Skip to main content

Table 2 The prediction performance of each model in internal validation and external validation sets

From: Interpretable machine learning model for early prediction of 28-day mortality in ICU patients with sepsis-induced coagulopathy: development and validation

Model

AUROC

AUPRC

ACC

SE

SP

PPV

NPV

F1 score

DeLong test for AUC

Internal validation set

 XGBoost

0.828

0.807

0.785

0.646

0.904

0.739

0.2

0.63

Reference

 LR

0.809

0.605

0.770

0.546

0.886

0.712

0.209

0.62

P < 0.001*

 NB

0.768

0.499

0.733

0.481

0.862

0.643

0.236

0.55

P < 0.001*

 SVM

0.808

0.610

0.774

0.504

0.914

0.751

0.219

0.60

P < 0.001*

 SOFA

0.715

0.507

0.713

0.310

0.921

0.671

0.279

0.42

P < 0.001*

 SAPS II

0.746

0.517

0.722

0.397

0.889

0.649

0.259

0.49

P < 0.001*

 SIC score

0.602

0.443

0.664

0.212

0.964

0.431

0.336

0.35

P < 0.001*

External validation set (MIMIC-IV)

 XGBoost

0.913

0.796

0.885

0.523

0.974

0.828

0.107

0.69

Reference

 LR

0.841

0.589

0.847

0.403

0.956

0.692

0.133

0.58

P < 0.001*

 NB

0.814

0.479

0.807

0.511

0.879

0.510

0.200

0.51

P < 0.001*

 SVM

0.839

0.591

0.842

0.347

0.964

0.700

0.142

0.54

P < 0.001*

 SOFA

0.773

0.452

0.713

0.310

0.921

0.671

0.279

0.48

P < 0.001*

 SAPS II

0.763

0.446

0.816

0.192

0.970

0.606

0.170

0.47

P < 0.001*

 SIC score

0.599

0.437

0.803

0.121

0.982

0.478

0.197

0.35

P < 0.001*

External validation set (EICU)

 XGBoost

0.923

0.921

0.891

0.555

0.976

0.842

0.102

0.70

Reference

 LR

0.831

0.797

0.847

0.403

0.956

0.692

0.133

0.58

P < 0.001*

 NB

0.808

0.744

0.744

0.628

0.829

0.728

0.247

0.50

P < 0.001*

 SVM

0.831

0.799

0.770

0.645

0.861

0.772

0.231

0.58

P < 0.001*

 SOFA

0.724

0.628

0.665

0.518

0.772

0.624

0.313

0.43

P < 0.001*

 SAPS II

0.775

0.704

0.718

0.518

0.772

0.624

0.313

0.48

P < 0.001*

 SIC score

0.592

0.647

0.579

0.638

0.536

0.501

0.330

0.35

P < 0.001*

  1. LR regression model, NB naive bayes, SVM support vector machine, SOFA sequential organ failure assessment, SAPS II simplified acute physiology score II, AUROC Area Under the Receiver Operating Characteristic Curve, AUPRC Area Under the Precision-Recall Curve
  2. * indicates a statistical difference