Introduction: Recently 5 established regional quality improvement collaboratives agreed to partner on assessing and improving care across their regions under the umbrella of the IMPROVE Network. The IMPROVE Network has chosen to begin by assessing regional signatures for potentially discretionary [<3 red blood cell (RBC)] transfusions. Hypothesis: Adjusted rates of potentially discretionary RBC transfusions differ across regional collaboratives. Methods: We enrolled patients who between 1/1/2008 through 6/30/2012 underwent isolated non-emergent coronary artery bypass (CABG) surgery across 58 medical centers and 4 IMPROVE Network regions. The study dataset of 11,600 consists of the most recent 200 patients from each of the 58 centers who received 0, 1 or 2 units of RBC transfusion during the index admission. We collected patient and disease characteristics, and percent of cases receiving intra- or post-operative RBC transfusions. Logistic regression models were constructed to model transfusion rates and adjust for variation in case mix across centers within each region, including: age, sex, body surface area, pre-operative hematocrit, vascular disease, diabetes, creatinine, and number of diseased vessels. We calculated crude and risk-adjusted transfusion rates across centers, and then stratified by regions. Results: RBC transfusions occurred among 24.8% (2,875 of 11,600) of CABG procedures. Significant variation remained across the four regions, even after adjusting for case mix (crude range: 15.3% - 29.8%; adjusted range: 18.5% - 26.6%, p<0.001), Figure. Conclusion: Potentially discretionary RBC transfusions were common, and varied across centers and regions, even after adjusting for differences in case mix. These data suggest that differences in regional and center practice environments, including transfusion triggers and anemia management, may be the primary drivers for variations in RBC rates.