Clinical assessment of pain and its measurement and reporting for research: A state of discomfort

For the better part of a century we have been attempting to measure pain for both clinical and research purposes. ¹ While various assessment tools have been shown to be reasonably reliable in experimental models of pain, where the stimulus and the subjects can be relatively well controlled, measurement of pathological pain is more problematic, especially in the acute setting. Myriad factors, internal and external, influence both the experience and the reporting of pain. And as discussed by Miu and colleagues in this issue of the journal, there is considerable evidence that the very act of trying to assess pain can alter that experience for the patient. ²

There are profound and fundamental problems associated with the pain assessment tools in common use today. This is not surprising, given that most of these tools represent an attempt to reduce a multidimensional experience to a coarse unidimensional measure. Awareness of these problems is also not novel. It was recognized 50 years ago, for example, that the addition of descriptors along the length of a visual analog scale (VAS) resulted in the majority of subjects marking the scale at the point of the descriptors, rendering it no different to or better than a simple verbal descriptive tool. ¹ Orientation can be important, especially at extremes of age, with some people having difficulty assigning left, right, up or down to the more painful end of the scale. ³ It was also recognized long ago that around a tenth or more of subjects fail to adequately comprehend the use of a VAS at the first attempt. ¹ It is perhaps for these reasons that various modified forms of the VAS have found their way into widespread use: slide devices illustrated with faces; scales that include interval marks and numbers; scales that include various descriptors; scales with grayscale or rainbow backgrounds. Mostly these modifications reduce the tool to what is, probably at best, an ordinal scale. It may be visual, but it is not analog. The inclusion of centimetre marks with or without numbers along a 100 mm line, for example, effectively reduces what might loosely be considered ‘analog’ to something with essentially no more resolution than three-bit digital.

The misunderstanding, misuse and misreporting of pain assessment runs deep, both in research and the clinical setting. It is all too common to see, for example say, a ‘30% reduction’ in pain reported for an intervention that results in a change in mean verbal numerical rating scores (VNRS) from, say, 5 to 2. This is problematic for several reasons. While there have been arguments for and against the linearity of a properly administered VAS,^4–8 there is no adequate evidence that a VNRS can be treated as linear, nor that any given score represents the same thing from one individual to another or indeed from one day to another even for any given individual, especially in an acute setting. The use of a mean to summarize VNRS data is therefore likely to be inappropriate, as would be the application of parametric statistics to those data, although this is perhaps more controversial if the samples are very large. For the same reason, it is not appropriate to quantify a percent reduction in pain from such a study. Given the multidimensional and very individual nature of the pain experience, the concept of quantifying any alteration in pain as a percentage is arguably absurd.

Another pervasive problem is the incorrect use of pain terminology in scientific communication, the most common being use of the term ‘visual analog scale’ or ‘VAS’ to describe assessment tools that are no such thing. Several papers a week are submitted to this journal reporting what is, without question, a VNRS (score from 0–10, usually), but which is instead incorrectly described as a ‘VAS’. Unfortunately this loose and incorrect use of terminology starts at the top, with examples to be found in the guidelines and teaching materials of many peak bodies including our own. ^{9 , 10} We encourage authors to ensure their submissions contain accurate descriptions of the pain assessment tools used in their research, including what measures were taken, or not, to ensure the reliable administration of those assessments.

Clinically, we have also been witness over the last 20 years to an iatrogenic disaster, with the well-intentioned but rather predictably misguided use of ‘pain as the 5th Vital Sign’ contributing significantly to what is now widely known as ‘the opioid crisis’. ¹¹ There are other factors involved, so the true extent of the harm specifically caused by this initiative is, like pain itself, rather difficult to quantify, but there is no question of a substantial contribution. ¹² This realization has now led, allegedly, to the abandonment of the 5th Vital Sign and numerical pain scores in some parts of the world, ¹¹ but their use remains prevalent in many others.

In the light of these limitations, is there a place for the numerical scoring assessment of ‘comfort’ rather than pain? While Miu and colleagues present some evidence in support of the concept that pain scores and ‘inverted’ comfort scores may be moderately correlated, they also provide rather convincing evidence of the lack of any reliable utility for either of them. ² From their figure, it can be seen that at least one subject scored 8/10 for pain (commonly considered severe) yet still scored 8/10 for comfort, while another scored 3/10 for pain (commonly considered mild) yet also scored only 2/10 for comfort. There were some patients with, apparently, no pain at all yet they reported being moderately uncomfortable according to their scores. And even if we did replace pain scores with some form of comfort score, we would also still be left using yet another unidimensional measure with the same limitations, along with the same potential consequences.