A comprehensive evaluation of survey questions for adherence to antiretroviral medications and exploratory analyses for identifying optimal sets of survey questions.

Although many methods for assessing adherence have been developed, most are not feasible for busy clinical settings. Using patients from the Adherence and Efficacy of Protease inhibitor Therapy (ADEPT) study (1998-2000), we systematically evaluated the relationship between psychosocial, environmental, clinical, and other factors with adherence to create composite variables (CVs) that are efficient with high sensitivity for detecting nonadherence and great potential for busy clinics. Eligible patients were protease inhibitor naïve or started a regimen within 3 months from baseline. Of the 128 patients who responded to survey at baseline, weeks 8, 24, and/or 48, mean (standard deviation [SD]) age was 39.3 (8.2) years with 81% male. About half of the patients were Latino, followed by 28% African American and 14% Caucasian. Sixteen percent reported injection drug use, and 40% had male-male sex. Mean CD4 count was 184.8 cells/mm(3) with a range from 1 to 1130 cells/mm(3). Thirty-two variables had a significant association with adherence at one or two time points and 9 were significantly associated with adherence over time. Among these significant factors, 8 also had a relationship with a clearly monotonic trend, by which 219 CVs were formed. Among these CVs, 8 were significantly associated with adherence and had a relationship with monotonic trend. Compared to traditional self-reported adherence, CVs had much higher sensitivities (p < 0.001) for detecting nonadherence. We conclude that CVs consisting of a combination of psychological, behavior, and adherence questions may be reasonable substitutes for direct adherence questions, which are limited by problems with recall and social biases. Trust in physicians, having a child, history of substance use, CD4 count, and belief that antiretrovirals can help living longer or improve quality of life can efficiently predict nonadherence. Because these variables are readily obtainable in clinical settings, these selected questions may provide a clinically useful means of screening patients for antiretroviral medication nonadherence.