed. However even in these industries, there is sometimes unease about using quantitative data that is essentially historical.
We’ve always done it that way
Very typically, when investigating incidents or injury, the case for the defence is based on ‘we have always done it that way, it’s never happened before, so it was not forseeable’. Although that might seem initially to hold some logic, unfortunately it isn’t sound reasoning. The phrase ‘accident statistics’ holds the clue. When there’s a hazard present, it’s a numbers game, it’s basically a wait until something goes wrong, because eventually, it will. There is a useful phrase in an industry technical bulletin that sums this up, paraphrased as:
“lack of accident / injury history does not necessarily indicate safety, but may indicate that, so far, ‘the gambler has had good luck’.”
Another example to emphasise what is going on with an over reliance on historical data; if someone flips a coin and gets six heads in a row (odds are 1/64 by the way – which in the grand scheme of things are pretty low!), you wouldn’t then change the odds from 50/50 for the next flip (in fact you might even fall into the trap of saying the odds must surely now be in favour of tails – another error in using historical data). You (hopefully) wouldn’t say, “look, it’s always been heads, it’s bound to carry on being heads!”. The right thing to do is examine the current situation (it’s just a coin – and no – it’s not double sided!!), use a predictive model if you have one and place your bets so to speak – me personally, I’d stick at 50/50 every time!! In the same way, if there is a clear hazard present in the workplace, we should base our decisions about what to do, on the present situation and any objective, forward looking, risk rating models we have, and not leave it because nothing has happened before. History does not equal foreseeability.
High score looking forward, low ‘score’ looking back – what to do!?
What about when you use a risk rating tool and the risk scores come out high, but historically you’ve not had any injuries? This is a question that came up recently at a training course I ran on using the ART and MAC tools. If you know or think that your ART or MAC based risk assessments scores will indicate high risks – then does this mean those high scores wrong? should I not place too much faith in them?
In this sort of case, until you get to the stage of allocating resources to the issues, my advice would be not put any significant reliance on the historical data. One of the key reasons being that your historical data might not be the full picture. If we stick with the example of musculoskeletal injuries – maybe some people just put up with them, maybe some people just decide to leave. So again it is a numbers game – x nr of people may be affected, and out of them eventually it will affect someone who; a) isn’t predisposed to put up with it and b) decides to take civil action or raise a complaint with the regulator. Just because it hasn’t come to a head yet, doesn’t mean it isn’t going on. Historical data is potentially a calm looking sea with rocks and sandbanks just under the surface; hiding near misses and issues that have simply remained unrecorded.
Allocation of resources
Where historical data might be useful, once a more objective assessment of the risks has been done, might be when it comes to allocating resources. Say you have two or three manual activities all rating similarly for risk exposure. You might want to say, okay we know for one of these tasks there have been reports of problems – and focus on that task first (but you should still focus on all three of the tasks in due course). It would be difficult to argue against that being a reasonable use of historical data. And within reason you could even take that same approach when considering how to allocate funds to make improvements or changes to the tasks.
Surely it will be taken into account if things go wrong?
If something does go wrong, as things unfortunately have a habit of doing, and an HSE investigation happens, the argument that ‘it hasn’t happened before so it wasn’t foreseeable’ is very unlikely to hold any water. It just won’t form much, if any, defence against the investigation and decisions about what enforcement action to take. HSE’s view of foreseeability is likely as far as possible to be based on forward looking predictive models, guidance and logic. The only point when history could be raised with a hope of it attracting some favour, would be in court as mitigation (not defense). Because although logically, it doesn’t really hold water, once the idea is ‘out there’, it could appeal to the forgiving nature of jurors or judges!!
Another scenario where you might reasonably rely on historical accident / injury data is when you simply have no predictive, objective model for foreseeing the risks. In reality this is what happens a lot of the time when risk matrices are used instinctively – taking the ‘finger in the air’ approach. It might seem as if it’s a forward looking process, but in reality, the decision is likely to be heavily weighted by knowledge of the history of the hazard on site.
This is one of the problems with risk matrices. A low risk score becomes a pivot point in deciding what to do about it (or whether to do anything), rather than simply saying – “Is there a hazard present and if so, what can we do about it?” It might be that there is a simple low-cost solution which is never arrived at, because the low risk score has diverted attention onto other matters. And at the same time, the potential incident / injury could materialize tomorrow – when the gambler’s luck runs out.
In summary, where you don’t have objective predictive data, take great care in relying on a risk score (which in all likelihood has a substantial backwards looking element) to inform a decision about what can be done about a hazard viagra 100mg prix pharmacie. Don’t allow low scores to become a diversion and always always look for any low hanging fruit.
Historical accident data – what’s it good for?
- Allocating resources once risks have been objectively assessed.
- Impressing visitors and investors.
- Appealing to jurors and judges.
Not so good for:
I hope this is a useful post and sparks some discussion. Feel free to leave comments – shoot me down in flames or offer any sage words of your own!!!