Table 4 Performance statistics of the different machine learning models, and EKFCcrea+cys-equation, in the external validaton dataset.

From: Comparison between the EKFC-equation and machine learning models to predict Glomerular Filtration Rate

External validation cohort (n = 5389)

Methods

Median bias (95%CI)

IQR (P25–P75)

P10 (95%CI)

P30 (95%CI)

MSE

EKFC

−0.60 [−0.90; −0.30]

15.60 (−7.9,5.4)

0.45 [0.44; 0.46]

0.89 [0.88; 0.90]

247.98

RF

−0.71 [−1.09; −0.27]

15.01 (−8.03,5.65)

0.44 [0.43; 0.45]

0.88 [0.87; 0.89]

227.3

LR

2.73 [2.06; 3.31]

23.49 (−10.4,14.47)

0.30 [0.28; 0.31]

0.68 [0.67; 0.69]

551.78

XGBoost

−1.38 [−1.83; −1.08]

15.58 (−9.02,5.9)

0.4 [0.39; 0.41]

0.86 [0.85; 0.87]

245.83

  1. EKFC = European Kidney Function Consortium equation; RF = Random Forest model; LR = Linear Regression model; XGBoost = eXtreme Gradient Boosting model; IQR = Interquartile range; P10/P30 = fraction of patients with eGFR within 10%/30% of mGFR; MSE = Mean Square Error.
Back to article page