Altiner 2007.

Methods	Study design: cluster‐randomised controlled trial Unit of randomisation: general practitioner (GP) Trial duration: November 2003 to March 2005 Recruitment: 2036 GPs from 9 regions in North‐Rhine and Westphalia‐Lippe, Germany, invited to participate (blinded to the primary outcome); of 239 GPs willing to participate and receiving baseline materials, 104 completed reliable baseline study documentation and were randomised (10 practice partners randomised as pairs) into intervention (GPs = 52, patients = 1389) and control groups (GPs = 52, patients = 1398) Methods of data collection: GPs recorded all consecutive and eligible patients during each documentation period on study specific paper documentation Data collection time points: 3 documentation intervals of 6 weeks each: baseline (before randomisation), and 6 weeks and 12 months post‐intervention Length of follow‐up: 12 months
Participants	GPs documented all consecutive and eligible patients: ≥ 16 years of age with an initial episode of acute cough (without prior episode < 8 weeks) and could comprehend German Exclusion: patients with underlying chronic lung diseases (e.g. asthma, chronic obstructive pulmonary disease), immune deficiency or malignant diseases
Interventions	Brief intervention name: complex, peer‐led, educational intervention Recipients: GPs and patients (passive) Providers: GP peers were trained to provide (in 3 sessions) the outreach visits in clinics during normal working hours (methods of training these GP peers were not specified) Health professional components: focused on antibiotic 'misunderstanding' during a consultation, and aimed to motivate GPs to change attitudes to communication and empower patients. Peers addressed GP beliefs and attitudes by exploring and evaluating GPs 'opposite' motivational background using a standardised dialogue script and communication techniques derived from the elaboration likelihood model. Aspects of the intervention were also informed by previous qualitative work Patients: waiting room poster and leaflet focusing on the patients' role within the antibiotic misunderstanding (e.g. GP perceptions that patients expect an antibiotic) and also brief evidence‐based information about acute cough and antibiotics to enable patients to raise and clarify issues and make a joint decision about antibiotic use with their doctor Materials: waiting room poster and leaflet (patient only); script used by GP peers Mode of delivery: face‐to‐face (GPs) and waiting room posters and leaflets (patients) Duration and intensity: 1 peer outreach visit per GP (duration not specified) Comparator: nil active comparator; GPs provided usual care
Outcomes	Primary: rate of antibiotic prescriptions per acute cough and by GP (study specific paper documentation) Secondary: nil
Notes	Funding: yes Conflict of interest: none disclosed Published trial protocol: no Trial registration: yes Ethics approval: yes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Program‐generated complete randomisation list
Allocation concealment (selection bias)	Low risk	Not described. However, GPs recruited prior to randomisation
Blinding of participants and personnel (performance bias) All outcomes	High risk	Not possible (complex peer‐led educational intervention)
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Participating GPs sent data to researchers. Each patient was assigned a unique identification number that could be connected with the patient only by the participating GP
Incomplete outcome data (attrition bias) All outcomes	High risk	Randomised: 104 GPs (intervention = 52, 1389 patients; control = 52, 1398 patients) 6 weeks post‐intervention: 86 GPs (intervention = 42 (80%), patients = 1021; control = 44 (84%), patients = 1143) 12 months post‐intervention: 61 GPs (intervention = 28 (54%); 787 patients; control = 33 (63%); 920 patients) 17% (18/104) dropped out at 6 weeks and 41% (43/104) by 12 months (reasons for GPs' exclusion from analysis: poor data quality or did not return data) Cluster‐level sensitivity analysis performed to explore effect of differential missing values
Selective reporting (reporting bias)	Low risk	All indicated results reported. Prospective trial registration: Projektdatenbank Versorgungsforschung NRW, ID: 90/34/CHANGE
Other bias	Unclear risk	Sample size (power) calculation: yes. Sample size calculated on number of patients to detect a 10% difference in 6‐month prescription rates (50% control, 40% intervention). Allowing for 20% drop‐out rate, it was estimated 200 GPs would be required to contribute 20 patients during each observation period (i.e. 4000 at each of the 3 documentation periods) ITT or per protocol analysis: no, all analysis (with exception of sensitivity analyses) included only general practices with complete follow‐up Large baseline difference found in antibiotic prescription rates between intervention and control groups (36.4% versus 54.7%) (unadjusted and adjusted analysis performed) GPs were not monitored during the trial period and may have under‐reported patients who received an antibiotic Government regulatory change during study to exclude OTC medicines from reimbursement by German statutory health insurance funds may have increased antibiotic prescribing decisions to minimise patient out‐of‐pocket cost Generalised estimating equation (GEE) models applied Intraclass correlation (coefficient): 0.20