Skip to main content

Table 2 Comparative performance for severe complications risk prediction models

From: Prognostic models for short-term annual risk of severe complications and mortality in patients living with type 2 diabetes using a national medical claim database

 

Training sets

C-statistic

Validation sets

C-statistic

95%CI

p-value:

Se

Sp

Balance accuracy

vs. LR

vs. RF

CV complications

       

LR

0.738

0.715

0.658–0.772

Ref

0.1271

73%

60%

0.665

RF

0.944

0.786

0.747–0.825

0.1271

Ref

71%

72%

0.717

NN

0.739

0.738

0.686–0.79

0.3317

0.1787

70%

67%

0.685

Other complications

       

LR

0.738

0.706

0.653–0.759

Ref

< 0.0001

59%

70%

0.645

RF

0.981

0.847

0.800-0.894

< 0.0001

Ref

80%

81%

0.801

NN

0.732

0.670

0.626–0.714

0.2279

< 0.0001

74%

52%

0.630

All-cause mortality

       

LR

0.823

0.814

0.769–0.859

Ref

0.0776

76%

73%

0.745

RF

0.941

0.860

0.821–0.899

0.0776

Ref

80%

79%

0.794

NN

0.845

0.841

0.791–0.891

0.2354

0.3486

82%

71%

0.765

  1. Note: CV, Cardiovascular; LR, Logistic Regression; NN, Neural Network; RF, Random Forest
Back to article page

Diabetology & Metabolic Syndrome

ISSN: 1758-5996

Contact us