Variability and predictability of large-volume red blood cell transfusion in cardiac surgery: a multicenter study
Article first published online: 9 AUG 2007
Volume 47, Issue 11, pages 2081–2088, November 2007
How to Cite
Karkouti, K., Wijeysundera, D. N., Beattie, W. S., Callum, J. L., Cheng, D., Dupuis, J.-Y., Kent, B., Mazer, D., Rubens, F. D., Sawchuk, C., Yau, T. M. and the Reducing Bleeding in Cardiac Surgery (RBC) Research Group (2007), Variability and predictability of large-volume red blood cell transfusion in cardiac surgery: a multicenter study. Transfusion, 47: 2081–2088. doi: 10.1111/j.1537-2995.2007.01432.x
- Issue published online: 9 AUG 2007
- Article first published online: 9 AUG 2007
- Received for publication March 1, 2007; revision received May 10, 2007, and accepted May 12, 2007.
BACKGROUND: In cardiac surgery, excessive blood loss requiring large-volume red blood cell (RBC) transfusion is a common occurrence that is associated with significant morbidity and mortality. The objectives of this study were to measure the interinstitution variation and predictability of large-volume RBC transfusion.
STUDY DESIGN AND METHODS: Data were retrospectively collected on 3500 consecutive cardiac surgical patients at seven Canadian hospitals during 2004. The crude and risk-adjusted institutional odds ratios (ORs) for large-volume (≥5 U) RBC transfusion were calculated with logistic regression. The predictive accuracy of an existing prediction rule for large-volume RBC transfusion was calculated for each institution.
RESULTS: Large-volume RBC transfusion occurred in 538 (15%) patients. When compared to the reference hospital (median crude rate), the institutional unadjusted and adjusted ORs for large-volume RBC transfusion ranged from 0.29 to 1.26 and 0.14 to 1.15, respectively (p < 0.0001 for interinstitution variation). The variation was lower, but still considerable, for excessive blood loss, defined as at least 5-U RBC transfusion or reexploration; the ORs ranged from 0.42 to 1.22 (p < 0.0001). The prediction rule performed well at most sites; its pooled positive predictive value for excessive blood loss was 71 percent (range, 63%-89%), and its negative predictive value was 90 percent (range, 87%-93%).
CONCLUSIONS: There is marked interinstitution variation in large-volume RBC transfusion in cardiac surgery that is not explained by patient- or surgery-related factors. Despite this variation, patients at high or low risk for large-volume RBC transfusion can be accurately identified by a prediction rule composed of readily available clinical variables.