Background. —Previous interventions to promote performance of cancer prevention activities have largely targeted physicians in university-based practices. Methods. —We randomly assigned 40 primary care physicians in community-based practices to either (1) Cancer Prevention Reminders, computer-generated lists of overdue screening tests, and smoking and dietary assessment and counseling, supplemented by cancer education materials; or (2) controls. For each physician, we reviewed a random sample of 60 medical records for data about screening test, assessment, and counseling performance during 12-month preintervention and intervention periods. We calculated performance scores as percentage compliance with American Cancer Society and/or National Cancer Institute recommendations. Multiple regression analyses provided estimates of incremental differences in performance scores between intervention and control groups. Results. —Controlling for preintervention performance levels, significant incremental differences in performance scores between intervention and control groups (P<.05) were achieved for nine maneuvers: stool occult-blood test, + 14.5; rectal examination, + 10.5; pelvic examination, +11.8; Papanicolaou's smear, + 30.7; breast examination, + 8.7; smoking assessment, + 10.2; smoking counseling, + 17.3; dietary assessment, + 12.3; and dietary counseling, + 13.9. Increments for sigmoidoscopy and mammography were not significant. Conclusion. —Computerized reminders can significantly increase physicians' performance of cancer prevention activities in community-based practices. (JAMA. 1991;266:538-544)