Table 3:

Progressive increase of multiple entropy r² value and decrease of AIC with the incremental protocol^a

	Logistic Regression Model				GEEs Method
	r2	AIC	P Value	Odds Ratio (95% CI)	AIC	P Value
NCCT
Model fit statistics	0.015	451.1			451.3
Observed diagnosis (yes vs no)			.28	1.68 (0.65–4.41)		.29
Observed confidence score			.90	1.03 (0.67–1.58)		.90
NCCT + CTA-SI
Model fit statistics	0.170	409.2			408.7
Observed diagnosis (yes vs no)			.59	0.78 (0.31–1.94)		.57
Observed confidence score			<.001b	0.46 (0.31–0.68)		<.001b
NCCT + CTA-SI + CTP
Model fit statistics	0.329	357.3			357.1
Observed diagnosis (yes vs no)			.04b	0.33 (0.11–0.95)		.05
Observed confidence score			<.001b	0.30 (0.20–0.44)		<.001b

Note:—GEEs indicates generalized estimating equations.

^aWith logistic regression analysis and the GEEs method, the actual stroke diagnosis was modelled on different observed diagnoses (NCCT alone, NCCT + CTA-SI, NCCT + CTA-SI + CTP) when adjusting for the corresponding confidence score. OR < 1 indicates that patients with a positive diagnosis on MRI are more likely to have a lower level of confidence (1 = definitely present, 2 = probably present, 3 = possibly present, 4 = possibly absent, 5 = probably absent, 6 = definitely absent).

^bStatistically significant.