Effectiveness of computerized physician order entry with decision support to reduce inappropriate blood transfusions

• Objective: To reduce unnecessary transfusions, our hospital instituted transfusion guidelines and computerized physician order entry (CPOE) with decision support. Two years after implementation, we sought to determine the number of inappropriate transfusions, determine reasons for decision support bypasses and overrides, and determine patient specific predictors of inappropriate transfusions. • Methods: We conducted a retrospective 1-year cohort study of general medical inpatients with a hemoglobin ≥ 9 g/dL who received at least 1 transfusion of red blood cells. Inpatient charts and CPOE decision support were compared to determine transfusion appropriateness. Decision support was further evaluated to determine bypasses and overrides. Patient specific predictors of inappropriate transfusions were determined by logistic regression. • Results: Of 214 transfusions, 54% and 62% were inappropriate by chart review and CPOE decision support, respectively. Of the 141 transfusion orders that bypassed decision support by indicating active bleeding, active bleeding could not be found in 54% of chart reviews. Of the 45 deemed inappropriate by decision support, 73% of overrides indicated their superior instructed them to transfuse the patient. No patient-specific predictors associated with inappropriate transfusions were found. • Conclusion: After institution of transfusion guidelines and CPOE decision support, over half of our transfusion prescribing is still inappropriate. Decision support was bypassed altogether in two-thirds of transfusion orders, and over two-thirds of the overrides indicated a superior instructed the transfusion. This suggests that decision support alone will not completely remove inappropriate transfusions and other front-ended interventions targeted at the deciding provider may be needed.