A prospective cohort study develops and validates a machine learning model for predicting ecchymosis after total knee...

Area of Science:

• Orthopedic Surgery
• Biomedical Engineering
• Data Science in Medicine

Background:

• Total knee arthroplasty (TKA) is a common surgery for advanced knee disease.
• Perioperative bleeding and ecchymosis are significant complications after TKA.
• Accurate prediction of ecchymosis can improve patient outcomes and reduce healthcare costs.

Purpose of the Study:

• To develop and validate a machine learning model for predicting ecchymosis after TKA.
• To identify key clinical and laboratory risk factors associated with TKA-related ecchymosis.
• To enable personalized perioperative blood management strategies.

Main Methods:

• Prospective study of 416 TKA patients, with data split for training (312) and validation (104).
• Feature selection using LASSO, RF-RFE, and BORUTA to identify risk factors.
• Machine learning models including RF, XGBoost, SVM, and LGBM were employed for prediction.
• Key predictors identified: low prealbumin, reduced coagulation index (CI) and its change (XCI), high fibrinogen degradation products (FDP), and postoperative day 1 total blood loss (TBL).

Main Results:

• The developed prediction model demonstrated high performance.
• Area under the curve (AUC) values were 0.927 for the training set and 0.954 for the validation set.
• Significant risk factors identified include prealbumin, CI, XCI, FDP, and postoperative TBL.

Conclusions:

• A robust machine learning model accurately predicts ecchymosis in TKA patients.
• The model facilitates early and precise risk assessment.
• Findings support personalized anticoagulation therapy and optimized perioperative blood management in TKA.