Accurately predicting the need for perioperative RBC transfusions reduces costs and conserves blood products while optimizing safety for patients. Since the 1970s, surgical teams have used maximum surgical blood order schedules (MSBOS) to predict transfusion needs based solely on the type of surgery; patient factors are not considered. Researchers at University of California San Francisco used the artificial intelligence model Smart Match to account for 82 variables (including patient’s medical history, laboratory values, demographics, type of surgery, medications, transfusion history, and MSBOS recommendations) to predict RBC transfusions. The model was trained on 235,054 retrospective, elective adult surgeries (3.04% of patients received RBC transfusions) and tested on 24,003 more surgeries (2.2% of patients received RBC transfusions) with an area under the receiver operating characteristic curve of 0.94. The Smart Match was better able to predict perioperative RBC transfusions than clinician behavior with a missed case rate of 0.29 compared to 0.60 for clinicians. The top predictors of perioperative transfusions included hospital clinical codes, surgery length, and hemoglobin levels. Since results are institution specific, further studies at other hospitals are needed.
Reference:
Leave a Reply