Table 4. Results of pair-wise feature selection in the training cohort.

	AUC (LD)	BA (LD)	AUC (RF)	BA (RF)
Age	0.742	0.685	0.753	0.694
Sex	0.748	0.684	0.757	0.693
Smoking History	0.745	0.684	0.754	0.688
Healthcare worker	0.743	0.685	0.754	0.692
Alcohol use disorder	0.743	0.686	0.754	0.690
Pregnancy	0.742	0.685	0.754	0.692
Hypertension	0.743	0.689	0.754	0.698
Diabetes	0.748	0.691	0.757	0.698
Rheumatic diseases	0.742	0.686	0.755	0.697
Autoimmune disorder	0.743	0.687	0.755	0.696
Dementia	0.747	0.690	0.755	0.690
Cancer	0.743	0.686	0.755	0.687
COPD	0.743	0.688	0.755	0.692
Asthma	0.743	0.686	0.755	0.690
CHD	0.742	0.685	0.753	0.693
CCD	0.741	0.688	0.752	0.690
CND	0.745	0.696	0.759	0.700
CLD	0.743	0.685	0.754	0.688
CKD	0.743	0.686	0.754	0.687
AIDS	0.743	0.686	0.753	0.696
Cachexia	0.744	0.687	0.754	0.694
Organ transplant	0.743	0.686	0.753	0.694

The first row shows the performance of age alone, and subsequent rows show the performance of a pair of features, where age is always one half of the pair, e.g., Sex means Age + Sex and Diabetes means Age + Diabetes. The third significant figure is included to show minute differences (not statistically significant). COPD, chronic obstructive pulmonary disease; CHD, chronic hematologic disease; CCD, chronic cardiac disease; CND, chronic neurological disease; CLD, chronic liver disease; CKD, chronic kidney disease; AIDS, Acquired Immune Deficiency Syndrome; AUC, area under the Receiver Operating Characteristic curve; BA, balanced accuracy; LD, linear discriminant; RF, random forest.