Purpose The purpose of this study is to describe preliminary results of an innovative quality improvement intervention focused on improving physician practice patterns in diabetes care via Coaching Physicians by RN certified diabetes educators (CDEs), a program called “CPR for Diabetes Care.” The program identified primary care physicians with optimal diabetes control rates below the system aggregate (n = 195). Physicians with the lowest rates (n = 74) were targeted for comprehensive intervention. All other low-performing physicians practicing in the same clinic system (n = 121) comprised the comparison group. Data were obtained from electronic diabetes registries for 2007 and 2008. Each physician had a set of measures from 2 points in time. Measures included optimal diabetes scores and the 5 component measures of the optimal diabetes care bundle (A1C <7, low-density lipoprotein cholesterol <100, blood pressure <130/80, aspirin use if older than 40, and no tobacco use). T tests and difference-in-difference models were used to examine changes over time. Optimal diabetes scores increased 11.7 points (from 14.7% to 26.4%) for intervention physicians and 4.0 points (from 29.7% to 32.9%) for comparison physicians. The improvement was greater for the intervention group. The greatest component improvements were in control of blood pressure and cholesterol. Coaching low-performing physicians dramatically improved the proportion of diabetes patients with optimal diabetes control. The CPR for Diabetes Care program represents an innovative and effective way to address the long-standing problem of disseminating and sustaining quality improvement efforts by focusing on low-performing physicians.