Exploring adherence to antidiabetic medications in Singapore primary care: a comparison of four models of proportion of days covered.

Abstract

Introduction: There is currently no standardised approach to calculate and report proportion of days covered (PDC), a medication adherence measure. We aimed to assess adherence to antidiabetic medications by applying four PDC models to a primary care database and examine the factors associated with medication adherence.

Methods: Four models were used to calculate PDC for 789 patients with diabetes mellitus (DM) using the average PDC method. Models P1 and P2 incorporated prescribed and dispensed data, whereas models D1 and D2 used dispensed data only. Models P1 and D1 used an interval-based method, whereas models P2 and D2 used a prescription-based method. Gender, age at recruitment, race, number of chronic diseases, years of DM, glycated haemoglobin (HbA1c) levels and number of antidiabetic medication classes were tested in a univariate analysis. Stepwise selection method was used in the multivariate logistic regression model.

Results: The proportion of adherent patients (PDC ≥80%) was 64.1% for model P1, 73.9% for P2, 66.5% for D1, and 87.3% for D2. Patients with PDC <80% were more likely to have HbA1c ≥9% (odds ratios 2.54 [P1], 2.69 [P2], 2.48 [D1], and 3.33 [D2]). Additionally, PDC <80% was associated with Malay or Indian ethnicity and having four or more chronic diseases.

Conclusion: The PDC models that incorporate prescribed data and use interval-based methods may result in more patients being classified as having poor adherence. Compared to the other models, we postulate that model P2 may provide the most accurate estimate of adherence, as it takes into account the prescribers' intent by including prescribed data and changes in medication regimens by using prescription-based method.