Table 4.
Performance metrics between developed models developed with only postoperative blood test values.
| Machine Learning Models | |||||
|---|---|---|---|---|---|
| LASSO | DNN | RF | XGB | ||
| AUROC | Train | 0.81 | 0.82 | 0.77 | 0.90 |
| Test | 0.77 | 0.79 | 0.75 | 0.80 | |
| External validation * | 0.70 | 0.72 | 0.82 | 0.75 | |
| AUPRC | Train | 0.35 | 0.31 | 0.31 | 0.53 |
| Test | 0.26 | 0.27 | 0.29 | 0.30 | |
| External validation * | 0.09 | 0.08 | 0.13 | 0.09 | |
AUROC, area under receiver operating curve; AUPRC, area under precision recall curve; SASA, Surgical Apgar Score combined with the ASA-PS classification; DNN, deep neural network; RF, random forest; XGB, extreme gradient boosting. * External validation was performed on AUSOM DB, while train and test were performed on VitalDB.