Table 2 Predictive performance of AI model, ISS, and ICISS.

From: External validation of an artificial intelligence model using clinical variables, including ICD-10 codes, for predicting in-hospital mortality among trauma patients: a multicenter retrospective cohort study

Model

Sensitivity

Specificity

Accuracy

Balanced accuracy

Precision

F1-score

AUROC

Total dataset in both hospitals

 AI model

0.8995

0.8021

0.8069

0.8508

0.9539

0.8954

0.9448

 ISS-16

0.7808

0.8665

0.8623

0.8237

0.9498

0.8951

0.8807

 ISS-25

0.5388

0.9642

0.9432

0.7515

0.9493

0.9460

0.8807

 ICISS

0.7671

0.7440

0.7452

0.7556

0.9421

0.8169

0.7978

Patients with ISS < 9 in both hospitals

 AI model

0.4000

0.9372

0.9340

0.6686

0.9905

0.9604

0.8979

 ISS

0.0000

1.0000

0.9940

0.5000

0.9881

0.9911

0.5455

 ICISS

0.0000

1.0000

0.9940

0.5000

0.9881

0.9911

0.5000

Patients with ISS ≥ 9 in both hospitals

 AI model

0.9363

0.6056

0.6407

0.7709

0.9062

0.7089

0.9143

 ISS-16

0.8431

0.6725

0.6906

0.7578

0.8947

0.7498

0.8171

 ISS-25

0.5784

0.9122

0.8768

0.7453

0.894

0.8840

0.8171

 ICISS

0.6618

0.7376

0.7296

0.6997

0.8722

0.7781

0.7164

Total dataset at CHH

 AI model

0.8469

0.7919

0.7937

0.8598

0.9656

0.8598

0.9234

 ISS-16

0.6327

0.8733

0.8656

0.7530

0.9592

0.9041

0.8168

 ISS-25

0.4592

0.9645

0.9483

0.7119

0.9483

0.9535

0.8168

 ICISS

0.5204

0.8416

0.8313

0.6810

0.9535

0.8824

0.6905

Patients with ISS < 9 at CHH

 AI model

0.4615

0.9334

0.9298

0.6975

0.9883

0.9568

0.8987

 ISS-16

0.0000

1.0000

0.9923

0.5000

0.9847

0.9885

0.5110

 ISS-25

0.0000

1.0000

0.9923

0.5000

0.9847

0.9885

0.5110

 ICISS

0.0000

1.0000

0.9923

0.5000

0.9847

0.9885

0.5000

Patients with ISS ≥ 9 at CHH

 AI model

0.9059

0.6056

0.6244

0.7558

0.9363

0.7190

0.8969

 ISS-16

0.7294

0.7066

0.7080

0.7180

0.9232

0.7831

0.7621

 ISS-25

0.5294

0.9178

0.8936

0.7236

0.9254

0.9069

0.7621

 ICISS

0.5647

0.7300

0.7197

0.6474

0.9095

0.7908

0.6446

Total dataset at CNUH

 AI model

0.9421

0.8262

0.8364

0.8842

0.9363

0.8670

0.9653

 ISS-16

0.9091

0.8508

0.8559

0.8799

0.9355

0.8809

0.9351

 ISS-25

0.6033

0.9635

0.9319

0.7834

0.9314

0.9317

0.9351

 ICISS

0.8512

0.7762

0.7828

0.8137

0.9193

0.8267

0.8629

Patients with ISS < 9 at CNUH

 AI model

0.0000

0.9451

0.9428

0.4725

0.995

0.9682

0.8761

 ISS-16

0.0000

1.0000

0.9976

0.5000

0.9951

0.9963

0.6868

 ISS-25

0.0000

1.0000

0.9976

0.5000

0.9976

0.9963

0.6868

 ICISS

0.0000

1.0000

0.9976

0.5000

0.9976

0.9963

0.5000

Patients with ISS ≥ 9 at CNUH

 AI model

0.9580

0.6054

0.6804

0.7817

0.8571

0.7088

0.9303

 ISS-16

0.9244

0.5737

0.6482

0.7490

0.8389

0.6789

0.8381

 ISS-25

0.6134

0.8957

0.8357

0.7546

0.8357

0.8357

0.8381

 ICISS

0.7227

0.7959

0.7804

0.7593

0.8237

0.7940

0.7593

  1. AI, artificial intelligence; ISS, Injury Severity Score; ICISS, ICD-10–based Injury Severity Score; AUROC, area under the receiver operating characteristic curve; CHH, Cheju Halla General Hospital; CNUH, Chonnam National University Hospital.
Back to article page