Perioperative blood loss in cardiac surgery: Validation of machine learning-derived clusters

BackgroundThere is no universally accepted definition of perioperative blood loss in cardiac surgery. Existing methods are based on chest tube output and are not normalised for patient weight.ObjectiveTo validate machine learning-derived blood loss severity clusters based on a haemoglobin mass loss per kilogram (Hb/kg index).MethodsThis single-center prospective study included 195 patients undergoing cardiac surgery between October 2023 and November 2024. Three clusters derived using K-Medoids were mapped to the Hb/kg index to define cut-offs. Cluster discrimination was assessed by receiver operating characteristics (ROC) analysis (area under the curve (AUC)). Group comparisons were performed using analysis of covariance adjusted for age and gender. Associations between the Hb/kg index and clinical outcomes, including transfusion requirements and complications were analysed using Chi-square tests and adjusted two-way Analysis of Covariance (ANCOVA).ResultsClustering identified three groups (Mild, Moderate, Severe) defined by optimal Hb/kg thresholds of 1.72 and 2.10. The Severe cluster demonstrated strong discriminative performance (AUC = 0.790, 95% confidence interval 0.721-0.859). Chest tube output did not differ significantly between clusters (<i>p</i> = 0.097), while haemoglobin mass loss through chest tubes demonstrated a significant effect (<i>p</i> = 0.011).ConclusionsThe Hb/kg Index is a validated, data-driven, objective metric for perioperative blood loss, offering greater precision than traditional chest tube drainage volume. It effectively stratifies bleeding severity and identifies high-risk patients with lower BMI.