2066. Estimating HIV Drug Adherence: Pharmacy Records and Patient Self-Reports
Allen L Gifford, MD, VA QUERI HIV Coordinating Center and Center for Research in Patient-Oriented Care, VA San Diego Healthcare System and University of California San Diego, Y Chen, VA QUERI HIV Coordinating Center and Center for Research in Patient-Oriented Care, VA San Diego Healthcare System and University of California San Diego, D Looney, VA QUERI HIV Coordinating Center and Center for Research in Patient-Oriented Care, VA San Diego Healthcare System and University of California San Diego, M Goetz, VA Greater Los Angeles Healthcare System and University of California Los Angeles, N Ampel, Southern Arizona VA Healthcare System and University of Arizona, M Holodniy, VA Palo Alto Healthcare System and Stanford University, S Bozzette, VA QUERI HIV Coordinating Center and Center for Research in Patient-Oriented Care, VA San Diego Healthcare System and University of California San Diego

Objectives: To evaluate the validity of pharmacy dispensing records adherence (PDR) and patient self-report questionnaires (SR) as measures of antiretroviral (ARV) medication adherence, using electronic monitoring (EM) as a validation standard.

Methods: We studied 130 HIV+ veterans taking multidrug ARV therapy at 4 VA centers. PDR adherence was calculated from computerized pharmacy medical records. Participants completed brief self-report adherence questionnaires. EM adherence to recommended number and timing of ARV doses were determined for matched time periods. Statistical correlations and Receiver Operating Characteristic (ROC) curves were assessed for diagnosing nonadherence at the <95% level.

Results: Patients were 99% male, 52% white, 28% African-American, 12% Latino, and 75% with some college. EM proportion of prescribed doses taken (mean 0.81; SD 0.23) was higher than the proportion taken on time (mean 0.70; SD 0.24). PDR was moderately and significantly correlated with EM number (r=0.27;p=0.01) and timing (r=0.37;p=0.0004) of doses taken; SR was also correlated with EM number (r=0.32;p=0.0003) and timing (r=0.33;p=0.0002). PDR <95% was 92% sensitive and 33% specific (area under ROC curve [AROC] 0.74) for diagnosing nonadherence, but SR <95% was 64% sensitive and 85% specific (AROC 0.78).

Conclusions: Veterans’ ARV adherence is well below the recommended 95% level, and adherence to ARV dose timing is lower. PDR and SR adherence both contribute to adherence measurement, with PDR adding sensitivity to clinical nonadherence detection.

Impact: Effective interventions for improving ARV adherence in VA HIV care are needed. PDR and SR adherence may be complementary clinical tools for identifying those with lower adherence.