BACKGROUND We investigated the effect of implementing adaptive plasma ordering criteria in the computerized physician order entry (CPOE) system, with alerts that were automatically generated if the recipient's antecedent international normalized ratio (INR) did not meet the institutional criteria. METHODS In a regional health care system consisting of 11 hospitals using a common CPOE, data on the number of plasma orders and alerts that were generated were collected over a 4-month period before prescribers were required to select an indication for plasma. When adaptive ordering was implemented prescribers had to choose from prepopulated indications for plasma: INR of 1.6 or greater with bleeding, INR of 1.6 or greater before an invasive procedure, therapeutic exchange, massive transfusion, and other. Regardless of the antecedent INR the alert did not trigger if massive transfusion or plasmapheresis was selected. Information on prescribers and recipients was collected during this 5-month period. RESULTS In the 4-month period before the adaptive alerts were implemented, 42.9% of the plasma orders generated an alert; in the 5-month period thereafter the alert rate was significantly lower at 27.9% (p < 0.0001). The percentage of heeded alerts increased during the adaptive alert period (24.3% vs. 17.1%, respectively, p = 0.004). A significant percentage (45%) of other plasma orders were for periprocedure or bleeding patients whose antecedent INR was less than 1.6. There were significant differences in prescriber specialties among those who ordered plasma using the other indication compared to all plasma orders. CONCLUSIONS Electronic interventions improve compliance with plasma guidelines but as implemented are not sufficient to completely curtail non-evidence-based ordering.