Ann Pharmacother 2009 Mar 3;43(3):413-22. Epub 2009 Mar 3.
Division of Pharmaceutical Outcomes and Policy, Eshelman School of Pharmacy, University of North Carolina at Chapel Hill, Campus Box 7360, 2205 Kerr Hall, Chapel Hill, NC 27599, USA.
Background: Medication adherence is suboptimal, and clinicians and researchers struggle with identifying nonadherent patients. Various measures of medication adherence exist, but there is controversy regarding which measures provide acceptable data and how nonadherence should be defined.
Objective: To assess agreement among patient self-report, pharmacy refill, and electronic adherence measures and compare the sensitivity and specificity of different cut-points for defining nonadherence.
Methods: Data were analyzed from 2 similarly designed randomized controlled trials that assessed a pharmacist's intervention to improve medication adherence among patients with hypertension or heart failure. For each participant, adherence was measured by patient self-report, prescription refill records, and electronic lids on medication containers. Agreement among measures was assessed using Spearman's correlation coefficient rho. Correlation coefficients were compared by patient characteristics using Fisher's Z transformation. The sensitivity and specificity of different cut-points for defining nonadherence were calculated.
Results: Median adherence was 84% for self-report, 86% for electronic, and 91% for prescription refill adherence measurement. Refill and electronic adherence demonstrated the best agreement among measures (rho = 0.48). Age, depression, and other comorbid conditions influenced agreement among measures. Measures were generally in agreement, regardless of how nonadherence was defined. A cut-point of 80% illustrated a fair balance between sensitivity and specificity for all measures.
Conclusions: All measures provided similar estimates of overall adherence, although refill and electronic measures were in highest agreement. In selection of a measure, practitioners should consider population and disease characteristics, since measurement agreement could be influenced by these and other factors. The commonly used, clinically based cut-point of 80% had a reasonable balance between sensitivity and specificity in studies of adherence in patients with heart failure or hypertension.