Table 3.

AUC^a, optimum cutoff, Sn^b, Sp^c, positive predictive value, negative predictive value, and diagnostic accuracy for AI^d scores (probability and ALA^e) and BSS^f in discriminating 86% (258/300) of cases of moderate to critical disease from 14% (42/300) of cases of mild disease^g.

	AUC, mean (SD)	P valueh	Value, 95% CI	Cutoff	Sn, %	Sp, %	Acci, %	PPVj, %	NPVk, %
AI probability score	0.787 (0.033)	<.001	0.722-0.852	≥56	68.2	78.6	69.7	95.1	28.7
AI ALA score	0.857 (0.024)	<.001	0.809-0.905	≥1	84.5	73.8	83.0	95.2	43.7
BSS	0.863 (0.023)	<.001	0.818-0.908	≥1	76.7	95.2	79.3	99.0	40.0

^aAUC: area under the curve.

^bSn: sensitivity.

^cSp: specificity.

^dAI: artificial intelligence.

^eALA: affected lung area.

^fBSS: Brixia scoring system.

^gInterpretation: <0.60: fail; 0.60 to 0.70: poor classification; 0.70 to 0.80: fair classification; 0.80 to 0.90: good classification; 0.9 to 1: excellent classification.

^hP<.05 was considered statistically significant and emphasized by bold texts.

ⁱAcc: accuracy.

^jPPV: positive predictive value.

^kNPV: negative predictive value.