26 September 2024 : Database Analysis
Development and Internal Validation of Machine Learning to Predict Postoperative Worse Functional Status after Surgical Treatment for Thoracic Spinal Stenosis
Tun Liu12ABCE, Jia Li2D, Huaguang Qi3F, Bin Guo2B, Songchuan Zhao4CD, Baoping Zhang2F, Langbo Li5D, Gang Wu2A, Gang Wang1D*DOI: 10.12659/MSM.945310
Med Sci Monit 2024; 30:e945310
Table 3 Performance metrics of the best-performing predictive ML models at the corresponding follow-up visits on the training set and the testing data, representing internal validation.
| Metric | 30-Day worse function status | 6-Month worse function status |
|---|---|---|
| Training (10-fold cross-validation) | ||
| Accuracy (%) | 87.02 | 85.32 |
| AUC | 0.82 | 0.78 |
| Testing (internal validation) | ||
| Accuracy (%) | 88.17 | 86.03 |
| AUC | 0.83 | 0.80 |
| Brier score | 0.06 | 0.08 |
| The following ML algorithms were used: (1) 30-day postoperative worse function status: XGBoost machine (); (2) 6-month postoperative worse function status: Naïve Bays (). The F1 score is a composite performance measure and is defined as the harmonic mean of precision and sensitivity. AUC – area under the curve; ML – machine learning. In , preoperative variables were utilized to predict postoperative worse function status; In , preoperative variables and intraoperative variables were implemented in the model to predict postoperative worse function status. | ||