Machine Learning Approaches for Predicting Intraoperative Blood Transfusion in Partial Hip Arthroplasty

Area of Science:

• Orthopedic Surgery
• Medical Informatics
• Artificial Intelligence in Medicine

Background:

• Partial hip arthroplasty (PHA) carries a risk of significant blood loss, especially in elderly patients with comorbidities.
• Predicting the need for intraoperative blood transfusion is vital for patient safety and effective surgical planning.
• Machine learning (ML) offers potential for data-driven clinical decision support in predicting transfusion requirements.

Purpose of the Study:

• To evaluate the efficacy of various ML algorithms in predicting intraoperative blood transfusions during PHA.
• To identify key preoperative and intraoperative predictors of transfusion needs.
• To assess the interpretability of ML models using SHAP analysis.

Main Methods:

• Retrospective cohort study of 202 PHA patients.
• Trained and tested six ML algorithms: Logistic Regression, Decision Tree, SVM, ANN, Random Forest, and Gradient Boosting.
• Evaluated model performance using accuracy, F1-score, and AUC; utilized SHAP for interpretability.

Main Results:

• 42.1% of patients received intraoperative transfusions.
• Key predictors included low preoperative hemoglobin, high ASA score, prolonged operative time, increased blood loss, and elevated INR.
• Random Forest and Decision Tree models showed high accuracy (95.1%) and F1-score (0.960); SVM achieved the highest AUC (0.992).
• SHAP analysis highlighted hemoglobin, age, ASA score, INR, and operative time as most influential.

Conclusions:

• ML models, particularly Random Forest, Decision Tree, and SVM, effectively predict intraoperative transfusion needs in PHA.
• Explainable AI (SHAP) enhances clinical interpretability, supporting personalized patient management.
• Findings support integrating ML into clinical decision support systems, pending external validation.