Prediction of delayed graft function after kidney transplantation: comparison between logistic regression and machine learning methods

Table 3 Performance of the statistical methods after 10-fold stratified cross-validation

Statistical method	Sensitivity (%)		PPV (%)		AUROC (%)
Statistical method	No DGF	DGF	No DGF	DGF	AUROC (%)
Decision tree	75.4 ± 6.64	29.5 ± 16.29	88.2 ± 2.73	14.2 ± 8.13	52.5 ± 8.55
Gradient boosting	98.8 ± 1.55	16.2 ± 12.94	89.2 ± 1.67	58.3 ± 38.19	77.2 ± 9.64
Random forest	96.3 ± 4.05	16.4 ± 14.92	89.0 ± 2.09	43.9 ± 38.19	73.9 ± 9.94
Random forest (full)	100.0 ± 0.00	0.0 ± 0.00	87.5 ± 0.64	0.0 ± 0.00	71.6 ± 12.38
LDA	94.7 ± 2.92	27.6 ± 15.10	90.2 ± 2.00	42.3 ± 19.94	82.2 ± 6.14
QDA	89.9 ± 5.35	37.6 ± 17.26	91.0 ± 2.55	37.9 ± 20.82	79.6 ± 7.55
Linear SVM	72.0 ± 6.29	83.8 ± 7.51	96.9 ± 1.34	30.6 ± 5.60	84.3 ± 4.11
Radial SVM	57.9 ± 7.45	88.8 ± 7.38	97.2 ± 1.87	23.6 ± 4.14	83.3 ± 4.05
Polynomial SVM	97.5 ± 1.90	10.9 ± 12.20	88.5 ± 1.14	24.0 ± 24.17	79.8 ± 5.33
Logistic regression	65.0 ± 8.25	85.5 ± 8.94	96.9 ± 1.84	26.5 ± 4.75	81.7 ± 5.82

Abbreviations: AUROC area under the receiver operating characteristic curve, DGF delayed graft function, LDA linear discriminant analysis, PPV positive predictive value, QDA quadratic discriminant analysis, SVM support vector machine

ISSN: 1472-6947