Table 3.

Assessment of predictive performance for prediction of 90-day mortality using EHR features in the external validation set

Models	AUROC	AUPRC	Sensitivity	Specificity	PPV	NPV	F1-score
LR	0.62 (0.49 to 0.74)	0.6 (0.43 to 0.75)	0.71 (0.56 to 0.85)	0.48 (0.33 to 0.63)	0.56 (0.42 to 0.71)	0.63 (0.47 to 0.80)	0.63 (0.50 to 0.74)
SVM	0.64 (0.52 to 0.76)	0.63 (0.46 to 0.78)	0.71 (0.56 to 0.85)	0.43 (0.28 to 0.57)	0.54 (0.39 to 0.67)	0.61 (0.43 to 0.78)	0.61 (0.48 to 0.72)
RF	0.69 (0.58 to 0.80)	0.66 (0.50 to 0.81)	0.84 (0.72 to 0.94)	0.43 (0.28 to 0.56)	0.58 (0.45 to 0.70)	0.74 (0.56 to 0.90)	0.69 (0.57 to 0.78)
XGB	0.75 (0.64 to 0.85)	0.74 (0.58 to 0.86)	0.82 (0.69 to 0.92)	0.6 (0.45 to 0.75)	0.66 (0.52 to 0.79)	0.77 (0.62 to 0.91)	0.73 (0.61 to 0.83)
MLP	0.71 (0.59 to 0.82)	0.70 (0.53 to 0.84)	0.84 (0.72 to 0.95)	0.33 (0.19 to 0.47)	0.54 (0.41 to 0.67)	0.68 (0.56 to 0.75)	0.66 (0.54 to 0.76)
LGB	0.66 (0.54 to 0.78)	0.66 (0.49 to 0.80)	0.89 (0.79 to 0.97)	0.48 (0.33 to 0.63)	0.62 (0.49 to 0.75)	0.83 (0.67 to 0.96)	0.73 (0.62 to 0.82)

All numbers are presented with 95% CI.

AUPRC, area under the precision-recall curve; AUROC, area under the receiver operating characteristic curve; LGB, light gradient boosting; LR, logistic regression; MLP, multilayer perceptron; NPV, negative predictive value; PPV, positive predictive value; RF, random forest; SVM, support vector machine; XGB, extreme gradient boosting.