Background: Adherence assessment is an essential component of monitoring HIV antiretroviral therapy. Prior studies suggest that medical providers frequently estimate individual patient adherence inaccurately.
Objective: We compared provider estimates of nonadherence to antiretroviral therapy with unannounced pill counts and structured patient interviews to determine the accuracy of adherence information obtained by providers and patients.
Design, setting, and participants: Comparison of three adherence measures in homeless or marginally housed persons receiving HIV antiretroviral therapy (n = 45) and their providers (n = 35).
Measurements: Provider estimate of percentage of pills taken; three successive patient structured reports of number of doses missed in the last 3 days; and three successive unannounced pill counts.
Results: 13% (95% confidence interval [CI], 4%-22%) of patients were not following their regimen as directed. Provider-adherence estimate explained only 26% (95% CI, 6%-47%) of the variation in pill count adherence, whereas patient report explained 72% (95% CI, 52%-96%). The sensitivity and specificity of provider estimates of nonadherence, defined as <80% of pills taken by pill count, were 40% and 85%, respectively. The sensitivity and specificity of patient interview were 72% and 95%, respectively.
Conclusions: Provider estimate of adherence was inaccurate whereas structured patient report was more closely related to pill count. Structured assessment over several short intervals may improve accuracy of adherence assessment in clinical practice.