Table 3.

Performance of classification models predicting severity of COVID-19 from laboratory findings (Abu Dhabi dataset) or individual risk factors (Dubai dataset) with and without age and sex.

ML method		Prediction of severity from					Prediction of disease severity from
		laboratory findings, age and sex					individual risk factors inclusively age and sex
		Precision	Recall	F1	AUC	Acc	Precision	Recall	F1	AUC	Acc
AdaBoost	w/o	0.63	0.64	0.62	0.6444	0.64	0.72	0.73	0.72	0.641	0.73
	with	0.64	0.64	0.63	0.6585	0.64	0.72	0.74	0.73	0.7222	0.74
ExtraTrees	w/o	0.68	0.68	0.65	0.6658	0.68	0.73	0.76	0.74	0.6824	0.76
	with	0.71	0.69	0.67	0.6728	0.69	0.74	0.76	0.75	0.7452	0.76
Random forest	w/o	0.72	0.69	0.66	0.6633	0.69	0.73	0.78	0.73	0.709	0.78
	with	0.72	0.69	0.66	0.6887	0.69	0.74	0.78	0.73	0.7998	0.78
NN	w/o	0.70	0.71	0.70	0.7547	0.71	0.76	0.79	0.76	0.7208	0.79
	with	0.79	0.77	0.76	0.7906	0.77	0.80	0.81	0.80	0.8134	0.81
SVM (linear)	w/o	0.66	0.66	0.63	0.6465	0.66	0.68	0.76	0.69	0.6656	0.76
	with	0.67	0.67	0.64	0.6534	0.67	0.71	0.76	0.71	0.7806	0.76
LR	w/o	0.65	0.66	0.64	0.655	0.66	0.74	0.78	0.74	0.7239	0.78
	with	0.66	0.66	0.65	0.6643	0.66	0.74	0.78	0.74	0.7863	0.78
Gain after adding age and sex to predictors		+3%	+1.33%	+1.83%	+1.64%	+1.33%	+1.5%	+0.5%	+1.33%	+8.41%	+0.5%

Significant differences (p < 0.05) between models performance with and without such predictors as age and sex are marked in bold font.

AUC, area under the receiver operating characteristic curve; Acc, accuracy; LR, logistic regression; ML, machine learning; SVM, support vector machine; w/o, without.