We calculated cumulative incidence of mortality over the 5-year follow-up period and compared across alcohol measures. We fit Cox proportional hazards models for time to death and estimated mortality hazard ratios with 95% confidence intervals (CI) first using AUDIT-C as the only alcohol metric, adjusted for age, sex, race/ethnicity, HIV, HCV, and viral suppression (these variables were identified a priori and are commonly associated with mortality). We used the lower risk drinking category (AUDIT-C = 1–3 for men/1–2 for women) as the reference group as this has been suggested to be an ideal comparison group for studying the effect of alcohol on health outcomes.41 We then refit the model using the combined AUDIT-C/PEth alcohol exposure categories, and used the likelihood ratio test to determine whether model fit improved. We further adjusted for smoking status, injection drug use in the past year, and having at least some college education.
