Skip to main content

Table 2 Model performance for logit with addition of log HLOS and ROD (NC: not computed)

From: Modelling hospital outcome: problems with endogeneity

Model title Logit1: lnday Logit1: lnrod Logit2: lnday Logit2: lnrod Logit3:lnday Logit3:lnrod
Index
 ROC AUC 0.921 0.934 0.923 0.936 0.917 0.935
 H-L statistic; P-value 0.000 0.060 0.000 0.013 0.000 0.054
 Out-of-sample shrinkage % 0.940 0.020 −1.370 0.750   
 In-sample-shrinkage % 0.380 −0.500 −2.350 −0.380   
 Overfitting % 0.560 0.510 0.960 1.120   
 Calibration belt: P-value 0.000 NC 0.000 0.000 0.000 0.000
 AIC 31,306.41 30,574.13 31,077.06 30,492.24 31,148.98 30,523.52
 BIC 322,421.1 31,489.52 33,039.96 32,455.96 32,073.81 31,448.36
Development set
 CITL      0.005 0.017
 C-slope      1.005 1.004
 AUC      0.923 0.935
 E:O ratio      0.997 0.998
Validation set
 CITL      0.005 0.017
 C-slope      1.005 1.004
 AUC      0.923 0.935
 E:O ratio      0.997 0.990