Skip to main content

Table 5 Results of original Elixhauser models 1 and enhanced Elixhauser 2 models

From: The predictability of claim-data-based comorbidity-adjusted models could be improved by using medication data

MRDx3 c statistics Hosmer-Lemeshow test
Elixhauser4 Enhanced5 Elixhauser4 Enhanced5
  (95%CI6) (95%CI6) chi-square (P) chi-square (P)
ICH7 0.736 0.748 7.3 (0.50) 7.8 (0.46)
(0.723-0.749) (0.736-0.760)
Pneumonia 0.917 0.920 26.8 (<0.01) 26.8 (<0.01)
(0.912-0.922) (0.915-0.925)
Ischemic infarct 0.787 0.805 14.6 (0.07) 17.2 (0.03)
(0.767-0.807) (0.786-0.824)
AMI8 0.809 0.825 35.8 (<0.01) 35.4 (<0.01)
(0.795-0.823) (0.811-0.839)
Non-alcoholic liver disease 0.798 0.811 24.1 (<0.01) 25.4 (<0.01)
(0.782-0.814) (0.796-0.826)
Intracranial injury 0.778 0.790 12.3 (0.14) 13.5 (0.10)
(0.759-0.797) (0.771-0.809)
CRF9 0.832 0.840 19.3 (0.01) 28.0 (<0.01)
(0.816-0.848) (0.825-0.855)
COPD10 0.810 0.815 8.7 (0.37) 12.2 (0.14)
(0.789-0.831) (0.795-0.835)
Alcoholic liver disease 0.777 0.788 8.9 (0.35) 3.2 (0.92)
(0.753-0.801) (0.764-0.812)
Aspiration pneumonia 0.730 0.734 3.2 (0.92) 8.1 (0.42)
(0.703-0.757) (0.707-0.761)
CHF11 0.699 0.707 3.5 (0.90) 4.7 (0.79)
(0.672-0.726) (0.680-0.734)
Coronary atherosclerosis 0.881 0.889 22.7 (<0.01) 14.2 (0.08)
  (0.862-0.900) (0.869-0.909)   
  1. 1 & 2 Multiple logistic regression models for predicting in-hospital mortalities composed of age + sex + status of health insurance + admission category (emergent or not) + operation (yes or no) + presence of each Elixhauser comorbidity (yes or no), before (Elixhauser models) and after (enhanced Elixhauser models) adding comorbidities inferred by drug prescription information, 3 Most responsive diagnoses, 4 Elixhauser models, 5 Enhanced Elixhauser models, 6 95% confidence interval calculated by bootstrapping, 7 Intracranial hemorrhage, 8 Acute myocardial infarction, 9 Chronic renal failure, 10 Chronic obstructive pulmonary disease, 11 Congestive heart failure.