How well do patients report noncompliance with antihypertensive medications? a comparison of self-report versus filled prescriptions

Purpose To address poor patient compliance with antihypertensives, clinicians and researchers need accurate measures of adherence with prescribed regimens. Although self-reports are often the only means available in routine practice, their accuracy and agreement with other data sources remain questionable. Methods A telephone survey was conducted on 200 hypertensive patients treated with a single antihypertensive agent in a large health maintenance organization (HMO) or a Veterans Affairs medical center (VAMC) to obtain self-reports of the frequency of missing antihypertensive therapy. We then analyzed records of all filled prescriptions to calculate the number of days that patients actually had antihypertensive medications available for use. Agreement between the two data sources was measured with correlation coefficients and kappa statistics. Logistic regression models were used to identify demographic, clinical and psychosocial correlates of overstating compliance. Results There was very poor agreement between self-reported compliance and days actually covered by filled prescriptions (Spearman correlation coefficient = 0.15; 95%CI: 0.01, 0.28). Very poor agreement was also observed between a categorical measure of self-reported compliance (ever vs. never missing a dose) and categories of actual compliance defined by filled prescriptions (< 80% vs > 80% of days covered; kappa= 0.12, 95%CI: -0.02, 0.26). Surprisingly, few factors were associated with inaccurate self-reporting in either crude or adjusted analyses; fewer visits to health care providers was significantly associated with overstating compliance. Conclusions Compliance was markedly overstated in this sample of patients and few characteristics identified those who reported more versus less accurately. Clinicians and researchers who rely on self-reports should be aware of these limits and should take steps to enhance their accuracy. Copyright (C) 2003 John Wiley Sons, Ltd.