We designed and implemented 2 automated, computerized screens for use at the time of antiepileptic drug (AED) test order entry to improve appropriateness by reminding physicians when a potentially redundant test was ordered and providing common indications for monitoring and pharmacokinetics of the specific AED. All computerized orders for inpatient serum AED levels during two 3-month periods were included in the study. During the 3-month period after implementation of the automated intervention, 13% of all AED tests ordered were canceled following computerized reminders. For orders appearing redundant, the cancellation rate was 27%. For nonredundant orders, 4% were canceled when information on specific AED monitoring and pharmacokinetics was provided. The cancellation rate was sustained after 4 years. There has been a 19.5% decrease in total AED testing volume since implementation of this intervention, despite a 19.3% increase in overall chemistry test volume. Inappropriateness owing to repeated testing before pharmacologic steady state was reached decreased from 54% of all AED orders to 14.6%. A simple, automated, activity-based intervention targeting a specific test-ordering behavior effectively reduced inappropriate laboratory testing. The sustained benefit supports the idea that computerized interventions may durably affect physician behavior. Computerized delivery of such evidence-based boundary guidelines can help narrow the gap between evidence and practice.