Table 2 Study methods
Author (year) | Primary Outcome | Statistical Software and ML algorithms | Models | Model Evaluation | Training/Test Split | Missing Data Method |
|---|---|---|---|---|---|---|
Martin (2023) | Revision | R (Version 4.1.11 R Core Team) | Cox lasso Random survival forest Gradient boosting Super learner | Concordance - Harrell C-index Calibration | 75/25 | Multiple imputation |
Martin (2022a) | Revision | R (Version 3.6.1) | Cox Lasso | Concordance - Harrell C-index Calibration | NR (external validation study - original model 75/25) | Patients included if they had data for five predictive models from original model |
Martin (2022b) | Revision | R (Version 3.6.1) | Cox Lasso Survival Random Forest Generalized Additive Model (GAM) Gradient Boosted Regression Model (GBM) | Calibration Concordance | 75/25 | Multiple imputation |
Johnson (2023) | All-cause re-operation | SciPy version 1.6.2 | MLPClassifier GaussianNB LogisticRegression KNeighborsClassifier BaggingClassifier RandomForestClassifier AdaBoostClassifier GradientBoostingClassifier XGBClassifier | AUC Calibration AUPRC F1 Recall Accuracy Precision | 75/25 | Multiple imputation |
Lopez (2023) | ACLR post-op outcomes (revision included) | TensorFlow Python open-source coding platform (Google Brain, Alphabet Inc., Mountain View, CA) | Artificial Neural Network ML Logistic Regression | AUC Accuracy | 80/20 | Excluded |
Ye (2022) | Graft failure | SPSS (Version 25.0; IBM Corp) | Logistic Regression Gaussian Naïve Bayes Random Forest XGBoost Isotonic XGBoost Sigmoid XGBoost | AUC Accuracy F1 | 90/10 | NR |
Martin (2024) | Revision | R (RStudio 2022.07.1) | Cox Lasso | Concordance - Harrell’s C-index Calibration | 75/25 | NA |
Jurgensmeier (2023) | Secondary meniscus tear | R 4.1.2 using RStudio version 1.4.1717 (RStudio, Boston, MA) | SVM Random Forest XGBoost Elastic Net | Discrimination - AUROC Calibration Brier score | 0.632 bootstrapping with 1000 resampled datasets | Multiple imputation |
Lu (2022) | Secondary meniscus tear | R 4.1.2 using RStudio version 1.2.5001 (RStudio, Boston, MA). | Random Survival Forests | C-statistic (AUROC) (Concordance) Calibration Brier Score | 0.632 bootstrapping with 1000 resampled datasets | Multiple imputation |