Table 3.

Performances of models using logistic regression, classification and regression tree, conditional inference tree, and random forest.

	C-Statistics	Kappa	Acc and 95% CI	Sen	Spe	PPV	NPV
LRM	0.765 (0.728–0.801)	0.350	70.71 (67.08–74.16)	49.60	83.78	65.45	72.86
CART	0.690 (0.652–0.728)	0.375	71.32 (67.70~74.75)	55.16	81.33	64.65	74.55
CARTcat	0.638 (0.603–0.672)	0.298	69.35 (65.67–72.85)	40.08	87.47	66.45	70.22
CIT	0.751 (0.714–0.788)	0.379	71.32 (67.70–74.75)	56.75	80.34	64.13	75.00
CITcat	0.759 (0.722–0.796)	0.334	69.95 (66.29–73.43)	48.81	83.05	64.06	72.38
RF	0.753 (0.715–0.791)	0.355	71.02 (67.39–74.46)	49.21	84.52	66.31	72.88

Note: No information rate (NIR), 61.76. All models showed better accuracies than NIR. Abbreviations: LRM, logistic regression model; CART, classification and regression tree; CARTcat, classification and regression tree using categorically transformed variables; CIT, conditional inference tree; CITcat, conditional inference tree using categorically transformed variables; RF, random forest; Kappa, Cohen’s kappa; Acc, accuracy; CI, confidence intervals; Sen, sensitivity; Spe, specificity; PPV, positive predictive value; NPV, negative predictive value.