To maximize disease control, patients must participate effectively in their medical care. The authors developed an intervention designed to increase the involvement of patients in medical decision making. In a 20-minute session just before the regular visit to a physician, a clinic assistant reviewed the medical record of each experimental patient with him/her, guided by a diabetes algorithm. Using systematic prompts, the assistant encouraged patients to use the information gained to negotiate medical decisions with the doctor. A randomized trial was conducted in two university hospital clinics to compare this intervention with standard educational materials in sessions of equal length. The mean pre-intervention glycosylated hemoglobin (HbA1) values were 10.6 +/- 2.1% for 33 experimental patients and 10.3 +/- 2.0% for 26 controls. After the intervention the mean levels were 9.1 +/- 1.9% in the experimental group (p less than 0.01) and 10.6 +/- 2.22% for controls. Analysis of audiotapes of the visits to the physician showed the experimental patients were twice as effective as controls in eliciting information from the physician. Experimental patients reported significantly fewer function limitations. The authors conclude that the intervention is feasible and that it changes patient behavior, improves blood sugar control, and decreases functional limitations.