Skip to main content

Table 8 Optimal feature vectors selected by different models from bootstrap test data

From: A comparative analysis of predictive models of morbidity in intensive care unit after cardiac surgery – Part II: an illustrative example

No. BL BQ k NN LR HS DS ANN1 ANN2
1 O2ER O2ER Post-CI O2ER O2ER SvO2 O2ER O2ER
2 VCO2 DO2 O2ER VCO2 VCO2 Card-ID Card-ID VO2
3 Card-ID Card-ID Card-ID Card-ID Card-ID DO2 VO2 Card-ID
4 PVD   PVD PVD PVD O2ER PVD PVD
5 TBU   TBU TBU TBU EM TBU Gly
6 EM    EM EM BSA Pre-CI Gender
7 SAP    SAP SAP AD EM MVR
8 SaO2    Pre-CI Pre-CI CHF WBC Cr
9     WBC WBC MVR Age AVO2
10     SaO2 SaO2 MR SaO2 Arrhy
11     PvO2 PvO2 PVD AD  
12     AD AD Diab P/F  
13     PaO2 PaO2 VCO2 PVS  
14     Cr Cr CABG-C   
15       PAH   
16       IABP   
  1. BL, Bayes linear model; BQ, Bayes quadratic model; k NN, k-nearest neighbour model; LR, logistic regression model; HS, Higgins score system; DS, direct score system; ANN1, one-layer artificial neural network; ANN2, two-layer artificial neural network. Predictor variable abbreviations are indicated in Tables 1-6.