Table 2.

Performance metrics of the developed machine learning models, along with SOFA^a score and SAPS II^b,^c.

Model	AUROCd, mean (95% CI)	AUPRCe, mean (95% CI)	Cohen κ, mean (95% CI)	F1-score, mean (95% CI)
VCf	0.861 (0.853-0.869)g,h	0.589 (0.564-0.614)	0.413 (0.404-0.421)	0.554 (0.550-0.558)
CBCi	0.860 (0.852-0.868)g,h	0.590 (0.570-0.610)	0.400 (0.383-0.417)	0.546 (0.536-0.557)
RFCj	0.855 (0.848-0.863)g,h	0.577 (0.551-0.603)	0.392 (0.380-0.404)	0.540 (0.534-0.547)
RLRCk	0.823 (0.813-0.832)g,h	0.497 (0.570-0.525)	0.359 (0.348-0.369)	0.515 (0.509-0.521)
SAPS II	0.749 (0.742-0.756)g	0.438 (0.414-0.462)	0.280 (0.263-0.297)	0.451 (0.440-0.462)
SOFA	0.588 (0.566-0.609)h	0.284 (0.264-0.304)	0.121 (0.096-0.147)	0.330 (0.311-0.349)

^aSOFA: Sequential Organ Failure Assessment.

^bSAPS II: Simplified Acute Physiology Score II.

^cValues were calculated from 5-fold cross-validation. Hypothesis tests were conducted to determine whether the AUROC values of the models using machine learning algorithms were equal to those of conventional scores.

^dAUROC: area under the receiver operating characteristics curve.

^eAUPRC: area under the precision-recall curve.

^fVC: voting classifier.

^gP<.001 compared to SOFA score.

^hP<.001 compared to SAPS II.

ⁱCBC: CatBoost classifier.

^jRFC: random forest classifier.

^kRLRC: regularized logistic regression classifier.