Alcohol use disorders (AUDs) are highly prevalent in Europe, but only a minority of those affected receive treatment. It is therefore important to identify factors that predict treatment in order to reframe strategies aimed at improving treatment rates.

Representative cross-sectional study with patients aged 18–64 from primary health care (PC, six European countries, n = 8476, data collection 01/13–01/14) and from specialized health care (SC, eight European countries, n = 1762, data collection 01/13–03/14). For descriptive purposes, six groups were distinguished, based on type of DSM-IV AUD and treatment setting. Treatment status (yes/no) for any treatment (model 1), and for SC treatment (model 2) were main outcome measures in logistic regression models.

AUDs were prevalent in PC (12-month prevalence: 11.8%, 95% confidence interval (CI): 11.2–12.5%), with 17.6% receiving current treatment (95%CI: 15.3–19.9%). There were clear differences between the six groups regarding key variables from all five predictor domains. Prediction of any treatment (model 1) or SC treatment (model 2) was successful with high overall accuracy (both models: 95%), sufficient sensitivity (model 1: 79%/model 2: 76%) and high specificity (both models: 98%). The most predictive single variables were daily drinking level, anxiety, severity of mental distress, and number of inpatient nights during the last 6 months.

Variables from four domains were highly predictive in identifying treatment for AUD, with SC treatment groups showing very high levels of social disintegration, drinking, comorbidity and functional losses. Earlier intervention and formal treatment for AUD in PC should be implemented to reduce these high levels of adverse outcomes.