Initial medication adherence-review and recommendations for good practices in outcomes research: an ISPOR medication adherence and persistence special interest group report

Background: Positive associations between medication adherence and beneficial outcomes primarily come from studying filling/consumption behaviors after therapy initiation. Few studies have focused on what happens before initiation, the point from prescribing to dispensing of an initial prescription. Objective: Our objective was to provide guidance and encourage high-quality research on the relationship between beneficial outcomes and initial medication adherence (IMA), the rate initially prescribed medication is dispensed. Methods: Using generic adherence terms, an international research panel identified IMA publications from 1966 to 2014. Their data sources were classified as to whether the primary source reflected the perspective of a prescriber, patient, or pharmacist or a combined perspective. Terminology and methodological differences were documented among core (essential elements of presented and unpresented prescribing events and claimed and unclaimed dispensing events regardless of setting), supplemental (refined for accuracy), and contextual (setting-specific) design parameters. Recommendations were made to encourage and guide future research. Results: The 45 IMA studies identified used multiple terms for IMA and operationalized measurements differently. Primary data sources reflecting a prescriber's and pharmacist's perspective potentially misclassified core parameters more often with shorter/nonexistent pre- and postperiods (1-14 days) than did a combined perspective. Only a few studies addressed supplemental issues, and minimal contextual information was provided. Conclusions: General recommendations are to use IMA as the standard nomenclature, rigorously identify all data sources, and delineate all design parameters. Specific methodological recommendations include providing convincing evidence that initial prescribing and dispensing events are identified, supplemental parameters incorporating perspective and substitution biases are addressed, and contextual parameters are included.