Freyer‐Adam 2008.

Methods	Quasi‐RCT.
Participants	595 general hospital patients. 25% were alcohol abusers, 57% at ‐risk drinkers, and 18% heavy episodic drinkers. Germany.
Interventions	1. MI by liaison service (n= 249) 2. MI by hospital physicians (n= 121) 3.TAU (n= 225).
Outcomes	Physiological primary: None. Non‐physiological primary: Gram alcohol per day, gram alcohol past week. Secondary: Readiness to change drinking.
Notes	It was not possible to monitor the fidelity of the intervention in physician arm.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	"...randomisation was conducted by time‐frame, based on the date of admission."
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement.
Blinding (performance bias and detection bias) Patients and providers	High risk	No blinding.
Blinding (performance bias and detection bias) Assessors	High risk	"...the staff was not blind to the study group to which the participants had been assigned."
Incomplete outcome data (attrition bias) All outcomes	High risk	29% lost to follow‐up at 12 months, not balanced, reasons provided. Not ITT.
Selective reporting (reporting bias)	Low risk	The published report included all expected outcomes based on the study hypotheses.
Other bias	High risk	Only self‐reported outcomes. Because staff became more experienced over time, they might have recruited different patients in the first period when they recruited the controls than in the later period when they recruited to the intervention groups. There were differences between groups at baseline on satisfaction with health, age, and having an intimate partner.