Recommendation #4: Prioritize high-risk groups
In many cases, deprescribing can be beneficial but has only small to
moderate effects on outcomes of interest. For example, deprescribing
fall-risk increasing drugs (FRIDs) may reduce the risk of falls – yet
many other factors influence fall risk, so intervening solely on
medications will only go so far. In these settings, testing
deprescribing interventions in a population whose baseline risk of the
outcome is low almost guarantees that the study will be underpowered to
detect small to moderate (but still meaningful) effects.
We should thus test our interventions in populations who have high rates
of the outcome we seek to reduce, making it easier to detect a
beneficial effect for a given sample size. Consider an intervention that
reduces rates of a harmful outcome by one-third (i.e., relative risk
0.67). If the baseline outcome rate among participants in a controlled
clinical trial is 10%, a sample size of more than 2300 people would be
required to detect the expected reduction from 10% to 6.7% (at P value
<0.05 and power of 0.80). In contrast, if 50% of the
population has the outcome at baseline, fewer than 300 study subjects
would be required to detect the expected reduction in outcome rates.