Evaluation
Trials commonly include both outcome measures and process measures. Outcome measures indicate whether or not the intervention has worked. For example, the intended outcome of a deprescribing trial may be an improvement in quality of life or reduction in hospital readmissions. The outcome measure is used to inform the required sample size of a trial. The larger the difference that is anticipated between intervention and control, the smaller the required sample size to detect this difference with adequate precision. Researchers may choose a non-inferiority study; however, in most cases the required sample size prohibits such studies. In order to prove that the outcome from the intervention is ‘no worse’ than the control, the study requires a sample size that can detect a difference that is deemed negligible in clinical significance. This means that the study has to be powered to detect a very small difference unlike superiority studies when the research has to prove that the intervention is better than the control and choose a sample size accordingly.
Process measures represent intermediate steps to achieving an outcome. An example of a process measure to help explain the observed outcome is the number of medicines stopped. For example, in a trial powered to detect a difference in quality of life, if intervention participants have a significantly higher quality of life than control participants then we might assume that the intervention is effective. If, however, we find that the average number of medicines stopped is the same in both intervention and control arms, then we know that something other than stopping medicines is attributable to the difference.
Given that deprescribing is a complex behaviour with multiple determinants, effective interventions will also be complex as they need to comprise multiple components to address the target determinants. For example, the practitioner behaviour change intervention being tested in the CompreHensive geriAtRician-led MEdication Review (CHARMER) trial, aims to address five determinants of proactive deprescribing in hospital and thus comprises five intervention components. It is essential to evaluate the intervention’s mechanism of action in order to understand how the intervention leads to any changes in deprescribing activities and how this then impacts on the primary outcome measure. Understanding mechanisms of actions can be achieved by developing questionnaires or surveys (6). To understand how the CHARMER deprescribing intervention changes the behaviours of pharmacists and geriatricians, we developed a questionnaire to identify changes in the five target determinants. This will enable identification of any targeted determinants that the intervention does not address and help explain how the intervention works (or does not work) in changing proactive deprescribing.
The role of condition-orientated measures in evaluating trials is unclear. For some research funders, condition-orientated measures such as blood pressure or lipid profile are acceptable trial outcome measures whilst for other research funders, these are deemed process outcomes. This disparity in expectations generates variation in practice and therefore reduces opportunity for the results from deprescribing trials to be compared, and trial data to be aggregated to increase the power and thus precision of the estimated effect of deprescribing. A Core Outcome Set (COS) is an agreed, standardised set of outcomes to be measured and reported in all clinical trials of a particular health condition or specific area of healthcare. Research teams can measureadditional outcomes in their deprescribing trials, in addition to the COS, to suit their context and particular focus. A COS for hospital deprescribing trials for older people under the care of a geriatrician was recently reported (10). Consistent use of this COS by deprescribing trials will enhance uniformity of reporting; there is also a need for funders to align expectations on acceptable outcome measures.