Determinants of hospital variability in perioperative red blood cell transfusions during coronary artery bypass graft surgery

Abstract

Objective: To identify to what extent distinguishing patient and procedural characteristics can explain center-level transfusion variation during coronary artery bypass grafting surgery.

Methods: Observational cohort study using the Perfusion Measures and Outcomes Registry from 43 adult cardiac surgical programs from July 1, 2011, to July 1, 2017. Iterative multilevel logistic regression models were constructed using patient demographic characteristics, preoperative risk factors, and intraoperative conservation strategies to progressively explain center-level transfusion variation.

Results: Of the 22,272 adult patients undergoing isolated coronary artery bypass surgery using cardiopulmonary bypass, 7241 (32.5%) received at least 1 U allogeneic red blood cells (range, 10.9%-59.9%). When compared with patients who were not transfused, patients who received at least 1 U red blood cells were older (68 vs 64 years; P < .001), were women (41.5% vs 15.9%; P < .001), and had a lower body surface area (1.93 m² vs 2.07 m²; P < .001), respectively. Among the models explaining center-level transfusion variability, the intraclass correlation coefficients were 0.07 for model 1 (random intercepts), 0.12 for model 2 (patient factors), 0.14 for model 3 (intraoperative factors), and 0.11 for model 4 (combined). The coefficient of variation for center-level transfusion rates were 0.31, 0.29, 0.40, and 0.30 for models 1 through 4, respectively. The majority of center-level variation could not be explained through models containing both patient and intraoperative factors.

Conclusions: The results suggest that variation in center-level red blood cells transfusion cannot be explained by patient and procedural factors alone. Investigating organizational culture and programmatic infrastructure may be necessary to better understand variation in transfusion practices.

Keywords: blood loss/surgical; bypass/cardiopulmonary; cardiac surgical procedures; coronary artery bypass; erythrocyte transfusions; perioperative care.

Figure 1:

The turnip plot displays the coefficient of variation for each of the models. Each dot represents a hospital that is centered symmetrically on the horizontal x-axis based according to RBC transfusion rate (y-axis). After controlling for patient and procedural covariates, there is no significant change in the coefficient of variation.

Figure 2:

The authors conducted an observational study of 22,272 patients undergoing coronary artery bypass grafting across 43 centers. Iterative modeling approaches were used to evaluate the contribution of patient and intraoperative factors on explain hospital-level transfusion rates. The majority of center-level variation could not be explained through the patient and intraoperative factors.