| Abstrakt: |
SUMMARY The purpose of this study was to determine and quantify those factors that anesthetists use when deciding to transfuse blood and to create a model that could be used to make decision making more consistent. Two hundred audit forms, completed during intermediate or major surgery, were analyzed to determine the strengths of certain factors in the decision-making process related to transfusion. Eighty of these had a satisfactorily complete data set with hemoglobin concentrations ([Hb]) between 7.0 and 10.0 g/dL. This audit demonstrates that, of the physiological factors, the risk of tissue hypoxia and the hemoglobin concentration were the most important, and the presence of ongoing bleeding an important contextual factor. Peer pressure to transfuse became apparent at a [Hb] of about 8.5 g/dL. A binary logistic regression (BLR) model, a probabilistic model (PM) and a neural network (NN) were created using this database and were used to predict transfusion decisions in a further 50 audit returns. Using [Hb], risk of tissue hypoxia and ongoing hemorrhage, the sensitivities, specificities, positive predictive values and negative predictive values, respectively, were as follows: BLR, 83%, 81%, 86%, and 77%; PM, 100%, 62%, 78% and 100%; and NN, 86%, 82%, 86% and 82%. If physiological and environmental factors are combined into a statistical model, it is possible to predict the decision to transfuse with an overall accuracy of about 82-86%. [ABSTRACT FROM AUTHOR] |
Plný text