Patient- and provider-reported adherence: toward a clinically useful approach to measuring antiretroviral adherence

We seek to develop a clinically useful measure of antiretroviral medication adherence. Because there is no gold standard for adherence, we will assess the clinical validity of patient- and provider-reported adherence by the strength of their expected associations with current viral load, depressive symptoms, alcohol and illicit drug use, and homelessness. The Veterans Aging Cohort 3 Site Study (VACS 3) is a multisite study of 881 patients at Cleveland, Houston, and Manhattan Veterans Affairs health care systems. Data was collected on adherence using patient report and provider assessment; depressive symptoms using the Center for Epidemiological Studies Depression (CESD) and provider assessment; alcohol use using the Alcohol Use Disorders Identification Test (AUDIT) and provider assessment; and homelessness using patient report only. Viral load was collected from electronic laboratory data. Although agreement between providers and patients about the patient's adherence was not better than chance (61%; weighted kappa =.07), both patient and provider-reported adherence were related to viral load (P <.001), current alcohol use (P <.01), current drug use (P <.01), and depressive symptoms (P <.001). Patient-reported adherence was also associated with homelessness (P <.05). In multivariate regression models, provider assessment of adherence demonstrated independent associations with viral load (P <.001), current alcohol use (P <.001), current drug use (P <.001), and depressive symptoms (P <.001) after adjustment for the patient's report of adherence (also significantly associated). The consistent and largely independent association between patient and provider reported adherence and a range of variables previously shown to be associated with adherence suggests that patient- and provider-reported adherence independently measure actual adherence. Future work will explore how patient- and provider-reported adherence might best be combined, and whether the measure may be further enhanced with pharmacy refill data.