A model for predicting transfusion after coronary artery bypass grafting

Abstract

Background. Blood conservation has become an important issue in cardiac surgery. This study was undertaken to determine if the need of blood transfusion could be predicted from preoperative patient variables. Methods. From January 1, 1992, to December 31, 1993, 2,033 patients having isolated coronary artery bypass grafting procedures were studied; 1,446 (71%) were male and 587 (29%), female. The mean age was 65.1 ± 9.9 years (range, 31 to 88 years). Emergency operation, urgent operation, and reoperations were done in 78 (4%), 188 (9%), and 189 (9%) patients, respectively. In the entire group, 1,245 (61%) received transfusion during hospitalization, and 788 (39%) did not. Logistic regression analysis was used to construct a model that predicted the need of transfusion of packed red blood cells after coronary artery bypass grafting. A transfusion risk score was constructed by assigning points to independent predictive factors on the basis of the logistic regression coefficient and the odds ratio. Preoperative predictors of transfusion were emergency operation, urgent operation, cardiogenic shock, catheterization-induced coronary occlusion, low body mass index, left ventricular ejection fraction lower than 0.30, age greater than 74 years, female sex, low red cell mass, peripheral vascular disease, insulin-dependent diabetes, creatinine level greater than 1.8 mg/dL, albumin value lower than 4 g/dL, and redo operation. Results. The mean transfusion risk score for patients receiving 0, 1 to 4, and greater than 4 units of packed red blood cells was 2.3 ± 0.9, 5.2 ± 3.0, and 9.6 ± 3.5, respectively (p = 0.001). Patients with a score higher than 6 had a 95% transfusion incidence. The predictive model was validated on 422 patients having coronary artery bypass grafting from January 1 to May 31, 1994. The observed rates of the validation group fell within the 95% confidence intervals of the predicted rates. Conclusions. These data demonstrate that readily available patient variables can predict patients at risk for transfusion. Routine use of aprotinin and other adjustments of cardiopulmonary bypass should be considered to reduce transfusion in high-risk patients.