Utilization of machine learning methods for predicting surgical outcomes after total knee arthroplasty
Hina Mohammed1,2, Yihe Huang1, Stavros Memtsoudis3,4
1Milken Institute School of Public Health, The George Washington University, Washington, DC, United States of America.
Plos One
|March 22, 2022
Summary
Gradient Boosting Method (GBM) models demonstrated superior prediction of adverse events after total knee arthroplasty (TKA). Key predictors included admission month, patient location, income, anemia, and length of stay, aiding clinical risk assessment.
Area of Science:
- Orthopedic Surgery
- Data Science
- Predictive Analytics
Background:
- Adverse events after total knee arthroplasty (TKA) pose clinical challenges.
- Identifying risk factors is crucial for preoperative interventions and resource allocation.
Purpose of the Study:
- To develop and evaluate machine learning (ML) models for predicting outcomes after TKA.
- To compare the performance of Logistic Regression (LR), Gradient Boosting Method (GBM), Random Forest (RF), and Artificial Neural Network (ANN) models.
Main Methods:
- Utilized National Inpatient Sample (NIS) data from 2010-2014.
- Developed LR, GBM, RF, and ANN models for predicting discharge disposition, post-surgical complications, and blood transfusion.
- Evaluated models using Brier scores, Area Under the ROC Curve (AUC), and F1 scores.
Main Results:
- GBM models exhibited superior calibration and discrimination compared to other models.
- GBM model performance: Brier scores (0.09-0.14), AUCs (79-87%), F1 scores (41-73%).
- Significant predictors included admission month, patient location, income, anemia, length of stay, and age.
Conclusions:
- ML models, particularly GBM, effectively predict outcomes following TKA using NIS data.
- Demonstrated clinical utility for accurate prognostication of complications in orthopedic surgery.
