Table 6 Responsiveness and reliability in predicting length of stay, discharge disposition, and mortality for the best performing three models when trained with patient-specific variables only
| Â | Reliability (Accuracy) | Responsiveness (AUC) | ||||
|---|---|---|---|---|---|---|
Training | Testing | Validation | Training | Testing | Validation | |
LOS | ||||||
 LSVM | 64.19% | 64.11% | 63.98% | 0.646 | 0.649 | 0.642 |
 CHAID | 63.63% | 63.83% | 63.62% | 0.634 | 0.634 | 0.63 |
 Decision List | 85.64% | 85.50% | 85.45% | 0.586 | 0.59 | 0.586 |
Discharge | ||||||
 LSVM | 80.03% | 80.07% | 79.83% | 0.721 | 0.721 | 0.723 |
 CHAID | 80.04% | 80.02% | 79.80% | 0.706 | 0.707 | 0.708 |
 Decision List | 89.88% | 89.87% | 89.59% | 0.648 | 0.649 | 0.651 |
Mortality | ||||||
 XGT Boost Tree | 99.87% | 99.89% | 99.89% | 0.851 | 0.883 | 0.888 |
 LSVM | 99.87% | 99.89% | 99.89% | 0.907 | 0.951 | 0.941 |
 Decision List | 99.90% | 99.90% | 99.91% | 0.845 | 0.925 | 0.871 |