Comparison of various measures for assessing medication refill adherence using prescription data

• Published in: Pharmacoepidemiology and drug safety Vol. 18; no. 2; pp. 159 - 165
• Main Authors: Vink, N. M., Klungel, O. H., Stolk, R. P., Denig, P.
• Format: Journal Article
• Language: English
• Published: Chichester, UK John Wiley & Sons, Ltd 01.02.2009
• Subjects: administrative data; Antihypertensive Agents - administration & dosage; Cohort Studies; diabetes mellitus type 2; Diabetes Mellitus, Type 2 - drug therapy; Drug Prescriptions - statistics & numerical data; drug utilization; Drug Utilization Review - statistics & numerical data; Humans; Hypoglycemic Agents - administration & dosage; Hypolipidemic Agents - administration & dosage; Netherlands; patient compliance; Patient Compliance - statistics & numerical data; primary health care; Netherlands
• ISSN: 1053-8569; 1099-1557
• DOI: 10.1002/pds.1698

Background Several measures using prescription data have been developed for estimating medication refill adherence. Few studies have made direct comparisons, and little is known about the accuracy of these measures in patients on a multiple‐drug regimen. Purpose To compare different calculation methods using prescription data for assessing refill adherence. Method An observational cohort study among type 2 diabetes patients was conducted. Adherence to oral glucose‐lowering, antihypertensive and lipid‐lowering medication was assessed for 2004. We calculated medication possession ratios in a flexible period (MPRF), per calendar year (MPRY) and gaps between refills (GAP) at drug class and therapeutic level. Individual review of drug prescription profiles was conducted to validate identified cases of suboptimal refill adherence. Differences in Area Under the Curve (AUC) of ROC‐curves were calculated to compare the methods. Results Of the 3877 patients, 2969 (77%) patients received oral glucose‐lowering medication, 2715 (70%) antihypertensives and 1797 (46%) lipid‐lowering medication. Using cutoffs for MPR < 80% and GAP > 30 days, overall rates of suboptimal adherence for these drug classes were 32, 35 and 23% respectively. AUC for measures calculated at drug class level (range 0.85–0.90) were significantly larger than those calculated at therapeutic level (0.72–0.90). For oral glucose‐regulating medication and antihypertensives, AUCs were largest for the MPRY and GAP measures (0.87–0.88). For lipid‐lowering medication, the AUC was largest for the GAP measure (0.90). Conclusions Differences between adherence measures were small and favoured calculation on drug class level. For multiple drug use, both MPRY and GAP were good measures for identifying suboptimal refill adherence. Copyright © 2008 John Wiley & Sons, Ltd.