Skip to main content

Table 3 Variable importance for detecting sepsis

From: Deep-learning model for screening sepsis using electrocardiography

Rank Logistic regression (deviance difference) Random forest (mean decrease Gini) Deep learning (relative importance)
1 QTc (492) Heart rate (473.2) QTc (0.193)
2 Age (274) T-wave axis (472.1) QT interval (0.168)
3 QRS duration (207) R-wave axis (443.2) PR interval (0.121)
4 T-wave axis (145 QTc (429.5) T-wave axis (0.085)
5 QT interval (101) P-wave axis (413.3) QRS duration (0.082)
6 Heart rate (63) Age (394.5) Age (0.079)
7 P-wave axis (18) QRS duration (386.2) Heart rate (0.078)
8 R-wave axis (11) QT interval (367.0) P-wave axis (0.075)
9 PR interval (2) PR interval (363.0) R-wave axis (0.063)
10 Sex (− 1) Sex (0.1) Sex (0.055)