• Marcin Wasowicz, Stuart A McCluskey, Duminda N Wijeysundera, Terrence M Yau, Massimiliano Meinri, W Scott Beattie, and Keyvan Karkouti.
• Toronto General Hospital, 200 Elizabeth St., 3EN-402 Toronto, ON, Canada M5G 2C4.
• Anesth. Analg. 2010 Aug 1;111(2):331-8.

BackgroundAccurate risk stratification may help reduce the burden of excessive blood loss after cardiac surgery. We measured the incremental value of thrombelastography to an existing risk prediction model for excessive blood loss in cardiac surgery.MethodsThis observational study included 434 patients who underwent cardiac surgery with cardiopulmonary bypass (CPB) and had thrombelastographic measures before and during CPB, their risk of excessive blood loss could be calculated with an existing risk prediction model and they had not received clopidogrel or warfarin within 5 days of surgery. Excessive blood loss was defined as transfusion of > or = 5 U of red blood cells from termination of CPB to 1 day after surgery. Logistic regression models including an existing risk prediction model without and with thrombelastographic measures were constructed. Improvement in risk prediction was measured by the area under the curve and net reclassification improvement.ResultsExcessive blood loss occurred in 59 of 434 patients (13.6%). The only thrombelastographic measure that improved risk stratification was maximum amplitude during CPB, which reflects maximum clot strength. Although the addition of this variable to the existing prediction model did not have a material effect on the area under the curve (increased from 0.780 to 0.784; P = 0.8), it did improve the net reclassification improvement by 12% (P = 0.05), primarily by improving the detection of high-risk cases.ConclusionsRisk stratification for excessive blood loss after cardiac surgery is improved when on-CPB thrombelastography is added to an existing risk prediction model that incorporates readily available patient- and surgery-related variables, but large, multicenter trials are needed to verify this finding and create a new risk prediction model.

27 keywords...

hide…