What is ‘uncertainty tolerance’ and how can we measure it?
Some thoughts from the development of the ‘TofU’ project
Each decision we make is made with some degree of uncertainty, as it is impossible to predict the future perfectly. Ideally, decision-making relies on evaluating the pros and cons of each choice (often with limited information available to us) and deciding the best course of action.
For some decisions, such as choosing to catch a bus (method of transport prone to lateness) versus driving to get to a friend’s house, we may effortlessly make our choice as the consequences are relatively low stakes (being fashionably late and annoying your friend). However, this exact same choice in the context of travelling to work may be more taxing - as the consequences to making an incorrect choice are more salient (being fashionably late and getting fired). Thus, in the former example, we may allow for a risk of arriving late and catch a bus. In the latter, we may never catch a bus as the uncertain risk of arriving late, however small, has large consequences.
Unsurprisingly, the relationship between uncertainty and risk taking is more complex than it first seems. Aside from complexities regarding sources and contexts of uncertainty (for example, do we have accurate data on the frequency of bus delays?), some individuals can handle or ‘tolerate’ uncertainty more than others. Thus, for some, catching that bus to work may always be their preferred choice, despite the fact that there is a risk of being late - and importantly, that the exact risk is unknown.
For these individuals, we may say that they have high ‘risk tolerance’ (i.e. comfort in making a potentially risky decision) borne of a high ability to tolerate the uncertainty in this particular decision (i.e. ability to make a choice comfortably, even though the precise risks of a choice are unknown). Put another way, if the individual was omniscient and knew with certainty that a bus would be late 50% of the time, yet still chose to take it, that would be high risk tolerance. If instead the individual thought there was a 0-50% chance, and took the decision, this would be high uncertainty tolerance. As here, if the bounds of that interval were wide (i.e. uncertain) and the upper limit close to a ‘risky’ level (which would change, depending on how serious the risks are), and the individual took the bus, this may be a situation in which the individual shows high risk tolerance - perhaps because they have high uncertainty tolerance. This is because the individual has identified a possible strong risk likelihood but ignored it, as the uncertainty implies that there is a corresponding possible low risk likelihood - cancelling it out.
Uncertainty tolerance may be a very useful work trait for some high-risk professions. In medicine, for example, research has shown that doctors with low uncertainty tolerance feel more comfortable making conservative decisions, such as ordering more diagnostic tests or trying more treatment options. This may be problematic as it exposes patients to associated side effects and anxiety, and . ultimately, as this uses extra vital NHS resources, may also then affect other patients. Of course, this may work both ways - a doctor with very high tolerance may instead miss many patients that need treatment - for instance by not ordering enough diagnostic tests.
This is however an idea that needs testing.
It is possible that high uncertainty tolerance may instead facilitate a decision-making process where potential risks are judged and balanced more appropriately. This is because evidence shows that high tolerance may lead to lower doctor stress, burnout, psychological distress and higher wellbeing, resilience and wellbeing - perhaps therefore more mental capacity to think things through (i.e. less ‘cognitive load’).
Uncertainty is endemic in many aspects of healthcare. Patients present with an infinite combination of complaints, comorbidities, demographics, preferences and more, meaning it is difficult to not only determine the cause of their complaints, but also their prognosis and the best way to help them. As such, doctors and other health professionals must learn to make decisions in an environment both high in uncertainty and with high stakes. Nowhere is this more relevant than in emergency medicine. Here, doctors must quickly choose a course of management for patients. Many of these patients may have conditions that can quickly lead to serious health problems or death, and thus any delays or ‘incorrect’ decisions (insofar as a decision can be deemed wholly correct or incorrect) has a high chance of leading to poor patient outcomes. These doctors also rarely get feedback on the outcomes of their decisions as they don’t find out what happened to patients they met; they only have a short meeting with patients to get an accurate picture of their health needs; and often have limited prior information about patients.
Given the above, we believe that supporting emergency doctors in tolerating uncertainty may be not only beneficial for their own wellbeing, but may help reduce resource use, and thus benefit the NHS more widely. As such, the aim of our ‘TofU project’ is to find out whether there is a link between uncertainty tolerance and resource use in the emergency department context, and, for the first time, test the hypothesis that increased tolerance is not associated with worse patient outcomes.
Our project will also assess doctor factors that may be associated with uncertainty tolerance to inform the development of interventions, be they psycho-social (such as face to face support), educational (such as computer programmes that help make better decisions) or other learning (such as patient or colleague feedback). Given that more experience as a doctor had been observed to correlate with higher tolerance - we believe uncertainty tolerance is somewhat malleable and amenable to change.
One of the central challenges in designing our project was deciding how to measure ‘uncertainty tolerance’ amongst doctors. A paper by Hillen and colleagues (2017) helpfully identifies 22 measures of uncertainty tolerance, including the most widely used in the medical literature: the Physician’s Reaction to Uncertainty scale (Gerrity et al. 1990; 1995). Hillen and colleagues reviewed the questions used across all of the measures and categorised them into broad groupings.
For instance, in Figure 1, they determined that uncertainty may arise from randomness or indeterminacy of future outcomes (for example, a bus may often be late, but inconsistently so, so it’s impossible to predict), lack of reliability, credibility or adequacy of information (for example, you may not have good data on the frequency of late buses, so are unsure whether your lateness prediction is true), and features of information that limit understanding (for example, a bus being late is reliant on many interacting factors such as time, date and weather, so it is highly unpredictable). Further, in Figure 2, we can see that in response to uncertainty (influenced by the source) individuals have three broad ways of reacting to the uncertainty - cognitively (thoughts about what uncertainty and whether and how to deal with it), emotionally (the effect uncertainty has on their mood), and behaviourally (how they behave in response to uncertainty).
Thus, building on the work of Hillen and colleagues, we took the Physician’s Reaction to Uncertainty scale and then added new items to this ensure that we included items tapping all of the sources and responses to uncertainty identified in Hillen et al.’s model. All adaptation was conducted in a collaborative manner amongst the research team. Our final tolerance of uncertainty measure is 34 items long and takes around 7 minutes to complete.
Alongside our main study, we aim to assess the properties of our new measure - such as whether it is internally consistent (do all the items assess the same thing?) and whether it is associated with things that have previously been shown to, or should logically, correlate with uncertainty tolerance (such as years in the profession, burnout, risk aversion). We will also take feedback from completers about the wording, possible future items and the general structure and layout of the measure. It is hoped that future research will use our measure in order to capture the concept of uncertainty tolerance more holistically. This should also serve to ease comparability between studies and capture more variation in the concept.
We welcome all reasonable requests to use our measure for future research. If you would like a copy, please do not hesitate to contact me as the lead author of the study (email@example.com). We only ask that you tell us first what you are using it for, and a little detail about your wider related projects and background.
Figure 1. Sources of uncertainty.
Figure 2. Integrative model of uncertainty tolerance.
Gerrity, M.S., DeVellis, R.F. and Earp, J.A. (1990). Physicians' reactions to uncertainty in patient care: a new measure and new insights. Medical Care, pp.724-736.
Gerrity, M.S., White, K.P., DeVellis, R.F. and Dittus, R.S. (1995). Physicians' reactions to uncertainty: refining the constructs and scales. Motivation and Emotion, 19(3), pp.175-191.
Hillen, M.A., Gutheil, C.M., Strout, T.D., Smets, E.M. and Han, P.K. (2017). Tolerance of uncertainty: Conceptual analysis, integrative model, and implications for healthcare. Social Science & Medicine, 180, pp.62-75. https://doi.org/10.1016/j.socscimed.2017.03.024.
11th January 2021