RESUMO
OBJECTIVE: Transfusion is becoming safer but is not free of risk. It is important to establish a good approach to transfusion management and calculate real losses. Risk factors for transfusion should be identified. MATERIAL AND METHODS: This was a prospective study of 102 patients who did not receive intraoperative autotransfusion of shed blood, selected from a group of 127 who were undergoing primary knee arthroplasty. We initially calculated the amount of blood shed. Then, by multivariate logistic regression analysis we identified the model that best predicted that a patient would require transfusion. Receiver operating characteristic curves were constructed and the area under the curves calculated. RESULTS: Mean (SD) blood loss was calculated to be 1786 (710) mL. The best model considered initial hemoglobin (Hb), weight, height, and sex as predictive factors: Probability = 1/ (1+e(-Z)), where Z = 11.542 - 1.074 x initial Hb (g/dL) - 0.039 x Weight (kg) + 0.031 x Height (cm) + 0.267 x (sex: male=1 or female=0). The area under the ROC curve was 0.805 (0.44). CONCLUSION: Initial Hb, which can be modified before surgery, is one of the factors that most affects whether or not the patient will need a transfusion. Therefore, one of our first objectives in the process of managing transfusion is to improve preoperative Hb values.