Henihan 2016.

Methods	Study design: cluster‐RCT. Recruitment modality of participants: 16 general practitioners selected by random stratified sampling (by location and level of methadone provision training), each recruited 10 consecutive patients receiving addiction care, including methadone.
Participants	Number of participants: 81 participants were included in the trial (34 in the intervention group, 47 in the control group). Only participants with positive AUDIT at baseline (n = 30) were included in this systematic review. Gender: 61.7% male Mean age (mean ± SD): 42 ± 8.5 years Condition: participants were > 18 years old, receiving primary care addiction treatment including methadone (80/81 patients). For the purpose of this review, only patients who were AUDIT‐positive at baseline were included. The threshold for positive scores was 8 or more points. Other relevant information Participants were excluded if they were age < 18 years, acutely intoxicated, cognitively impaired including severe mental illness, or had language difficulties.
Interventions	(1) BI (n = 13): Physicians randomised to the intervention group screened participants using the AUDIT‐C and provided a brief intervention to patients who were positive for 'hazardous' or 'harmful' alcohol use. Delivery of the intervention was confirmed by interviews with physicians, patients, and chart reviews. Cointeventions: GPs received a complex intervention to promote screening and brief intervention which included practice visits, best practice guidelines and education, multimedia educational tools, MI‐related training presentations, and demonstration of interventions. (2) TAU (n = 17): Physicians randomised to the control group were given training at the end of the trial (3 months). Duration of follow‐up: 3 months Country of origin, setting: 16 general practices in Ireland (Health Service Executive Mid‐West and Dublin Mid‐Leinster regions)
Outcomes	Alcohol use as AUDIT scores at 3 months Retention at 3 months Feasibility of the study was assessed by physician and patient retention. Acceptability of the intervention was assessed by a physician Short Alcohol and Alcohol Problems Perception Questionnaire, as well as by qualitative interviews of both physicians and patients. Efficacy among patients was assessed by an AUDIT‐C and AUDIT assessment at baseline and 3 months, as well as rates of patients self‐reporting screening, brief interventions and referral to treatment at 3 months after intervention.
Notes	The BI was part of a complex implementation strategy to increase the uptake of BIs by general practitioners. Actual delivery of the intervention by GPs and fidelity of the intervention were not tested. The study was funded by Health Research Board Ireland; no conflicts of interest were reported. The trial analysis calculated an intra cluster correlation coefficient (ICC) for care process and outcome measures. The ICC for the proportion of patients with positive AUDIT‐C (follow up) results was 0.11 (standard error [SE] = 0.013). The ICCs for screening, BI and referral to treatment were 0.016 (SE 0.014), –0.06 (SE 0.017), and 0.22 (SE 0.026), respectively.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Personal communication: “external statistician”.
Allocation concealment (selection bias)	Low risk	"External statistician emailed anonymous sequence of numbers to researcher who matched them with the unique study identifiers of participants.”
Blinding of outcome assessment (detection bias) Subjective outcomes	High risk	Blinding of participants and personnel not possible for the kind of intervention. “Outcome assessors not blinded.”
Incomplete outcome data (attrition bias) End of Study outcomes	Low risk	"Baseline: 1 GP dropped out." Random allocation: "dropouts: two intervention GP practices and 11 patients from these practices."
Incomplete outcome data (attrition bias) Follow up	Low risk	Follow up: "14 patient dropouts (intervention (n=7) and Control (n=7))” information on drop out from the study provided; Missing outcome data balanced in numbers across intervention groups, with no reasons for missing data provided;
Selective reporting (reporting bias)	Low risk