OBJECTIVE To identify key systems issues from the Indian Health Service (IHS) experience that must be addressed to improve metabolic control among patients with non-insulin-dependent diabetes mellitus (NIDDM) who were followed in primary care settings. METHODS Records of diabetic patients seen in IHS facilities in specific geographic regions from 1987 to 1994. METHODS A representative sample of charts from each facility was reviewed yearly to measure key variables. The sampling frame was the number of diabetic patients currently active on the registry and the sample size calculated to measure a 10% change in selected practices at each facility. METHODS Regional diabetes coordinators reviewed charts or trained local providers to sample and extract data in a standard format. RESULTS Regional data were examined to show trends in the performance of immunizations and foot examinations and in other variables such as hypertension and metabolic control. The percentage of diabetic patients who received a single dose of pneumococcal vaccine improved from 24% in 1987 to 1988 to 57% in 1994 (P < 0.01 for trend) among diabetic patients in Minnesota, Wisconsin, and Michigan. Rates of yearly comprehensive foot examination increased from 36% to 58% (P < 0.01 for trend) over the same period. In Montana and Wyoming, the percentage of diabetic patients with uncontrolled hypertension (defined as the mean of three systolic blood pressure measurements of > or = 140 mm Hg or diastolic pressure measurements > or = 90 mm Hg, or both, during the previous year) decreased from 36% in 1992 to 25% in 1993 after the regional diabetes coordinator emphasized hypertension control. In 1994, when less emphasis was placed on hypertension, 33% of the diabetic patients had uncontrolled hypertension. Estimates of metabolic control from records of diabetic patients in Washington, Oregon, and Idaho in 1994 showed that 29% of patients had excellent metabolic control (a hemoglobin A1c [HbA1c] level < or = 7.5% or mean blood glucose level < or = 9.2 mmol/L) within the past year; only 9% experienced poor control (a HbA1c level > 12% or mean blood glucose level > 18.9 mmol/L). CONCLUSIONS The IHS experience shows that standard, ongoing monitoring of key variables allows facilities to improve diabetes care. Simple, reliable methods of defining metabolic control combined with a feedback system in the primary care setting are needed to improve metabolic control in patients with NIDDM.