Table 3: Ten highest contribution features identified from the best-performing EI model for predicting mortality due to COVID-19.

The table provides detailed information on the features, including their names, the modality they were included in, and their description. The features are sorted in terms of increasing rank product scores (third column), which is the EI interpretation metric our method calculates (Section 2.2.)

Modality	Feature	Rank product score (RPS)	Description
Admission	Age	0.023	A patient’s age at the time of hospital admission.
Laboratory Tests	Chloride	0.051	The amount of chloride in a patient’s blood sample.
Vital Signs	Minimum O2 Saturation	0.055	The minimum oxygen saturation recorded during the first 36 hours of hospitalization.
Laboratory Tests	Blood Urea Nitrogen	0.072	The amount of nitrogen from the waste product urea in a patient’s blood sample.
Laboratory Tests	Calcium	0.088	The amount of calcium in a patient’s blood sample.
Admission	Respiratory Rate	0.093	The respiratory rate recorded at the time of hospital admission.
Laboratory Tests	Sodium	0.118	The amount of sodium in a patient’s blood sample.
Laboratory Tests	Venous PCO2	0.120	The partial pressure of carbon dioxide measured in a patient’s venous blood.
Admission	O2 Saturation	0.120	A patient’s oxygen saturation recorded at the time of hospital admission.
Co-morbidities	Atrial Fibrillation	0.123	A diagnosis of atrial fibrillation at the time of hospital admission.