Adjusting alcohol quantity for mean consumption and intoxication threshold improves prediction of nonadherence in HIV patients and HIV-negative controls

Background: Screening for hazardous drinking may fail to detect a substantial proportion of individuals harmed by alcohol. We investigated whether considering an individual’s usual drinking quantity or threshold for alcohol‐induced cognitive impairment improves the prediction of nonadherence with prescribed medications. Method: Cross‐sectional analysis of participants in a large, multi‐site cohort study. We used the timeline followback to reconstruct 30‐day retrospective drinking histories and the timeline followback modified for adherence to reconstruct 30‐day medication adherence histories among 3,152 individuals in the Veterans Aging Cohort Study, 1,529 HIV infected and 1,623 uninfected controls. We categorized daily alcohol consumption by using quantity alone, quantity after adjustment for the individual’s mean daily alcohol consumption, and self‐reported level of impairment corresponding to each quantity. A standard drink was defined as 14 g of ethanol. Nonadherence was defined as the proportion of days with ≥1 medication doses missed or taken ≥2 hours late, and clinically significant nonadherence was defined as ≥5% absolute increase in the proportion of days with nonadherence. Results: The mean adjusted‐ and impairment‐based methods showed greater discrimination of nonadherence risk compared to the measure based on quantity alone (quantity‐based categorization, 3.2‐fold increase; quantity adjusted for mean daily consumption, 4.6‐fold increase, impairment‐based categorization, 3.6‐fold increase). The individualized methods also detected greater numbers of days with clinically significant nonadherence associated with alcohol. Alcohol was associated with clinically significant nonadherence at a lower threshold for HIV infected versus uninfected patients (2 standard drinks vs. 4 standard drinks) using quantity‐based categorization, but this difference was no longer apparent when individualized methods were used. Conclusions: Tailoring screening questions to an individual’s usual level of alcohol consumption or threshold for impairment improves the ability to predict alcohol‐associated medication nonadherence.