We have built a clinical workstation to help doctors and nurses care for patients with HIV infection. This knowledge-based medical record system provides medication alerts, reminders about primary care, and on-line information to support the care of patients with HIV infection. We are conducting a controlled clinical trial of this computer system in a single practice setting, which consists of 18 staff physicians, 13 nurses, and 113 residents, who cooperatively practice in four teams. Two teams of physicians are assigned to an intervention group and two teams to a control group. This paper reports preliminary results from the first year of study, January 15, 1992, through January 14, 1993. During this period 274 patients with HIV infection were followed by the general medical practice--130 in a control group and 144 in an intervention group. Physicians in the intervention group more rapidly and more completely followed primary care guidelines than did physicians in the control group. Patients in the intervention group had 2476 ambulatory or emergency visits (17.2 visits per patient) compared with 1882 visits (14.5 visits per patient) for the control patients (p < 0.01). There were 101 hospitalizations for 51 patients in the intervention group (an admission rate of 0.7) compared with 104 admissions for 54 patients in the control group (an admission rate of 0.8) (p = NS). There were 8 deaths in the intervention group (5.6%) compared with 13 (10%) in the control group (p = NS).(ABSTRACT TRUNCATED AT 250 WORDS)