Although “probable” and “likely” are not precise synonyms, if your risk assessments use any other qualitative terms (likely, often, frequent) assume that the same concerns apply.
For coins and dice we know how many possible outcomes there are and how many of those outcomes are of interest. It is easy to calculate the probability of any given outcome. As explained in “Q is for quantitative risk assessment” in IOSH Magazine, June 2017, in which historical failure rates are available for a system that can be fully described, it is possible to make an estimate of the probability of the system’s failure. However, for most occupational health and safety risk assessments (as opposed to process safety) there are too many variables, such as human behaviour, and too much missing data.
Even if you could access data on how many times people fell from a ladder, we don’t know how many times the ladder was climbed successfully. It’s like knowing how many times the die came up with a six, but not knowing how many sides it has, or how often it was tossed.
Hence, rather than numbers, qualitative probability terms are used. This addresses the problem of having to calculate probabilities from poor data and provides a rule of thumb for prioritising the management of a hazard. However, research for the Intergovernmental Panel on Climate Change (IPCC) in 2009 suggests that using probability descriptors creates “an illusion of communication”.
The findings do undermine the myth that there is a universal concept of ‘probable’
Researchers led by David Budescu asked subjects to assign frequency ranges to qualitative terms used in sentences such as, “It is very likely that hot extremes, heat waves, and heavy precipitation events will continue to become more frequent”. Estimates for “very likely” ranged from 60% to 95%. Estimates for “very unlikely” were even wider – from below 5% to above 80%. More than 75% of subject estimates for these terms were outside the ranges defined by the IPCC experts.
Specialist staff are no better. Richard Earlam and colleagues asked 105 consultants, medical staff and students at the Royal London Hospital to assign probabilities to words used to describe the risk of adverse outcomes after surgery. Using a statistical approach that excludes extreme answers, the central range of answers for “possible” was from 15% to 50%, and “probable” from 50% to 80%.
Of significance for OHS risk assessments is the tendency for people to assign different probabilities to the same qualitative probability term depending on the severity. In one study students assigned an average probability of 8.8% to the term “rarely” when told a treatment would rarely cause a dry mouth. This dropped to 2.3% when a more severe outcome, impaired vision, was described.
Another study found that rare deaths were given significantly lower percentages than rare occurrences of pneumonia after surgery. This suggests the possibility that, when we carry out risk assessments, our estimates and those of our colleagues are conflated by what we imagine as the outcome.
Applying “unlikely” to the chance of falling from a step ladder could mean something quite different from applying “unlikely” to the chance of falling from a greater height, with the result that the same priority is mistakenly given to an event with a lower risk.
If the probability of someone falling each time a ladder was climbed was 1% that would be extremely high – tolerable levels of risk would be a 1 in 1,000 chance a year, not each climb. And yet Earlam reports on earlier studies in which “very rare” and “almost never” are given values from 5% to 10%.
Although the focus of research in the use of qualitative terms has been in a medical context or in understanding climate change, the findings do undermine the myth that there is any universal concept of “probable” or any other qualitative terms used by OSH professionals.
Earlam’s advice that surgeons focus more on “reducing the incidence of those risks than in the actual meaning of words” is equally relevant to OSH.