Table 3.

Performance metrics of sepsis prediction models

Training set		Real	Synthetic	Synthetic
Testing set		Real	Synthetic	Real
Train	Accuracy	0.845	0.869	0.852
	Precision	0.803	0.840	0.812
	Recall	0.704	0.758	0.719
	F1	0.750	0.797	0.763
	AUROC	0.809	0.842	0.818
5-fold cross-validation	Accuracy	0.795	0.802	0.799
	Precision	0.712	0.73	0.723
	Recall	0.637	0.67	0.639
	F1	0.672	0.69	0.678
	AUROC	0.855	0.86	0.847
Test	Accuracy	0.834	0.833	0.834
	Precision	0.811	0.759	0.829
	Recall	0.677	0.678	0.654
	F1	0.738	0.716	0.731
	AUROC	0.887	0.885	0.892