O'Connor 2009b.

Study characteristics
Methods	Simulated physician learning intervention to improve safety and quality of diabetes care: a randomized trial Cluster‐RCT, 1) The study was conducted at HealthPartners Medical Group, an 18‐clinic multispecialty group that provides care to 8000 adults with diabetes. 2) Primary care physicians were randomised to a simulated case‐based physician learning intervention (software developed by the authors) with a) printed feedback comparing actions taken with the ones taken by an expert physician or b) verbal interaction and feedback from a physician opinion leader who observed the physician Three arms: 1. Control (no intervention ‐ group A), 2. Intervention 1 (simulated physician education and printed feedback ‐ group B) and 3. Intervention 2 (simulated physician education and verbal feedback ‐ group C)
Participants	Control arm N: 691 Intervention arms N: 725, 604, NA Diabetes type: unclear/not reported Mean age: 64.0 ± 13.0 % Male: 58 Longest follow‐up: 12 months
Interventions	Control arm: (no intervention ‐ group A) Intervention arm:  (simulated physician education and printed feedback ‐ group B) 1) Audit and feedback 2) Clinician education Intervention arm:  (simulated physician education and verbal feedback ‐ group C) 1) Audit and feedback 2) Clinician education
Outcomes	1) Glycated haemoglobin 2) Low‐density lipoprotein
Funding source	This project was supported by the Agency for Healthcare Research and Quality (grant no. RO1 HS 10639)
Notes	—
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported. Before randomisation, 67 consenting physicians were blocked into groups of 3 based on 1) same specialty (family medicine or internal medicine) and 2) whether they provided care to 50 vs 50 or more diabetic patients.
Allocation concealment (selection bias)	Low risk	Clustered RCT.
Provider's baseline characteristics (selection bias)	Low risk	Table 1. All P values above 0.05 for physicians, but group C seems to have fewer female physicians (16% vs 37% in group A and 26% in group B).
Patient's baseline characteristics (selection bias)	High risk	Table 1. Randomisation at the physician level resulted in similar patient samples except that patients of physicians in group B more often had coronary artery disease and higher Charlson scores (Charlson comorbidity index).
Patient's baseline outcomes (selection bias)	Low risk	Table 1. P values above 0.05 for HbA1c and LDL (pre‐intervention).
Incomplete outcome data (attrition bias)	High risk	Table 2. For HbA1c, they have data for 1686 patients out of 2020 at 12 months (16.5% loss) and for LDL, they have data for 1178 patients (41.7% loss). High numbers. Attrition occurred evenly across randomised groups, and final analysis included 19 physicians in each group.
Blinding of participants and personnel (performance bias) and of outcome assessors (detection bias)	Low risk	Our outcomes of interest were objectively measured (HbA1c and LDL).
Selective reporting (reporting bias)	Unclear risk	Retrospectively registered protocol (protocol posted on December 2005, study started on December 2001 and was completed on September 2002). The secondary outcomes do not match between protocol (secondary analysis will assess rates of screening for microvascular complications) and paper (pharmacotherapy intensification rates in patients not at clinical goals, and risky prescribing events).
Risk of contamination (other bias)	Unclear risk	Clustered‐RCT but randomisation was at the physician level, so communications might have happen between intervention and control physicians working in the same clinic. Groups B and C are pretty similar. Physicians in groups B and C received the same simulated learning intervention, but group B received printed feedback and group C received verbal feedback. The control group received no education.
Other bias	Low risk	None identified.