Differences in common medication adherence measurements illustrated through a diabetes outcomes study

Abstract

Background: Medication Adherence is a key aspect of type 2 diabetes (T2D) patient education and self-management counseling. Adherence measures based on pharmacy dispensing data and patient surveys are commonly used, but clinicians and researchers should be aware of the strengths and limitations of these indirect measurement approaches. Methods: Data from a T2D treatment outcomes study will be discussed to illustrate measurement approaches, behaviors the measures reflect, and approach limitations. Diabetes medication adherence was quantified in 166 adult patients using the modified Medication Possession Ratio (mMPR) and the 5-item Medication Adherence Reporting Scale (MARS-5). Correlation between measures was assessed using a Kappa coefficient. A structural equation model (SEM) assessed the associations between patient characteristics, diabetes medication class, adherence, and treatment outcomes. Findings: Diabetes medication adherence was 77.1% per mMPR and 72.3% per MARS-5. Only 58.4% were classified as adherent per both measurements; correlation was weak (kappa coefficient 0.142). Self-reported adherence was associated with weight loss. Both measurements were associated with improved glycemic control. No association was found between medication beliefs or diabetes drug class and adherence. Discussion: Low correlation between adherence measures was anticipated. Self-reported adherence represents patient medication taking behavior and intentions but is subject to reporting bias. Claims-based measures objectively report medication purchasing behavior, but do not reveal information on actual consumption. Thus, neither approach is considered a gold standard. Both are predictive of diabetes outcomes and could be used to target patients for adherence counseling that addresses the behavior reflected by the measurement approach.