184601 A novel approach to assessing medication adherence using clinical databases

Tuesday, October 28, 2008: 5:00 PM

Donald Robinaugh, MA , Research & Development Service, VA New York Harbor Healthcare System, New York Campus, New York, NY
Michelle Ulmer, BA , Research and Development Service, Department of Veterans Affairs New York Harbor Healthcare System & New York University, New York, NY
Marilena Antonopoulos, PharmD , Long Island University, Brooklyn, NY
Stuart R. Lipsitz, ScD , Brigham and Women's Hospital/Harvard Medical School, Boston, MA
Sundar Natarajan, MD, MSc , Research and Development Service, Department of Veterans Affairs New York Harbor Healthcare System & New York University, New York, NY
Background: Medication non-adherence is a significant public health issue, contributing to poor control of treatable conditions such as hypertension. Commonly used methods to identify non-adherence suffer from a variety of shortcomings, including feasibility, methodological concerns, subjectivity and intrusiveness. Refill compliance (RC) is an objective and reliable measurement based on pharmacy records, most notably in settings where patients receive medications from a single source, such as the VA. Many current RC measures, however, are only accurate in large samples over long periods of time. We developed a refill compliance measure using data collected from VA databases that accounts for a variety of medication, patient and pharmacy characteristics and compared this measure with the Morisky scale (MTS), a commonly used self-report assessment of adherence.

Methods: Refill information was collected individually for each patient using the Computerized Patient Record System. Doses were considered to be missed (Gap) if the number of days between two refill dates (Days Passed) exceeded the amount of medication prescribed (Supply). The summed Gaps over the observed period divided by the total days in that period represents non-compliance for that medication. Additionally, the algorithm accounts for overstocking, discontinuations, dosage changes, added medications, delivery type, inpatient hospitalizations and other factors affecting the core variables of Gap, Supply and Days Passed. Adherence was classified as full (>80%), partial (50-80%) or non (<50%) using RC. Pearson correlations and Mantel-Haenszel Chi-square analyses compared RC and MTS as continuous and categorical variables respectively. Using RC as the reference measurement, we used Chi-square analysis to assess MTS sensitivity to non-adherence.

Results: We evaluated 265 veterans with uncontrolled hypertension. The mean observed RC was 81.7% with 58.1% of the sample showing full adherence. MTS was positively associated with RC (r=.13, p<.001), and had a similar proportion of patients identified as adherent (55.6%). Chi-square analyses, however, showed significant differences existed between the RC and MTS in their classification of partial-adherence (25.2% v 39.5% respectively) and non-adherence (16.9% v 5.26%). Finally, the MTS showed extremely poor (6.6%) sensitivity to non-adherence, with only 3 of the 45 non-adherent participants identified as such by the MTS.

Discussion: The poor sensitivity of the MTS demonstrates the need for objective measurements of adherence in order to identify and address medication non-adherence. This novel algorithm for calculating medication adherence from VA databases allows for robust and reliable measurement of medication adherence for individual patients over short periods of time.

Learning Objectives:
Understand the advantages and disadvantages of various methods of measuring medication adherence. Recognize the usefulness of pharmaceutical databases in unobtrusively measuring adherence. Learn about the development and use of an objective measure useful for prospectively measuring adherence.

Keywords: Treatment Adherence, Veterans' Health

Presenting author's disclosure statement:

Not Answered