5 Continuous outcome measures – conundrums and conversions contributing to clinical application objectives

Abstract

Objectives To discuss how continuous outcome measures are commonly used to measure patient-relevant outcomes, the obstacles in employing these measures clinically, and methods available to facilitate use of bodies of evidence comprised of continuous outcomes. Method Many patient-relevant outcomes, particularly quality of life measures such as pain or function, are routinely measured on a continuous scale. However, the interpretation of continuous outcomes can be difficult, particularly when considering application to clinical practice and shared decision-making. Making matters worse is the frequent existence of multiple scales for any given construct. Therefore, quantitative syntheses of literature must find a way to combine different scales into a ‘common language’, and the longest-standing and most frequently used method to do so is the standardized mean difference. Unfortunately, the standardized mean difference is even more difficult to interpret clinically. However, there are validated methods to make these measures easier to understand and apply clinically. This presentation explores these issues and offers a resource to help make these continuous measures more clinically useful. Results Different methods to amplify the clinical use of continuous outcome measures have been discussed for at least three decades now, some better known than others. These methods seek to estimate the proportion of patients expected to achieve a specified degree of benefit or harm based on the observed continuous outcome. Exploring the strengths, limitations, and judicious use of these methods facilitates a greater understanding of how continuous outcomes data might be usable for clinical and shared decision-making. Of the methods that have been validated for this purpose, approximated and observed estimates of the proportion of people who achieved a specified degree of change in the continuous outcome measure have shown considerable consistency. Conclusions Although not necessarily a panacea, methods facilitating conversion of continuous outcomes data to proportions of people expected to achieve a specified degree of benefit adds considerably to our ability to make sense and clinical use of continuous outcome measures. As such, when these methods are used appropriately, they add substantially to our evidence-based medicine and shared decision-making ‘toolkit’.