Table 3.
Model performance for predicting intradialytic hypotension.
| IDH | Model | Internal validation | External validation | ||
|---|---|---|---|---|---|
| AUROC (min-max) | AUPRC (min-max) | AUROC (p-value) | AUPRC (p-value) | ||
| Nadir90 | DLM | 0.905 (0.892–0.913) | 0.287 (0.192–0.494) | 0.853 (reference) | 0.118 (reference) |
| LR | 0.900 (0.879–0.926) | 0.298 (0.193–0.572) | 0.833 (<0.001) | 0.110 (0.056) | |
| RF | 0.889 (0.847–0.903) | 0.292 (0.192–0.566) | 0.837 (<0.001) | 0.115 (0.444) | |
| XGB | 0.891 (0.855–0.906) | 0.270 (0.140–0.582) | 0.809 (<0.001) | 0.089 (<0.001) | |
| Fall20 | DLM | 0.864 (0.836–0.888) | 0.794 (0.698–0.847) | 0.872 (reference) | 0.831 (reference) |
| LR | 0.868 (0.840–0.888) | 0.788 (0.700–0.844) | 0.855 (<0.001) | 0.817 (<0.001) | |
| RF | 0.844 (0.812–0.869) | 0.750 (0.688–0.834) | 0.850 (<0.001) | 0.813 (<0.001) | |
| XGB | 0.860 (0.820–0.873) | 0.777 (0.701–0.812) | 0.860 (<0.001) | 0.815 (<0.001) | |
| Fall20/MAP10 | DLM | 0.863 (0.827–0.878) | 0.812 (0.729–0.858) | 0.853 (reference) | 0.841 (reference) |
| LR | 0.857 (0.825–0.873) | 0.804 (0.726–0.854) | 0.842 (<0.001) | 0.827 (<0.001) | |
| RF | 0.838 (0.801–0.859) | 0.773 (0.720–0.827) | 0.843 (<0.001) | 0.827 (<0.001) | |
| XGB | 0.851 (0.812–0.856) | 0.795 (0.735–0.824) | 0.843 (<0.001) | 0.829 (<0.001) | |
The performance measures of internal validation were calculated using 5-folds cross-validation; The min and max are the minimum and maximum values for 5 performance measures obtained through 5-folds cross-validation. P-values were calculated compared to the DLM. The Delong test was used to calculated p-values for comparison of AUROC. The bootstrap method was used to calculated p-values for comparison of AUPRC. IDH, intradialytic hypotension; MAP, mean arterial pressure; AUROC, area under the receiver operating characteristic curve; AUPRC, area under the precision-recall curve; LR, logistic regression; RF, random forest; XGB, extreme gradient boosting; DLM, deep learning model.