Table 3.

Prediction model and performance for the modified clinical severity score.

Response	Variable selection method	Variables, n	Sample size			Model		Training				Testing
			Training	Testing			Area under the curve		Sensitivity	Specificity	Area under the curve		Sensitivity	Specificity
y1a	Stepwise	16	2643	1331	Logistic regression		0.865		0.831	0.745	0.853		0.775	0.797
					Random forest		0.958		0.888	0.888	0.841		0.765	0.792
					Support vector machine		0.87		0.783	0.806	0.856		0.83	0.75
y1	Least absolute shrinkage and selection operator	5	2686	1354	Logistic regression		0.85		0.755	0.794	0.847		0.745	0.812
					Random forest		0.905		0.767	0.847	0.83		0.71	0.843
					Support vector machine		0.854		0.77	0.787	0.848		0.739	0.84
y2b	Stepwise	14	2931	1459	Logistic regression		0.891		0.803	0.807	0.864		0.82	0.765
					Random forest		0.901		0.832	0.85	0.834		0.757	0.837
					Support vector machine		0.887		0.834	0.771	0.854		0.868	0.687
y2	Least absolute shrinkage and selection operator	6	2683	1348	Logistic regression		0.881		0.757	0.832	0.877		0.812	0.793
					Random forest		0.943		0.87	0.841	0.865		0.772	0.842
					Support vector machine		0.886		0.835	0.768	0.879		0.812	0.816
y3c	Stepwise	11	2931	1460	Logistic regression		0.939		0.952	0.795	0.94		0.982	0.766
					Random forest		0.931		0.871	0.925	0.863		0.807	0.91
					Support vector machine		0.933		0.944	0.789	0.935		0.842	0.895
y3	Least absolute shrinkage and selection operator	6	2691	1357	Logistic regression		0.923		0.86	0.873	0.944		0.884	0.88
					Random forest		0.991		0.984	0.949	0.933		0.895	0.865
					Support vector machine		0.918		0.812	0.918	0.943		0.874	0.906

^ay₁: mild vs above moderate.

^by₂: below oderate vs above severe.

^cy₃: below severe vs critical.