Machine learning identifies ICU outcome predictors in a multicenter COVID-19 cohort

Table 2 Overall performance of the machine learning models for ICU outcome prediction

Prediction variable	ICU survival N = 1053		ECMO therapy N = 1053		Renal Replacement Therapy N = 1000
Prediction variable	Balanced accuracy	PR-AUC	Balanced accuracy	PR-AUC	Balanced accuracy	PR-AUC
ML model
RF	0.65	0.84	0.72	0.75	0.69	0.66
SVC	0.65	0.81	0.79	0.53	0.7	0.64
EBM	0.61		0.72		0.68
EBM (10 interactions)	0.64	0.81	0.73	0.69	0.7	0.69

EBM, explainable boosting machine; ICU, intensive care unit; ML, machine learning; RF, random forest classifier; SVC, support vector classifier

ISSN: 1364-8535