Skip to main content

Table 4 Weights of the selected features

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

Feature

Odds ratio (LR)a

Z-score (linear SVM)a

Gini index (RF)b

Donor

Age (per 1 year)

1.060

0.744

0.037

(#9)

BMI (per 1 kg/m2)

0.751

−1.700

0.023

(#20)

Terminal SCr (per 1 mg/dL)

6.512

1.126

0.024

(#17.5)

Hypotensive episodes: yes vs. no

1.784

0.165

0.001

(#48.5)

Diabetes mellitus: yes vs. no

0.013

−1.041

0.001

(#48.5)

History of hypertension: yes vs. no

3.585

0.940

0.011

(#28)

Donor after cardiac death: yes vs. no

25.789

1.534

0.080

(#1)

Preservation/Operation

Machine perfusion: yes vs. no

0.003

−1.078

0.000

(#60)

Perioperative graft reperfusionc

0.740

−0.844

0.027

(#14.5)

Preservation solution

   

 HTK + UW

0.00005

−0.510

0.000

(#60)

 UW

0.080

−1.557

0.016

(#25)

 HTK

0.050

−1.725

0.007

(#32.5)

Male donor-to-female recipient: yes vs. no

0.352

−0.750

0.019

(#23)

Recipient

BMI (per 1 kg/m2)

1.144

0.941

0.054

(#4)

Duration of dialysis (per 1 day)

1.0005

0.324

0.057

(#3)

PRA at time of Tx (per 1 %)

0.977

−0.557

0.008

(#30.5)

Peak PRA (per 1 %)

1.017

0.585

0.025

(#16)

Acute CNI toxicity: yes vs. no

22.044

0.964

0.007

(#32.5)

Reduced cardiac function: yes vs. no

5.570

0.897

0.033

(#13)

Impaired ECV: yes vs. no

0.003

−1.141

0.000

(#60)

Urinary tract obstruction: yes vs. no

6.638

0.942

0.004

(#38.5)

Iliac artery

   

 normal

1.520

0.221

0.001

(#48.5)

 atheromatosis

2.389

0.573

0.006

(#34.5)

 stenosis

28.465

0.948

0.037

(#9)

  1. aFitted on the reduced data set
  2. bFitted on the full data set. Tied rank amongst all 68 features is given in parentheses
  3. cPerioperative graft reperfusion is an ordinal feature (poor – patchy – moderate – good)
  4. Abbreviations: BMI body mass index, CNI calcineurin inhibitor toxicity, ECV effective circulating volume, HTK histidine-tryptophan-ketoglutarate, LR logistic regression, PRA panel reactive antibody, RF random forest, SCr serum creatinine, SVM support vector machine, Tx transplantation, UW University of Wisconsin