Results: Objectives: an 8% increase in patients achieving two or more goals could be realized. Conclusions: There is considerable variation in efficiency of diabetes care across MGs. Health plans can use existing data to assess MG efficiency. Improved diabetes care in terms of both cost reduction and quality improvement is possible and could have substantial benefits for both patients and payers. The objectives of this secondary analysis of a retrospectively constructed dataset are: 1) to explore methods of measuring the efficiency of diabetes mellitus (DM) care across medical groups(MG)s, 2) to quantitate variation in the provision of diabetes care provided by MGs, and 3) to estimate potential resource savings and quality improvements. Methods: Study subjects were 2,511 adults with DM, insured by a single payer during 2003-2004, and receiving care at one of 33 Minnesota medical groups. Chart abstractions assessed A1c <8%, BP<130/85, and LDL<130 mg/dl. Standard methods were used to calculate resource utilization by converting all billing into RVUs scaled by a single cost factor. Quality of care(QofC) was defined as the proportion of patients achieving two or more clinical goals (A1c < 8%, BP < 130/85, or LDL < 130mg/dl). Non-parametric Data Envelopment Analysis(DEA) model the relationship between resource utilization and QofC, rank MGs in terms of patient likelihood of achieving two or more clinical goals, and identify performance benchmarks for MGs. Results: MGs with fewer than 30 DM patients (n=5) were excluded. The remaining 28 MGs averaged 109 sampled DM patients (range: 174 to 38). Median per patient costs were $19,045 with MG-specific medians ranging from $8,822 to $38,842. 73% of all patients (n=1,824) achieved two or more goals with MG-specific range from 57% to 82%. After controlling for age and comorbidities, three MGs were identified as relatively efficient “benchmarks”, four were classified as outliers, and twenty-one were found inefficient. If inefficient MGs performed at benchmark levels an estimated 35% decrease in resource use OR