Table 4:

The AUC from prediction of treatment-requiring ROP birth weight and gestational age without or with inclusion of first image session findings

Analysis Approach	BW + GA*			BW + GA + First image session findings *			Difference in AUC (95% CI)			P-value
Two Eyes Analysis	AUC	95% CI	Width	AUC	95% CI	Width	AUC	95% CI	Width
Naïve approach	0.802	0.769, 0.835	0.066	0.878	0.853, 0.903	0.050	0.076	0.049, 0.102	0.053	<0.0001
Nonparametric clustered ROC analysis	0.802	0.755, 0.848	0.093	0.878	0.844, 0.912	0.068	0.076	0.041, 0.111	0.070	<0.0001
Cluster bootstrap-percentile: based in SAS	0.801	0.755, 0.845	0.090	0.878	0.842, 0.908	0.066	0.075	0.042, 0.112	0.070	<0.0001
Cluster bootstrap: BCa in R	0.802	0.753, 0.844	0.091	0.878	0.842, 0.908	0.066	0.076	0.043, 0.112	0.069	NA
Left Eye Analysis (N=771)
Standard logistic regression	0.800	0.753, 0.846	0.093	0.876	0.839, 0.912	0.073	0.076	0.038, 0.114	0.076	<0.0001
Right Eye Analysis (N=771)
Standard logistic regression	0.804	0.757, 0.850	0.093	0.881	0.847, 0.916	0.069	0.078	0.041, 0.114	0.073	<0.0001
Worse Eye Analysis (N=771)
Standard logistic regression	0.800	0.753, 0.846	0.093	0.871	0.835, 0.908	0.073	0.072	0.032, 0.112	0.080	0.0005

BW=birth weight; GA=gestational age; AUC=area under the ROC curve; BCa=Bias-corrected/accelerated; NA=not available.

^*Including the number of quadrants with preplus disease, ROP stage and zone, weight gain, and respiratory support.