Hazards of progression to severe AUD (≥6 criteria) were estimated for individuals who, at any previous time point, had endorsed a single criterion or met criteria for mild, moderate, or mild-to-moderate AUD via Cox proportional hazards survival analyses conducted using the survival (version 3.4-0)34 and adjustedCurves (version 0.9.1)35 packages in R. Low-risk (n = 699) and high-risk (n = 317) mild-to-moderate AUD subgroups were defined by a prior mild-to-moderate AUD diagnosis and differentiated by endorsement of 1 or more high-risk criteria identified in the cross-sectional IRT analysis. Additional survival analyses were conducted examining hazards of progression to severe AUD based on alcohol involvement milestones, MDD, ASPD, and other SUD diagnoses. Covariates for all models included sex, race and ethnicity, and criterion count to examine associations independent of individual differences in the number of prior criteria endorsed. A family grouping variable was used to estimate robust standard errors to account for familial clustering.36 Violations of the proportional hazards assumption for predictor variables were tested using Schoenfeld residuals and resolved by including an interaction term with age at onset of severe AUD. Statistical analyses were conducted from December 2022 to June 2023.
