Human Psychology, Nobel Laureates and Radiology Demand Management

Demand management; responsible requesting; appropriate referring; clinical vetting. Call it what you like: managing the demand for medical imaging is a hot topic. When it’s cheaper, easier and apparently more objective to get a scan than to get a senior and holistic medical opinion the demand for imaging will only increase.

Whether demand management is a good or a bad thing depends on your point of view. When I was a registrar, the fellow told a story about a placement he had done in a large hospital in the United States. After vetting a sorry litany of poorly justified inpatient ultrasound requests by chucking a third of them in the bin (as was his normal NHS practice), he was called aside by the Senior Attending to be told in vituperative terms and with a healthy smattering of agricultural language, that he had cost the department $25,000 in one morning and to please cease and desist.

On the other hand, in the increasingly austere environment of United Kingdom healthcare, the catchy “supporting clinical teams to ensure diagnostic test requesting that maximises clinical value and resource utilisation” is an important component of effort to increase the productivity of the service. This remains true even if the inexorable increase in demand, driven by increasing hospital attendance, direct requesting from primary care and widening indications for complex imaging such as CT and MRI, is unavoidable, and even mandated.

What levers can we put in place to try to ensure responsible requesting? There are some lessons we can learn here from two Nobel Laureates: Daniel Kahneman and Richard Thaler. Kahneman won the Nobel Prize in Economics in 2002 for his work on Prospect Theory, and Thaler in 2017 for his work on behavioural economics. Their work on how people make decisions has implications for radiology requesting.

In his book ‘Thinking Fast and Slow’, Kahneman describes the multiple biases and cognitive traps that distort the way we think and form judgements. One of these is ‘base-rate neglect’ in which we make narrative judgements about individual cases without thinking about the statistical likelihood of that judgement being correct. One of the simplest examples of this in his book is the following:

A young woman is described as a ‘shy poetry lover’. Is she more likely to be a student of Chinese or of Business Administration? 

The base-rate (numbers of people who study the two subjects) tends to suggest the latter, and the fact that she is female and is a ‘shy poetry lover’ tells you nothing of objective relevance as to which subject she chose. But which subject jumped into your mind? Worse than this, even when we are made aware of the base-rate, we tend to ignore this information. We continue to do this even when we are also reminded of the tendency to neglect base-rate: you probably find, even now, you cannot quite shake the image of the young woman studying Chinese.

In a radiology requesting context, the base-rate might be the statistical likelihood of a specific diagnosis. It might also be the rates of requesting of an individual, department or Trust relative to relevant peers (‘over-‘ or ‘under-requesting’) or other averaged metrics. 

But what the base-rate neglect phenomenon tells us is that this information is irrelevant when a clinician forms a judgement about whether to request imaging. The referrer’s thought processes create a narrative image of representativeness (of a patient’s presentation and a likely diagnosis) which may be completely divorced from the statistical likelihood of that diagnosis – hence referrals with the irritating query: “please exclude…”. Similarly, colleagues who are informed of their imaging practice relative to peers are unlikely to weigh this information when making decisions about imaging a specific patient at the point of requesting. If our goal is behaviour change, the base-rate neglect phenomenon tells us it’s pointless to describe the base-rate, to use non-binding clinical decision rules describing the likelihood of a particular diagnosis or to spotlight systemic over-requesting relative to peers. This information simply will be neglected, often subconsciously. This is not how anyone, clinicians included, makes decisions.

Even for conscious thought processes there will always be reasons why an expert feels their judgement will outperform an algorithm (or a decision rule) despite evidence that they frequently do worse. Kahneman describes this as the illusion of skill though there is some debate about the extent of this illusion and about the added value of expertise. However, when perceptions of skill are intimately bound to doctors’ social role and idea of personal worth, it is singularly difficult for them to accept algorithmic decisions which undermine these perceptions and the utility of their subjective judgement.

What else might work? Binding decision rules (for example not being allowed to request an imaging test unless certain criteria are met) and strict clinical pathways can help though can be proscriptive and rarely result in less imaging.

It is here that the work of our other Nobel Laureate, Richard Thaler, might help. In his book ‘Nudge’ he describes how people can be encouraged to make better decisions by careful design of the systems within which those decisions are made: something he describes as ‘choice architecture’. A simple example is auto-enrolment in pension schemes: the design (architecture) of the choice offered (stay-in or opt-out) favours enrolment over an alternative way of presenting the choice such as asking people to enrol themselves (opt-in or opt-out). 

In radiology requesting, decision rules could default to a particular scan type in particular clinical scenarios; information could be presented about relative cost, complexity, patient discomfort or radiation dose; alternatives could be suggested including senior clinical review; imaging choices could be limited by requester seniority or prior imaging studies; duplicate requests could be flagged. None of this requires complex software logic development and much of this work has already been done (eg. The Royal College of Radiology iRefer resources) – the critical step is to embed these resources into the requesting system choice architecture at the point of imaging request. The referrer still has all options available to them but has to consciously decide to override a recommendation and consider the consequences of their choice.

Finally both Kahneman and Thaler emphasise the importance of feedback in affecting behaviour. In order to learn, feedback needs to be timely, personal and specific. That is why we learn quickly not to put our finger into a flame but find it difficult to reconcile our individual contribution to global warming. Although there are obvious difficulties with the lost art of imaging vetting by intimidation, sighing deeply while tearing a quaking junior doctor’s request card into small pieces certainly provided the opportunity for immediate feedback and learning, especially if accompanied by a patient explanation. Vetting undertaken remotely (spatially and temporally) means this feedback is diluted, making it much less likely that learning will occur and requesting behaviour will alter. If departmental processes and electronic systems can be designed to allow prompt feedback to the requester that an imaging request has been rejected (or at least needs more discussion) this is much more likely to shift behaviour and improve quality, even for an itinerant and ever changing requester workforce.

There are other ways in which radiology demand can be managed: imaging protocols (eg. follow up after cancer or surveillance imaging) can be revised, financial disincentives can be created to suppress imaging use, waiting lists can be reassessed and validated. Some of these methods are more acceptable clinically and ethically than others.

What behavioural psychology and decision science tell us is that in order to alter requesting behaviour and culture, nudges, feedback and narrative story are more likely to get results than generic exhortations to reduce imaging use or to consider base-rates and statistical probability.

There are simple wins and the IT systems facilitating these nudges and feedback need not be complex.