Skip to main content
. 2022 Aug 6;6:100071. doi: 10.1016/j.ibmed.2022.100071

Table 3.

Global importance of the top 20 most important predictors for the logistic Regression (LR) and random forest (RF) model, ranked based on mean SHAP magnitude.

Predictor Predictor importance LR model (mean |SHAP|, log-odds scale) Predictor Predictor importance RF model (mean |SHAP|, probability scale)
Age [y] 0.453 pH (arterial) 0.0044
pH (arterial) 0.333 SpO2/FiO2 0.0037
FiO2 [%] 0.332 FiO2 [%] 0.0034
Sodium [mmol/L] 0.257 PaO2/FiO2 [mmHg] 0.0026
Haemoglobin [mmol/L] 0.211 SpO2 [%] 0.0017
SpO2 [%] 0.207 Potassium [mmol/L] 0.0016
Heart rate [bpm] 0.181 Age [y] 0.0014
Chloride [mmol/L] 0.175 SBP [mmHg] 0.0013
Potassium [mmol/L] 0.160 Base excess [mmol/L] 0.0011
PaO2 (arterial) [mmHg] 0.131 PaCO2 (arterial) [mmHg] 0.0010
SBP [mmHg] 0.130 White cell count [109/L] 0.0008
PaCO2 (arterial) [mmHg] 0.116 ICU length of stay [hours] 0.0008
Platelet Count [109/L] 0.116 Creatinine [μmol/L] 0.0006
White cell count [109/L] 0.114 Heart rate [bpm] 0.0006
ICU length of stay [hours] 0.112 PaO2 (arterial) [mmHg] 0.0006
SpO2/FiO2 0.111 ASAT [U/L] 0.0006
Urea Creatinine ratio 0.092 Urea [mmol/L] 0.0006
Ionised calcium [mmol/L] 0.078 Haemoglobin [mmol/L] 0.0005
Haematocrit 0.069 Ionised calcium [mmol/L] 0.0005
Respiratory rate [/min] 0.068 Respiratory rate [/min] 0.0005