Table 1.

Baseline prediction results using ECG on-chip model, and demographics and co-morbidity data. Late fusion model performance—Setting 1: using demographic and ECG data; Setting 2: using demographic, co-morbidity and ECG data.

ML classifier	Accuracy (%)
	1 h	2 h	3 h	4 h	5 h	6 h
ECG model
On-chip ANN	95	87.5	86.2	86	83.7	77.5

ML classifier	Accuracy (%)
EMR model
Linear SVM	49
Logistic Regression	53
Random Forest	76
Neural Network	51

Meta classifier	Accuracy (%)
	1 h	2 h	3 h	4 h	5 h	6 h
Late fusion model
Setting 1
Linear SVM	94	90	86	89	82	76
Logistic Regression	94	89	88	89	84	76
Random Forest	88	88	84	81	74	74
Neural Network	94	91	85	90	84	76
Setting 2
Linear SVM	93	96	95	91	90	86
Logistic Regression	93	96	95	93	90	86
Random Forest	88	96	89	88	88	81
Neural Network	92	95	91	92	89	84

Optimal performance for every prediction task is highlighted in bold.