Skip to main content

Table 4 Performance of DLM for predicting in-hospital mortality of patients with infectious diseases

From: Deep-learning model for screening sepsis using electrocardiography

Predictive model AUC (95% CI) SEN (95% CI) SPE (95% CI) PPV (95% CI) NPV (95% CI)
DLM using 12-lead ECG 0.817 (0.793–0.840) 0.785 (0.735–0.835) 0.710 (0.697–0.724) 0.137 (0.120–0.155) 0.983 (0.978–0.987)
DLM using 6-lead ECG 0.815 (0.794–0.836) 0.844 (0.799–0.888) 0.682 (0.668–0.695) 0.135 (0.118–0.152) 0.987 (0.983–0.991)
DLM using 1-lead ECG 0.802 (0.780–0.825) 0.945 (0.917–0.973) 0.524 (0.509–0.539) 0.105 (0.092–0.117) 0.994 (0.991–0.997)
SOFA score 0.817 (0.786–0.847) 0.734 (0.680–0.788) 0.785 (0.773–0.797) 0.167 (0.145–0.189) 0.980 (0.976–0.985)
Quick SOFA score 0.797 (0.767–0.828) 0.594 (0.534–0.654) 0.885 (0.875–0.894) 0.233 (0.200–0.265) 0.974 (0.969–0.979)
NEWS 0.808 (0.777–0.839) 0.738 (0.684–0.792) 0.771 (0.759–0.784) 0.160 (0.139–0.180) 0.980 (0.976–0.985)
MEWS 0.778 (0.747–0.808) 0.773 (0.722–0.825) 0.670 (0.656–0.684) 0.121 (0.105–0.137) 0.981 (0.976–0.985)
Lactate 0.801 (0.758–0.844) 0.646 (0.569–0.724) 0.854 (0.829–0.879) 0.466 (0.397–0.534) 0.925 (0.905–0.944)
WBC 0.591 (0.552–0.630) 0.504 (0.442–0.566) 0.679 (0.665–0.694) 0.086 (0.072–0.101) 0.958 (0.951–0.965)
C-reactive protein 0.541 (0.499–0.583) 0.300 (0.242–0.359) 0.811 (0.799–0.823) 0.084 (0.065–0.103) 0.953 (0.946–0.960)
  1. AUC area under the receiver operating characteristic curve; ECG electrocardiography; MEWS Modified Early Warning Score; NEWS National Early Warning Score; NPV negative predictive value; PPV positive predictive value; SEN sensitivity; SOFA sequential organ failure assessment; SPE specificity