OBJECTIVES: To develop a new method that objectively evaluates and compares provider efficiency in diabetes care integrating multiple measures of both resource use and clinical outcomes. Contemporary performance assessment technology only evaluates one outcome measure at a time and does not consider resource use. METHODS: The basis for comparison is the technical efficiency with which providers transform resources into outcomes relative to peers. By identifying the highest level of performance - the efficiency frontier - the relative efficiency of each provider under evaluation is quantified. Relative efficiency is determined among diabetes care teams at Steno Diabetes Center, Denmark, over a 6-year period (1995-2000). Applied outcome measures follow recommendations of the Diabetes Quality Improvement Project and resources include clinical tests and patient care time of physicians, nurses and dieticians. Data Envelopment Analysis (DEA), a linear programming optimization technique, estimates relative efficiency, while regression analysis (RA) examines the robustness of the method. Impact of case-mix differences is examined by subgroup analysis. RESULTS: Over the study period, DEA identified a total of four teams on the frontier for clinical tests and ten teams for patient care time. Below frontier practice implied a 7% excess use of HbA1c tests, 12% blood pressure tests and 22% lipid measurements and 6% physician, 5% nurse and 13% dietician care time. Subgroup estimates implied less variation across teams and up to 30% reduction in predicted excess use of resources. When adjusting for clinical outcomes, RA identified an excess use of resources among the same teams operating below frontier in DEA. CONCLUSIONS: The method appears robust in identifying performance below frontier set by peers and a useful diagnostic tool for identifying efficiency gaps and setting targets for performance improvement in diabetes care. Identification of major subgroups in outcomes and resource use is important for practical implementation of the method.