Prompting physicians increases performance of preventive procedures, but the long-term effects of prompting, and of different types of prompting (manual versus computer), on various procedures is unclear. Nor has the effect of the optional enrollment of patients by physicians into a prompting system been studied. We examined performance of eight preventive procedures in a university-based general medical practice during three successive periods over five years: no prompting, nurse-initiated prompting, and computerized prompting. Performance of seven prompted procedures for all patients (regardless of whether they had been enrolled) was significantly increased over the five-year period from 38% (no prompting) to 43% (nurse prompting) to 53% (computer prompting). Among the procedures, influenza vaccination (12% to 59%) and mammography (4% to 33%) showed the greatest increases in performance, while fecal occult blood testing and Pap smears showed no increase or slight declines. Enrollment of patients in the optional prompting system was strongly related to performance. Overall performance of procedures for the enrolled group increased to 68%, while that of the unenrolled group remained at 37%. However, after we adjusted for differences in the percentage of patients enrolled, overall performance for the nurse system (49%) differed little from that for the computer system (55%). Manual and computer prompts had similar effects, but the computer system prompted for more patients. Though prompting remained effective five years after initiation, performance did not increase for unenrolled patients or for some procedures. Understanding these patient- and procedure-specific barriers not addressed by prompting is necessary to increase further physician performance of preventive procedures.