Table 4 Performance of the mortality prediction models tested in the Korean Acute Myocardial Infarction Registry-National Institutes of Health.

From: Machine learning enhances the performance of short and long-term mortality prediction model in non-ST-segment elevation myocardial infarction

	STEMI					NSTEMI
	AUC (95% CI)	Specificity	Sensitivity	Accuracy	F1-score	AUC (95% CI)	Specificity	Sensitivity	Accuracy	F1-score
In-hospital mortality						In-hospital mortality
Machine learning algorithms						Machine learning algorithms
LR with Lasso	0.923 (0.897–0.948)	0.877	0.807	0.876	0.124	0.916 (0.891–0.941)	0.848	0.787	0.847	0.096
LR with Ridge	0.923 (0.898–0.948)	0.755	0.982	0.757	0.081	0.918 (0.894–0.942)	0.868	0.770	0.867	0.107
LR with Elastic net	0.923 (0.898–0.948)	0.884	0.754	0.883	0.123	0.917 (0.893–0.941)	0.845	0.803	0.845	0.096
RF	0.924 (0.897–0.952)	0.815	0.860	0.816	0.092	0.924 (0.903–0.946)	0.860	0.803	0.860	0.106
SVM	0.875 (0.844–0.907)	0.772	0.807	0.773	0.072	0.848 (0.815–0.880)	0.723	0.852	0.725	0.060
XGBoost	0.938 (0.920–0.955)	0.855	0.860	0.855	0.114	0.911 (0.885–0.937)	0.832	0.787	0.832	0.088
Traditional model						Traditional model
TIMI	0.866 (0.820–0.913)	0.774	0.807	0.774	0.072	0.672 (0.612–0.731)	0.693	0.590	0.692	0.038
GRACE	0.921 (0.891–0.950)	0.851	0.825	0.850	0.107	0.917 (0.890–0.944)	0.799	0.852	0.800	0.081
12-month mortality						12-month mortality
Machine learning algorithms						Machine learning algorithms
LR with Lasso	0.789 (0.719–0.860)	0.751	0.696	0.750	0.048	0.815 (0.781–0.848)	0.727	0.720	0.726	0.100
LR with Ridge	0.789 (0.718–0.859)	0.636	0.761	0.637	0.037	0.809 (0.774–0.843)	0.735	0.695	0.735	0.099
LR with Elastic net	0.789 (0.721–0.858)	0.721	0.696	0.721	0.044	0.814 (0.780–0.847)	0.749	0.695	0.748	0.104
RF	0.772 (0.702–0.843)	0.572	0.826	0.575	0.034	0.792 (0.751–0.832)	0.746	0.703	0.745	0.104
SVM	0.687 (0.606–0.768)	0.425	0.804	0.429	0.025	0.721 (0.676–0.765)	0.662	0.695	0.663	0.080
XGBoost	0.796 (0.736–0.857)	0.701	0.717	0.701	0.042	0.808 (0.773–0.843)	0.783	0.653	0.781	0.111
Traditional model						Traditional model
TIMI	0.701 (0.633–0.769)	0.624	0.804	0.626	0.038	0.676 (0.635–0.717)	0.693	0.590	0.692	0.038
GRACE	0.738 (0.671–0.806)	0.650	0.761	0.651	0.038	0.778 (0.741–0.814)	0.799	0.852	0.800	0.081

AUC, area under the receiver operating characteristic curve; LR, Logistic regression; Lasso, L1 penalty; Ridge, L2 penalty; Elastic net, Elastic net penalty; RF, Random Forest; SVM, Support Vector Machine; XGBoost, Extreme Gradient Boosting.

Back to article page

Table 4 Performance of the mortality prediction models tested in the Korean Acute Myocardial Infarction Registry-National Institutes of Health.

Search

Quick links