A long-standing model-driven approach for optimizing blood resource allocation in cardiac elective surgery

Background: Blood shortages is global health challenge. Despite its importance, the optimal number of packed red blood cell (PRBCs) required for elective cardiac surgery is still underexplored. This study aimed to evaluate a model that estimates the number of PRBCs for elective cardiac surgery. Study design and methods: Data on the actual PRBCs used and predicted by the model for patients who underwent elective cardiac surgery at the University Hospital of Siena from 2013 to 2020 were retrospectively collected. Before model development, the hospital's standard practice was to use approximately 10 blood bags per patient. Results: This study included 2,337 patients. The total number of PRBCs calculated by the model (plus 3 additional bags) and prepared for the surgeries was 13,227, compared to 23,370 bags obtained with the previous strategy. The ratio between actual PRBCs and the model predicted PRBCs was 29.2% for coronary surgery, 18.7% for valve surgery, and 41.4% for combined procedures. In contrast, with the previous strategy the ratio was 14.0%, 10.9%, and 25.8%, respectively. Conclusions: Machine-learning models like the one used in this study can improve patient blood management by accurately predicting the required number of blood bags for elective cardiac surgery.