Skip to main content

Table 2 Discrimination of ANN and CLR in modeling and testing datasets with 16- and 6-variable models

From: Hip fracture risk assessment: artificial neural network outperforms conditional logistic regression in an age- and sex-matched case control study

 

1

2

3

4

5

6

7

8

9

10

Mean

SD

p

Modeling

             

AUROC

             

ANN 16v

0.888

0.867

0.866

0.869

0.864

0.880

0.873

0.861

0.886

0.866

0.872

0.009

0.005*

CLR 16v

0.835

0.828

0.829

0.832

0.824

0.850

0.837

0.828

0.840

0.836

0.834

0.007

ANN 6v

0.849

0.839

0.839

0.842

0.831

0.853

0.848

0.832

0.837

0.837

0.841

0.007

0.005*

CLR 6v

0.826

0.825

0.818

0.826

0.816

0.836

0.828

0.815

0.821

0.823

0.823

0.006

Accuracy

             

ANN 16v

0.805

0.790

0.790

0.785

0.785

0.810

0.805

0.785

0.815

0.805

0.797

0.011

0.005*

CLR 16v

0.769

0.763

0.761

0.771

0.769

0.781

0.774

0.774

0.778

0.768

0.771

0.006

ANN 6v

0.770

0.761

0.760

0.786

0.775

0.780

0.775

0.760

0.770

0.765

0.770

0.008

0.005*

CLR 6v

0.753

0.746

0.746

0.769

0.763

0.766

0.758

0.751

0.758

0.753

0.756

0.008

Sensitivity

             

ANN 16v

0.790

0.760

0.800

0.800

0.860

0.820

0.830

0.780

0.820

0.790

0.805

0.027

0.444

CLR 16v

0.770

0.857

0.779

0.872

0.779

0.892

0.856

0.830

0.785

0.781

0.820

0.044

ANN 6v

0.780

0.840

0.820

0.860

0.840

0.800

0.850

0.840

0.860

0.810

0.830

0.025

0.959

CLR 6v

0.724

0.704

0.867

0.888

0.872

0.887

0.882

0.876

0.728

0.796

0.822

0.072

Specificity

             

ANN 16v

0.820

0.820

0.780

0.770

0.710

0.800

0.780

0.790

0.810

0.820

0.790

0.032

0.012*

CLR 16v

0.767

0.668

0.742

0.668

0.758

0.670

0.691

0.718

0.772

0.755

0.721

0.041

ANN 6v

0.760

0.680

0.700

0.710

0.710

0.760

0.700

0.680

0.680

0.720

0.710

0.028

0.368

CLR 6v

0.782

0.788

0.624

0.648

0.655

0.644

0.634

0.626

0.788

0.708

0.690

0.067

Testing

             

AUROC

             

ANN 16v

0.815

0.894

0.905

0.890

0.955

0.792

0.876

0.948

0.773

0.836

0.868

0.059

0.005*

CLR 16v

0.769

0.773

0.853

0.825

0.872

0.721

0.824

0.891

0.707

0.772

0.801

0.059

ANN 6v

0.806

0.878

0.865

0.807

0.908

0.777

0.842

0.948

0.838

0.866

0.854

0.048

0.005*

CLR 6v

0.778

0.793

0.863

0.801

0.845

0.758

0.810

0.904

0.817

0.800

0.817

0.041

Accuracy

             

ANN 16v

0.765

0.811

0.836

0.811

0.768

0.701

0.741

0.840

0.680

0.729

0.768

0.053

0.017*

CLR 16v

0.767

0.744

0.814

0.698

0.791

0.674

0.698

0.816

0.614

0.705

0.732

0.062

ANN 6v

0.743

0.860

0.857

0.676

0.743

0.651

0.697

0.906

0.730

0.795

0.766

0.081

0.028*

CLR 6v

0.721

0.698

0.791

0.674

0.698

0.698

0.698

0.811

0.682

0.727

0.720

0.043

Sensitivity

             

ANN 16v

0.760

0.760

0.860

0.760

0.910

0.730

0.770

0.830

0.680

0.760

0.782

0.063

0.759

CLR 16v

0.857

0.857

0.864

0.762

0.818

0.727

0.773

0.783

0.591

0.762

0.779

0.077

ANN 6v

0.810

0.860

0.950

0.620

0.860

0.680

0.770

0.960

0.820

0.900

0.823

0.104

0.575

CLR 6v

0.762

0.524

1.000

0.810

0.864

0.773

0.773

0.870

0.591

0.810

0.777

0.129

Specificity

             

ANN 16v

0.770

0.860

0.810

0.860

0.620

0.670

0.710

0.850

0.680

0.700

0.753

0.084

0.066

CLR 16v

0.682

0.636

0.762

0.636

0.762

0.619

0.619

0.850

0.636

0.652

0.685

0.075

ANN 6v

0.680

0.860

0.760

0.730

0.620

0.620

0.620

0.850

0.640

0.700

0.708

0.087

0.202

CLR 6v

0.682

0.864

0.571

0.545

0.524

0.619

0.619

0.750

0.773

0.652

0.660

0.103

  1. * Statistically significant difference.