In a population-based study, 5431 diabetic people in southern Wisconsin who were mature at onset of diabetes were identified, and their charts were reviewed. Recent glycemic control was evaluated from these charts. An index of recent glucose control with at least three glucose determinations abstracted from the doctor's chart was constructed. A sample of 1370 of the 5431 individuals was examined as part of this study. An index of glucose control could be derived for 568 of 674 people using insulin and for 565 of 696 nonusers of insulin. Correlation of the index with the glycosylated hemoglobin obtained at the time of study was significant in insulin users and nonusers. Patients with the poorest blood glucose control as described by index of past control or by current glycosylated hemoglobin had higher rates of retinopathy compared with patients who had the best control. When both the index of past control and current glycosylated hemoglobin values were combined, rate ratios for retinopathy for people with poorest control compared with those with best control were slightly better than when each measure was used alone. These analyses suggest that in planning programs, health-care impact may be increased if past and current data are considered.