Skip to main content

Table 2 Model Performance in predicting death up to 7 days post-discharge, as measured by AUC, AIC and BIC

From: Predicting the cumulative risk of death during hospitalization by modeling weekend, weekday and diurnal mortality risks

Model type Variables used Number of variables AUC AIC BIC Optimal cut-off point (ratio of deaths to survival) Sensitivity Specificity PPV NPV Accuracy
Model 1: Background risk Age, Gender 2 .711 (.708 -.715) 1.0710e + 006 1.0710e + 006 0.196 71.29 66.14 2.78 99.41 66.21
Model 2: Time exposure risk 2(a): Length of stay (LOS) 1 .710 (.705 - .714) 1.3155e + 006 1.3155e + 006 0.168 69.07 73.69 3.45 99.43 73.62
2(b): LOS (Weekday, weekend), admission time 3 .692 (.687 - .697) 1.1960e + 006 1.1961e + 006 0.145 71.97 70.21 3.18 99.5 70.23
2(c): LOS (Morning, evening, or night for each of seven days), admission time 22 .786 (.782 - .79) 1.3464e + 006 1.3467e + 006 0.228 64.62 74.57 3.34 99.4 74.44
Model 3: Disease model Charlson comorbidity index 1 .786 (.783 - .79) 1.0826e + 006 1.0826e + 006 0.1 91.54 66.27 3.56 99.8 66.60
Model 4: Background plus time exposure risk 4(a): 1 + 2(a) 3 .73 (.726 - .739) 0.9408e + 006 0.94097e + 006 0.196 76.14 61.25 2.6 99.5 61.5
4(b): 1 + 2(b) 5 .743 (.739 - .747) 1.1519e + 006 1.1520e + 006 0.227 79.95 68.72 3.36 99.6 68.87
4(c): 1 + 2(c) 24 .883 (.88 - .886) 1.4713e + 006 1.4716e + 006 0.32 82.41 77.41 4.73 99.7 77.47
Model 5: Full model 5(a): 1 + 2(a) + 3 4 .891 (.888 - .894) 1.5143e + 006 1.5143e + 006 0.197 87.09 78.32 5.18 99.8 78.43
5(b): 1 + 2(b) + 3 6 .893 (.89 - .896) 1.5118e + 006 1.5119e + 006 0.224 86.86 78.27 5.16 99.8 78.38
5(c): 1 + 2(c) + 3 25 .923 (.921 - .926) 1.6769e + 006 1.6772e + 006 0.271 88.18 81.61 6.13 99.8 81.71
Model 6: DRG-specific models 5(c) trained and tested only single DRG subpopulation R61: Lymphoma and Non-Acute Leukemia 25 .90 (.878 - .922) 4.7669e + 003 4.9223e + 003 0.149 81.45 82.63 34.67 97.5 82.51
E02: Other Respiratory System OR Procedures 25 .940 (.911 - .969) 2.6019e + 003 2.7281e + 003 0.398 80.0 95.25 21.05 99.6 94.99
F70: Major Arrhythmia and Cardiac Arrest 25 .762 (.736 - .802) 982.0188 1.1044e + 003 0.412 75.86 65.79 68.75 73.3 70.81
E64: Pulmonary Oedema and Respiratory Failure 25 .85 (.807 - .894) 466.95 476.32 0.616 71.34 88.41 87.5 73.05 79.32
J62: Malignant Breast Disorders 25 .903 (.874 - .933) 704.03 713.95 0.438 77.69 86.49 74.26 88.54 83.55
Model 7: Emergency Department admission 5(c) trained and tested only on ED or non ED subpopulations ED admission 25 .909 (.905 - .912) 3.0017e + 005 3.0044e + 005 0.16 87.80 77.32 11.91 99.45 77.67
Non-ED admissions 25 .925 (.921 - .928) 1.4626e + 006 1.4629e + 006 0.218 87.70 84.38 4.47 99.9 84.41
  1. An optimal cut-off value is selected as the point in the receiver operating characteristic curve where the sum of sensitivity and specificity is maximum. Sensitivity, specificity, model accuracy, positive predictive value (PPV) and negative predictive value (NPV) are reported at the optimal ratio of patient deaths to survival.