Advancing Post-Surgical Care: Machine Learning Prediction of Red Blood Cell Transfusion After Elective Aortic Surgery
Presented During:
Thursday, April 25, 2024: 5:38PM - 7:00PM
Sheraton Times Square
Posted Room Name:
Central Park
Abstract No:
P0032
Submission Type:
Abstract Submission
Authors:
Adam Carroll (1), Nicolas Chanes (1), Michael Kirsch (1), Ananya Shah (1), Muhammad Aftab (1), T. Brett Reece (1)
Institutions:
(1) University of Colorado Anschutz, Denver, CO
Submitting Author:
Adam Carroll
-
Contact Me
University of Colorado Anschutz
University of Colorado Anschutz
Co-Author(s):
Nicolas Chanes
-
Contact Me
University of Colorado Anschutz
University of Colorado Anschutz
Michael Kirsch
-
Contact Me
University of Colorado Anschutz
University of Colorado Anschutz
Ananya Shah
-
Contact Me
University of Colorado Anschutz
University of Colorado Anschutz
*Muhammad Aftab
-
Contact Me
University of Colorado Anschutz
University of Colorado Anschutz
*T. Brett Reece
-
Contact Me
University of Colorado Anschutz
University of Colorado Anschutz
Presenting Author:
Adam Carroll
-
Contact Me
University of Colorado Anschutz
University of Colorado Anschutz
Abstract:
Objective:
Post-operative blood transfusion in cardiac surgery has been linked to adverse outcomes including acute kidney injury and post-operative mortality. Machine learning models for post-operative red blood cell (RBC) transfusion have not been applied to aortic surgery specifically, despite the risk of bleeding in arch surgery. We sought to develop a machine learning model to predict the need for post-operative transfusion of RBC in aortic surgery.
Methods:
We identified all adult patients who underwent elective aortic surgery between June 2009 to October 2022 (n = 543) from our single institution prospectively maintained database. Patients were randomly divided into training (70%) and testing (30%) sets with various eXtreme gradient boosting (XGBoost) models constructed to predict postoperative RBC transfusion in the cardiothoracic intensive care unit (CTICU). From the index hospitalization, we extracted 64 input parameters, including 24 demographic characteristics as well as 8 preoperative and 32 intraoperative variables. To assess model performance, we employed various evaluation metrics, including accuracy, area under the receiver operating characteristic curve (AUC-ROC), and area under the precision-recall curve (AUC-PR, mean average precision). Additionally, we assessed model performance on various subsets within the broader study population, including root, hemiarch, and total arch procedures.
Results:
Postoperative RBC transfusion was required in 48.8% of patients (265 cases) following aortic surgery. The final XGBoost model demonstrated a 77% cross-validation accuracy and a 77% test accuracy, achieving an AUC-ROC of 0.79 and an AUC-PR of 0.56. When stratifying by aortic procedure, the model attained an AUC-ROC of 0.73 for root procedures, 0.82 for hemiarch cases, and 0.72 for total arch surgeries. Prominent factors associated with an increased risk of postoperative RBC transfusion included extended periods of circulatory arrest, SACP via innominate or axillary cannulation, low nadir hemoglobin, higher intraoperative blood product transfusion, low preoperative hemoglobin, and a history of prior sternotomy. Comparatively, baseline thrombocytopenia, extended cardiopulmonary bypass time, and lower nadir bladder temperature had significantly less effect on need for post-operative transfusion.
Conclusions:
The machine learning model developed had a strong performance in the overall cohort for predicting post-operative transfusion for all aortic surgery, and for specific procedure subsets. Extended periods of circulatory arrest, antegrade cerebral perfusion, baseline anemia and intraoperative bleeding were the strongest predictors of need for post-operative transfusion, while nadir temperature, baseline platelets, and extended cardiopulmonary bypass time did not have as significant of a model impact.
Post-operative blood transfusion in cardiac surgery has been linked to adverse outcomes including acute kidney injury and post-operative mortality. Machine learning models for post-operative red blood cell (RBC) transfusion have not been applied to aortic surgery specifically, despite the risk of bleeding in arch surgery. We sought to develop a machine learning model to predict the need for post-operative transfusion of RBC in aortic surgery.
Methods:
We identified all adult patients who underwent elective aortic surgery between June 2009 to October 2022 (n = 543) from our single institution prospectively maintained database. Patients were randomly divided into training (70%) and testing (30%) sets with various eXtreme gradient boosting (XGBoost) models constructed to predict postoperative RBC transfusion in the cardiothoracic intensive care unit (CTICU). From the index hospitalization, we extracted 64 input parameters, including 24 demographic characteristics as well as 8 preoperative and 32 intraoperative variables. To assess model performance, we employed various evaluation metrics, including accuracy, area under the receiver operating characteristic curve (AUC-ROC), and area under the precision-recall curve (AUC-PR, mean average precision). Additionally, we assessed model performance on various subsets within the broader study population, including root, hemiarch, and total arch procedures.
Results:
Postoperative RBC transfusion was required in 48.8% of patients (265 cases) following aortic surgery. The final XGBoost model demonstrated a 77% cross-validation accuracy and a 77% test accuracy, achieving an AUC-ROC of 0.79 and an AUC-PR of 0.56. When stratifying by aortic procedure, the model attained an AUC-ROC of 0.73 for root procedures, 0.82 for hemiarch cases, and 0.72 for total arch surgeries. Prominent factors associated with an increased risk of postoperative RBC transfusion included extended periods of circulatory arrest, SACP via innominate or axillary cannulation, low nadir hemoglobin, higher intraoperative blood product transfusion, low preoperative hemoglobin, and a history of prior sternotomy. Comparatively, baseline thrombocytopenia, extended cardiopulmonary bypass time, and lower nadir bladder temperature had significantly less effect on need for post-operative transfusion.
Conclusions:
The machine learning model developed had a strong performance in the overall cohort for predicting post-operative transfusion for all aortic surgery, and for specific procedure subsets. Extended periods of circulatory arrest, antegrade cerebral perfusion, baseline anemia and intraoperative bleeding were the strongest predictors of need for post-operative transfusion, while nadir temperature, baseline platelets, and extended cardiopulmonary bypass time did not have as significant of a model impact.
Aortic Symposium:
Aortic Arch
Image or Table
