Table 2 Discrimination and calibration metrics of clinical scores for predicting mortalities.

From: Prognostic performance of clinical scores and biomarkers in nonagenarians after TAVR requiring ICU admission

Variable

Outcome

AUC

CI lower

CI upper

Threshold

Sensitivity

Specificity

Cutoff

Youden index

APACHE III

ICU mortality

0.993

0.980

1.000

0.017

1.000

0.979

80.5

0.979

SOFA

ICU mortality

0.969

0.938

1.000

0.013

1.000

0.943

6.5

0.943

APACHE III

Hospital Mortality

0.916

0.834

0.998

0.026

0.778

0.898

69.5

0.676

SOFA

Hospital mortality

0.922

0.859

0.985

0.012

0.889

0.802

4.5

0.691

APACHE III

1-month mortality

0.749

0.604

0.894

0.038

0.538

0.897

69.5

0.435

SOFA

1-Month Mortality

0.723

0.563

0.883

0.018

0.692

0.650

3.5

0.342

APACHE III

6-month mortality

0.672

0.574

0.770

0.055

0.618

0.638

57.5

0.256

SOFA

6-month mortality

0.669

0.572

0.765

0.056

0.618

0.658

3.5

0.276

APACHE III

1-year mortality

0.641

0.559

0.722

0.142

0.321

0.895

68.5

0.216

SOFA

1-year mortality

0.635

0.554

0.715

0.096

0.571

0.665

3.5

0.236

APACHE III

Long-term mortality

0.608

0.559

0.658

0.320

0.600

0.561

54.5

0.161

SOFA

Long-term mortality

0.648

0.601

0.695

0.287

0.692

0.518

2.5

0.210

  1. Area under the receiver operating characteristic curve (AUROC) with 95% confidence intervals (CIs) was used to quantify discrimination. Thresholds were determined by maximizing the Youden Index (sensitivity + specificity – 1), representing the cut-off point that best differentiates high versus low risk. Sensitivity and specificity were calculated at the respective cutoff points. Mortality was assessed at ICU discharge, hospital discharge, and fixed timepoints (1 month, 6 months, 1 year, and long-term follow-up).
Back to article page