Machine-learning predictions for acute kidney injuries after coronary artery bypass grafting: a real-life muticenter retrospective cohort study

Table 3 Comparison of NRI and IDI between LightGBM model and the other three models (SVM, RF, Softmax)

		NRI			IDI
		Value	95%CI	P	Value	95%CI	P
Internal validation group
	LightGBM-SVM	0.5852	0.3651 to 0.8053	0	0.0144	-0.0022 to 0.0311	0.0895
	LightGBM-RF	0.8988	0.686 to 1.1116	0	0.1632	0.1168 to 0.2097	0
	LightGBM-Softmax	0.6847	0.4665 to 0.9028	0	0.1082	0.0708 to 0.1455	0
External validation group of QLH
	LightGBM-SVM	0.4398	0.2651 to 0.6146	0	0.0101	-0.0008 to 0.021	0.0696
	LightGBM-RF	1.0959	0.9317 to 1.2601	0	0.2017	0.1673 to 0.02361	0
	LightGBM-Softmax	0.3704	0.1966 to 0.5441	0	0.019	-0.0133 to 0.0513	0.2489
External validation group of GHN
	LightGBM-SVM	0.2237	-0.004 to 0.4514	0.0541	-0.027	-0.0558 to 0.0019	0.067
	LightGBM-RF	0.6735	0.4017 to 0.9452	0	0.1012	0.0494 to 0.153	0.0001
	LightGBM-Softmax	0.4722	0.1919 to 0.7525	0.0009	0.0248	-0.0153 to 0.0649	0.2259

Abbreviation: NRI, net reclassification improvement; IDI, integrated discrimination improvement;

ISSN: 1472-6947