Table 7.

Contribution of the predictors to the prediction ability.

	c-statistics		P-valueb	c-statistics		P-valueb
	Using clinical data and healthcare cost	Using only clinical data		Using clinical data and healthcare cost	Using patient age, gender, and healthcare cost
Reference modela	0.824 (0.813–0.835)	0.708 (0.695–0.721)	<0.001	0.824 (0.813–0.835)	0.821 (0.809–0.833)	0.72
Logistic regression with Lasso regularization	0.824 (0.813–0.835)	0.708 (0.695–0.721)	<0.001	0.824 (0.813–0.835)	0.821 (0.809–0.833)	0.69
Random forest	0.837 (0.826–0.848)	0.738 (0.725–0.751)	<0.001	0.837 (0.826–0.848)	0.816 (0.804–0.828)	0.01
Gradient-boosted decision tree	0.844 (0.833–0.855)	0.716 (0.703–0.728)	<0.001	0.844 (0.833–0.855)	0.841 (0.830–0.852)	0.66
Deep neural network	0.842 (0.831–0.853)	0.716 (0.703–0.729)	<0.001	0.842 (0.831–0.853)	0.839 (0.828–0.851)	0.63

^aWe used a non-penalized logistic regression model as the reference model.

^bWe compared the area under the curve between each machine-learning-based prediction model and the logistic regression model (the reference model) using the DeLong’s test.