Skip to main content

Table 2 Performance of DLM for screening sepsis and septic shock using electrocardiography

From: Deep-learning model for screening sepsis using electrocardiography

Prediction model

Internal validation

External validation

AUC (95% CI)

SEN (95% CI)

SPE (95% CI)

PPV (95% CI)

NPV (95% CI)

AUC (95% CI)

SEN (95% CI)

SPE (95% CI)

PPV (95% CI)

NPV (95% CI)

Performance for screening sepsis

DLM using 12-lead ECG

0.901

(0.882–0.920)

0.904

(0.854–0.953)

0.776

(0.767–0.785)

0.067

(0.055–0.078)

0.998

(0.997–0.999)

0.863

(0.846–0.879)

0.765

(0.725–0.804)

0.810

(0.805–0.816)

0.083

(0.075–0.091)

0.994

(0.992–0.995)

DLM using 6-lead ECG

0.882

(0.856–0.907)

0.815

(0.749–0.880)

0.824

(0.816–0.833)

0.076

(0.062–0.089)

0.996

(0.994–0.998)

0.856

(0.841–0.872)

0.794

(0.756–0.832)

0.766

(0.760–0.772)

0.071

(0.064–0.078)

0.994

(0.993–0.995)

DLM using 1-lead ECG

0.874

(0.850–0.899)

0.933

(0.891–0.975)

0.688

(0.678–0.698)

0.050

(0.042–0.059)

0.998

(0.997–0.999)

0.845

(0.827–0.863)

0.796

(0.759–0.834)

0.746

(0.739–0.752)

0.066

(0.059–0.072)

0.994

(0.993–0.995)

C-reactive protein

0.723

(0.670–0.776)

0.638

(0.550–0.725)

0.763

(0.748–0.778)

0.091

(0.071–0.111)

0.983

(0.977–0.988)

0.741

(0.716–0.766)

0.607

(0.561–0.654)

0.783

(0.776–0.791)

0.087

(0.077–0.097)

0.983

(0.981–0.986)

Body temperature

0.671

(0.621–0.722)

0.679

(0.599–0.759)

0.667

(0.655–0.679)

0.041

(0.032–0.049)

0.990

(0.987–0.993)

0.671

(0.642–0.700)

0.669

(0.624–0.714)

0.601

(0.593–0.608)

0.044

(0.039–0.049)

0.985

(0.983–0.987)

Performance for screening septic shock

DLM using 12-lead ECG

0.906

(0.877–0.936)

0.895

(0.815–0.974)

0.822

(0.814–0.831)

0.036

(0.026–0.045)

0.999

(0.998–1.000)

0.899

(0.872–0.925)

0.807

(0.743–0.870)

0.875

(0.871–0.880)

0.046

(0.038–0.054)

0.998

(0.998–0.999)

DLM using 6-lead ECG

0.881

(0.843–0.918)

0.842

(0.747–0.937)

0.792

(0.783–0.801)

0.029

(0.021–0.037)

0.999

(0.998–0.999)

0.893

(0.868–0.917)

0.880

(0.828–0.932)

0.749

(0.743–0.755)

0.026

(0.021–0.030)

0.999

(0.998–0.999)

DLM using 1-lead ECG

0.879

(0.841–0.916)

0.930

(0.864–0.996)

0.684

(0.673–0.694)

0.021

(0.016–0.027)

0.999

(0.999–1.000)

0.860

(0.833–0.888)

0.773

(0.706–0.840)

0.801

(0.795–0.806)

0.028

(0.023–0.033)

0.998

(0.997–0.999)

C-reactive protein

0.676

(0.590–0.762)

0.745

(0.625–0.865)

0.573

(0.556–0.590)

0.027

(0.019–0.036)

0.993

(0.989–0.997)

0.724

(0.680–0.768)

0.585

(0.505–0.665)

0.781

(0.774–0.788)

0.030

(0.023–0.036)

0.994

(0.992–0.995)

Body temperature

0.659

(0.584–0.734)

0.685

(0.561–0.809)

0.663

(0.651–0.674)

0.017

(0.011–0.022)

0.996

(0.994–0.998)

0.680

(0.631–0.730)

0.697

(0.622–0.773)

0.596

(0.588–0.604)

0.016

(0.013–0.019)

0.995

(0.994–0.997)

  1. AUC area under the receiver operating characteristic curve; DLM deep learning-based model; ECG electrocardiography; NPV negative predictive value; PPV positive predictive value; SEN sensitivity; SPE specificity