Pitfalls of medication adherence approximation through EHR and pharmacy records: Definitions, data and computation

Area of Science:

• Health Informatics
• Pharmaceutical Sciences
• Clinical Pharmacy

Background:

• Medication adherence is crucial for treatment success but challenging to measure accurately.
• Refill adherence, approximated through dispensation data and electronic health records (EHRs), is a common metric.
• Limited discussion exists on data quality and computational variations impacting adherence estimations.

Purpose of the Study:

• To evaluate the impact of common pitfalls in computing medication adherence using EHR data.
• To identify how data quality and operationalization issues affect adherence measures.
• To assess the reliability of EHR-derived adherence data.

Main Methods:

• Identified common pitfalls in EHR data and adherence measure operationalization.
• Defined refill adherence operationally and conducted experiments to test pitfall impacts.
• Utilized statistical significance testing against a baseline scenario to quantify effects.

Main Results:

• Minor definition changes substantially alter refill adherence estimates.
• Pickup patterns create significant discrepancies between adherence measures (e.g., proportion of days covered).
• Data issues had a small, statistically significant population-level impact but a considerable individual-level effect.

Conclusions:

• Data-related issues in real-world EHRs can significantly skew medication adherence values.
• Different operational definitions of refill adherence are affected uniquely by data issues.
• Inaccurate adherence estimations, particularly at the individual level, can lead to erroneous clinical conclusions.