Commentary

Asking patients directly but non-judgmentally about their medication taking is the most useful way to assess compliance in clinical practice.¹ Over the past 30 years, studies using different questionnaires and different methods of validation have consistently shown that the sensitivity of patient self reports of non-compliance is low, but the positive predictive value of such reports is high. In other words, patient reports of non-compliance are generally accurate, but many patients who are not compliant will not say so. The findings of de Klerk et al confirm this observation.

The use of electronic medication monitoring, the most accurate single measure of compliance, is a strength of their study, but the 12 minutes required to complete the CQR and the need for weighting of responses to enhance its predictive value are barriers to its clinical use. The finding that 33% of those offered the CQR did not provide analysable responses raises a concern about potential bias, since people non-compliant with surveys about their compliance may also be non-compliant with their pills. Additionally, the data were collected within 2 weeks of initiating a new rheumatological drug, so that these patient responses may be most predictive of their early decisions about whether to continue the drugs rather than their long term persistence with treatment.

Clinicians and researchers should remember the adage that “accurate measurements of compliance are not easy; easy measurement of compliance is not accurate.”² In the search for a clinically useful yet accurate compliance measure, de Klerk et al have validated a survey that identifies a subset of patients who do not take their medications as prescribed. Their finding that 4 of the 19 questions in the survey explain most of the variance in compliance raises the hope that a “short form” of their measure might be more readily incorporated into rheumatology practice (see web extra).