Novel criteria to classify ARDS severity using a machine learning approach

Table 3 Quality of the third ICU day severity predictive ML models for MIMIC-III

Algorithm	AUC, mean ± SD	CORR, mean ± SD
(a) PaO₂/FiO₂ results
Scenario I: Predicting ARDS Severity in the 3rd ICU day using the data in 1st ICU day
XGBoost	0.616 ± 0.039	0.190 ± 0.068
RF	0.622 ± 0.048	0.173 ± 0.089
LightGBM	0.612 ± 0.039	0.138 ± 0.084
*Scenario II: Predicting ARDS Severity in the 3rd ICU day using the data in 2nd ICU day
XGBoost	0.621 ± 0.023	0.147 ± 0.121
*RF	0.635 ± 0.020	0.139 ± 0.094
LightGBM	0.622 ± 0.025	0.126 ± 0.120
Scenario III: Predicting ARDS Severity in the 3rd ICU day using the data in 1st & 2nd ICU days
XGBoost	0.619 ± 0.030	0.150 ± 0.106
RF	0.627 ± 0.022	0.177 ± 0.108
LightGBM	0.618 ± 0.022	0.086 ± 0.101
(b) P/FP_E results
Scenario I: Predicting ARDS Severity in the 3rd ICU day using the data in 1st ICU day
XGBoost	0.711 ± 0.029	0.385 ± 0.064
RF	0.712 ± 0.027	0.408 ± 0.060
LightGBM	0.716 ± 0.029	0.376 ± 0.073
*Scenario II: Predicting ARDS Severity in the 3rd ICU day using the data in 2nd ICU day
XGBoost	0.785 ± 0.025	0.514 ± 0.053
RF	0.787 ± 0.023	0.546 ± 0.061
*LightGBM	0.788 ± 0.020	0.566 ± 0.044
Scenario III: Predicting ARDS Severity in the 3rd ICU day using the data in 1st & 2nd ICU days
XGBoost	0.782 ± 0.025	0.548 ± 0.049
RF	0.780 ± 0.023	0.538 ± 0.065
LightGBM	0.785 ± 0.021	0.511 ± 0.055

ISSN: 1364-8535