Predictive validity of clinical AUDIT‐C alcohol screening scores and changes in scores for three objective alcohol‐related outcomes in a Veterans Affairs population

Aims To evaluate the association between AUDIT-C alcohol screening scores, collected as part of routine clinical care, and three outcomes in the following year (Aim 1), and the association between changes in AUDIT-C risk group at one year follow-up and the same outcomes in the subsequent year (Aim 2). Design Cohort study Setting Twenty-four US Veterans Affairs (VA) healthcare systems (2004-2007), before systematic implementation of brief intervention Participants 486,115 outpatients with AUDIT-Cs documented in their electronic health records (EHRs) on two occasions ≥12 months apart (“baseline” and “follow-up”) Measurements Independent measures were baseline AUDIT-C scores and change in standard AUDIT-C risk groups (no use, low risk use, and mild, moderate, severe misuse) from baseline to follow-up. Outcome measures were 1) high density lipoprotein cholesterol (HDL), 2) alcohol-related gastrointestinal hospitalizations (“GI hospitalizations”) and 3) physical trauma, each in the years after baseline and follow-up. Findings Baseline AUDIT-C scores had a positive association with outcomes in the following year. Across AUDIT-C scores 0 to 12, mean HDL ranged from 41.4 (95% Cl 41.3-41.5) to 53.5 (51.4-55.6) mg/L, and probabilities of GI hospitalizations from 0.49% (0.48-0.51%) to 1.8% (1.3-2.3%), and trauma from 3.0% (2.95-3.06%) to 6.0% (5.2-6.8%). At follow-up, patients who increased to moderate or severe alcohol misuse had consistently higher mean HDL and probabilities of subsequent GI hospitalizations or trauma compared with those who did not (p-values all < 0.05). For example, among those with baseline low-risk use, in those with persistent low-risk use vs. severe misuse at follow-up, the probabilities of subsequent trauma were 2.65% (2.54-2.75%) vs. 5.15% (3.86-6.45%), respectively. However, for patients who decreased to lower AUDIT-C risk groups at follow-up, findings were inconsistent across outcomes, with only mean HDL decreasing in most groups that decreased use (p-values all < 0.05). Conclusions When AUDIT-C screening is conducted in clinical settings, baseline AUDIT-C scores and score increases to moderate-severe alcohol misuse at follow-up screening appear to have predictive validity for HDL cholesterol, alcohol-related gastrointestinal hospitalizations, and physical trauma. Decreasing AUDIT-C scores collected in clinical settings appear to have predictive validity for only HDL.