Table 2.

Summary of the five types of medication adherence measure: target population(s), advantages, and disadvantages.

Measures	Target population(s)	For primary/secondary nonadherence	Advantages	Disadvantages
Direct measures	Patients under single-dose therapy and intermittent administration and who are hospitalized	Both primary and secondary nonadherence	Most accurate Can provide physical evidence	Generate a Yes/No result only Intrusive Varied drug metabolism Nonquantifiable biomarkers/drug metabolites Drug-drug interactions and drug-food interactions Expensive Require qualified staff and techniques to perform Bias occurs if patients know the schedule of the tests (white coat adherence)
Measures involving secondary database analysis	Countries that allow refilling prescription; with centralized computerized system with a consistency among prescribers and dispensers; more common for research with a large population	Primary nonadherence	Able to assess multidrug adherence Can identify patients at risk for treatment failure Provide medication-refilling pattern Complete dataset used are generally verified by a third party for insurance claim purpose	Assumptions are made (the medication-taking behavior corresponds to prescription refilling and the medications are taken according to prescription) Fail to identify partial adherence Fail to identify barriers for the detected nonadherence Missing out prescriptions, if obtained outside the system Incomplete records, if drug discontinuation is verbally advised by prescriber
Measures involving Electronic Medication Packaging (EMP) devices	Studies with small population As reference standard to validate other measures	Secondary nonadherence	Highly accurate Identify medication-taking pattern Identify partial adherence	Expensive Technical supports required Overestimation if patients accidently or purposefully actuate the container Inconvenience due to bulky container Pressure to patients
Pill count	Routine clinical practice	Primary nonadherence	Low cost Simple Can be used in various formulations Highly accurate	Not for nondiscrete dosages or prn medications Underestimation due to early refill Arbitrary cutoff value Unable to identify medication-taking pattern
Measures involving clinician assessments and self-report	Routine clinical practice Less suitable for research	It depends on the type of assessments and questionnaires used	Low cost Easy to administer Real-time feedback Available Flexible to accommodate different conditions Identify belief and barriers to adherence Well-validated	Least reliable Relatively poor sensitivity and specificity Affected by communication skills of interviewers and questions in the questionnaire Patient's desirability can bias