Table 4.

Predictive performance of blending models, SAPS II and refitted SOFA on test set

	Accuracy [95% CI]	Recall [95% CI]	Precision [95% CI]	F1 score [95% CI]	AUPRC [95% CI]	AUROC [95% CI]	P-value for comparison between AUROCs
BM_total	0.845 [0.844, 0.846]	0.203 [0.190, 0.216]	0.748 [0.737, 0.758]	0.318 [0.302, 0.335]	0.536 [0.534, 0.539]	0.815 [0.808, 0.822]	vs. BM_reg: p = 0.251
BM_reg	0.845 [0.844, 0.845]	0.192 [0.183, 0.201]	0.768 [0.759, 0.776]	0.306 [0.295, 0.318]	0.542 [0.540, 0.544]	0.817 [0.810, 0.823]	vs. BM_SAPSII: p < 0.001 vs. BM_SOFA: p < 0.001
BM_SAPSII	0.842 [0.841, 0.844]	0.198 [0.179, 0.216]	0.726 [0.706, 0.746]	0.309 [0.286, 0.331]	0.515 [0.511, 0.518]	0.806 [0.799, 0.813]	vs. SAPSII: p < 0.001
BM_SOFA	0.835 [0.834, 0.836]	0.154 [0.145, 0.163]	0.679 [0.668, 0.690]	0.251 [0.239, 0.262]	0.428 [0.426, 0.431]	0.742 [0.734, 0.749]	vs. SOFA: p < 0.001
SAPS II	0.798	0.458	0.438	0.448	0.429	0.771 [0.764, 0.779]
Refitted SOFA	0.833	0.130	0.680	0.218	0.381	0.706 [0.698, 0.714]

Abbreviations: AUPRC area under the precision-recall curve, AUROC area under the receiver operating characteristic curve, CI confidence interval, SAPS II Simplified Acute Physiology Score II, SOFA Sequential Organ Failure Assessment, BM_total blending model based on the total variables, BM_reg blending model based on variables selected by stepwise regression model, BM_SAPSII blending model based on SAPS II related variables, BM_SOFA blending model based on SOFA related variables