Development of the medication adherence estimation and differentiation scale (MEDS) AthavaleAmod S. P. BentleyJohn BanahanBenjamin F. McCaffreyDavid J. PacePatrick F. VorhiesDouglas W. 2018 <p><b>Objectives:</b> To develop a self-reported measure for medication adherence and compare its ability to predict the proportion of days covered (PDC) with contemporary scales.</p> <p><b>Methods:</b> Retrospective prescription fill data from three community pharmacies in the Southeastern US were assessed to identify patients that were 18 years of age or older, and had received at least one medication for diabetes, hypertension, or dyslipidemia. A cross-sectional survey containing the Medication adherence Estimation and Differentiation Scale (MEDS) was administered among these pharmacy patrons. The MEDS assessed the extent and reasons for non-adherence. Survey responses were anonymously linked with retrospective prescription fill data. A total of 685 patients were sampled. The proportion of days covered (PDC) was used as the criterion measure. The Morisky, Green, and Levine Adherence Scale (1986 Morisky scale) and the Medication Adherence Reasons Scale (MAR-Scale) were used as comparators.</p> <p><b>Results:</b> The MEDS presented a five-factor solution—worries about side-effects, worries about addiction, worries about cost, lack of perceived need, and unintentional non-adherence (CFI = 0.97; RMSEA = 0.06; SRMR = 0.03; standardized factor loadings greater than 0.5, and statistically significant). The relationship between MEDS scores and PDC was statistically significant (unstandardized regression coefficient = –0.50, <i>p</i> < .01). The MEDS performed better than the 1986 Morisky scale (<i>R</i><sup>2</sup> = 0.02 vs 0.05, standardized regression coefficient = –0.13 vs –0.21) and the MAR-Scale (<i>R</i><sup>2</sup> = 0.02 vs 0.05, standardized regression coefficient = –0.12 vs –0.21) in predicting PDC.</p> <p><b>Conclusions:</b> The MEDS demonstrated good psychometric properties and performed better than the comparator scales in the prediction of PDC.</p>