This was a secondary analysis of datasets pooled from three ED-based randomized controlled studies of alcohol BIs. AUDIT-C was the primary outcome measure; secondary outcomes were binge drinking and achievement of NIAAA safe drinking levels. We conducted univariate comparisons and developed generalized linear models (GLM) for the primary outcome and generalized estimating equation (GEE) models for secondary outcomes to examine the intervention effect on the whole study group, gender-stratified subgroups, and gender- and violence-stratified subgroups.
Of 1219 participants enrolled, 30 % were female; 31 % of women and 42 % of men reported violence involvement at baseline. In univariate analysis, no differences in outcomes were found between intervention and control groups for any subgroup. However, in multivariable models, men demonstrated an intervention effect for likelihood of safe drinking limits. Stratifying further by violence, only men without violence involvement demonstrated a positive intervention effect for safe drinking limits. There was no evidence of an intervention effect on women.
Analyzing the overall effect of ED-based BI may mask its ability to improve alcohol-related outcomes in a subset of the population. Alternatively, interventions may need to be significantly improved in subsets of the ED population, e.g., in women and in men with involvement in violence.