Date of Award

January 2014

Document Type


Degree Name

Medical Doctor (MD)



First Advisor

Shamsuddin Akhtar

Subject Area(s)



Blood transfusion practice in cardiac surgery has evolved over the past few decades, with more studies suggesting that perioperative red blood cell (RBC) transfusions may be associated with worse mortality and morbidity. We sought to systematically study the effect of post-operative RBC transfusions in patients after cardiac surgery at our tertiary care institution, and whether the effects were dose-dependent. A total of 2,351 patients receiving coronary artery bypass graft (CABG) surgery with or without concurrent valve surgery between June 1, 2006 and April 30, 2013 were hospitalized in the CTICU. Eighteen preoperative variables were used to calculate propensity scores, which were then used in the greedy matching algorithm. Good balance of the variables could not be achieved for the cohort of patients receiving > 5 U packed RBCs (PRBCs) (compared to the non-transfused control group). In-hospital mortality rates were 1.6% for the 976 patients who were not transfused RBCs and 4.3% for 1,375 patients who received any blood transfusions. However, the in-hospital mortality rates were only 1.0% in the group of patients who were transfused 1-2 U PRBCs, 1.4% in the 3-5 U group, and 21.4% in the > 5 U group. Of the 11 post-operative outcomes studied, patients who were transfused had a higher in-hospital mortality rate (adjusted odds ratio, 2.63 [95% CI, 1.16-5.92], P = 0.0202), longer intensive care unit length of stay (P < 0.0001), higher readmission rate to the ICU (2.72 [1.14-6.45], P = 0.0239), higher rate of in-hospital post-operative events

(1.94 [1.48-2.53], P < 0.0001), prolonged ventilation (2.81 [1.59-4.98], P = 0.0004), renal failure (3.80 [1.42-10.20]), P < 0.0079), and higher rates of re-operation (10.42 [4.15- 26.32], P < 0.0001). When the analysis was restricted to patients who received < 5 units of PRBCs, many risks were attenuated including in-hospital mortality (1.13 [0.43-2.92], P = 0.8085), ICU readmission (1.72 [0.67-4.35], P = 0.2571) and prolonged ventilation (1.79 [0.93-3.44], P = 0.0824).

Our study showed that patients receiving transfusions overall had worse outcomes compared to non-transfused patients, but many of the outcomes were attenuated when the cohort of patients receiving > 5 U PRBCs was excluded. Also, the difference in in- hospital mortality between transfused and non-transfused patients was primarily due to the patients who received > 5 U PRBCs. Importantly, it was not possible to match the preoperative variables of this cohort of patients to the non-transfused control group, suggesting that their worse outcomes could be due to increased severity of illness even before surgery. Hence, our study provides evidence that smaller RBC transfusions have few negative impacts on patient outcomes, and larger transfusions may be a marker of a sicker population.


This thesis is restricted to Yale network users only. This thesis is permanently embargoed from public release.