Table 4

Summary of AIC and AUROCs of fitted Cox and Poisson regression models – training dataset

Predictors/model1-year all-cause mortality1-year CV/diabetes-related mortalityAny-cause hospitalisation1-year CV/diabetes-related hospitalisation*1-year aggregated any hospitalisation or mortality
1-year hospitalisation (single event)*Recurrent eventPoisson† (count in first year)
Demographics-only model
 Model 1AUROC=0.7865
(AIC=43 050)
AUROC=0.7962
(AIC=13 382)
AIC=995 523AIC=2.27e+07AIC=477 990AIC=704 241AUROC=0.6055
(AIC=1 008 481)
Severity score+demographics models
Model 2 (model 1+ever before severity score)AUROC=0.7912
(AIC=42 586.52)
AUROC=0.8030
(AIC=13 091.66)
AIC=992 032.8AIC=2.26e+07AIC=456 825AIC=699 582.5AUROC=0.6271
(AIC=1 004 865)
Model 3 (model 1+10-year severity score)AUROC=0.7912
(AIC=42 649.12)
AUROC=0.8032
(AIC=13 133)
AIC=992 096.3AIC=2.26e+07AIC=457 753AIC=699 635AUROC=0.6270
(AIC=1 004 939)
Model 4 (model 1 +
5-year severity score)
AUROC=0.7910
(AIC=42 732.13)
AUROC=0.8024
(AIC=13 203.79)
AIC=992 244.2AIC=2.27e+07AIC=460 109AIC=
699 853.5
AUROC=0.6265
(AIC=1 005 101)
  • *Competing risk analysis.

  • †Adjusted for age, gender and IMD only.

  • AIC, Akaike information criterion; AUROC, area under a Receiver Operating Characteristics curve; CV, cardiovascular; DM, diabetes mellitus.