Medication adherence is a significant problem in public health. Prescription-level pharmacy databases have great potential for monitoring actual drug adherence patterns at the healthcare system level. Many research papers have reported adherence estimates in different settings and populations. However, comparison between studies is not always straightforward due to different approaches taken when computing adherence. A crucial component to accurately estimate adherence is the availability of days' supply information for each dispensing event. Reasonable assumptions regarding medication dosage have to be made, when this information is not available. In this study, we evaluate adherence and persistence to medication in patients after myocardial infarction and show that corresponding estimates differ significantly, when using different dosage assumptions, namely, when using defined daily dose or tablet per day dosage regimens. Moreover, we demonstrate that observed differences between medications might be a result of inaccurate dosage assumptions. We propose a comparison of distribution of days between dispensing events to that of days' supplied as a relatively simple visual inspection to validate dosage assumptions.
Keywords: adherence; electronic health records; medication dosage.