To determine whether reminders are cost-effective for an adult primary care internal medicine center, we randomized 590 scheduled, follow-up appointments to no reminder, computer-generated letter reminders, and telephoned reminders. The no-show rate was reduced from 24% in the control group to 14% in the reminder groups, and letter and telephoned reminders were equally effective. An economic analysis showed that about two thirds of the savings realized from reminders was generated in 23% of the patients whose prior predicted probability of a no-show appointment was above 20%. However, in our primary care center, computer-generated letter reminders were estimated to be cost-effective whenever the probability of a no-show was above 5%, and telephoned or manual letter reminders were estimated to be cost-effective whenever the probability of a no-show was above about 11%. Based on our sensitivity analysis, telephoned or manual letter reminders should be cost-effective in many other ambulatory settings as well, although in some settings, reminders may be restricted to patients at high risk for no-show behavior.