Skip to main content
. Author manuscript; available in PMC: 2022 Apr 1.
Published in final edited form as: Ann Emerg Med. 2021 Jan 15;77(4):395–406. doi: 10.1016/j.annemergmed.2020.11.007

Table 2.

Top 20 most important features in septic shock prediction.

No. Feature
Development cohort and corresponding change in AUC ROC if removed
 1 SBP (0.062)
 2 BUN (0.019)
 3 Resp rate (0.012)
 4 Temp (0.008)
 5 ΔSBP (0.007)
 6 Hct (0.007)
 7 WBC (0.007)
 8 Lactate (0.007)
 9 Creatinine (0.007)
 10 HR (0.006)
 11 Calcium (0.004)
 12 DBP (0.003)
 13 Potassium (0.002)
 14 O2 sat (0.001)
 15 ΔO2 sat (0.001)
 16 ΔTemp (0.001)
 17 MAP (0.001)
 18 Δetco2 (0.001)
 19 ΔBase excess (0.001)
 20 ΔHCO3 (0.001)

Validation cohort
 1 SBP (0.049)
 2 HR (0.015)
 3 Temp (0.015)
 4 BUN (0.013)
 5 Hct (0.011)
 6 Lactate (0.009)
 7 Resp rate (0.005)
 8 Bilirubin total (0.005)
 9 Creatinine (0.003)
 10 DBP (0.003)
 11 Alkaline phos (0.003)
 12 Magnesium (0.002)
 13 Age (0.002)
 14 Time in ED (0.002)
 15 ΔSBP (0.001)
 16 pH (0.001)
 17 ΔBUN (0.001)
 18 ΔWBC (0.001)
 19 ΔFibrinogen (0.001)
 20 Men (0.001)

SBP, Systolic blood pressure; Resp,, respiratory; Temp, temperature; HR, pulse rate; DBP, diastolic blood pressure; O2 sat, oxygen saturation; ETCO2, end tidal carbon dioxide; HCO3, bicarbonate.

The numbers in parentheses show the amount of change in AUC if a feature is treated as missing in the model.