Utilizing visual symptoms to distinguish dry eye from glaucoma, cataract, and suspect glaucoma patients: a cross-sectional study

Table 2 Sets of visual symptoms best distinguishing dry eye from glaucoma suspect, glaucoma, and cataract

Dry Eye vs. Glaucoma Suspect ^aLikelihood Ratio Test p < 0.001 ^bCross-Validated Mean AUC (SD) = 0.84 (0.067) ^cSensitivity = 78%; Specificity = 83%
	Odds Ratio	95% CI	P value
Light Sensitivity	14.95	6.25–35.74	< 0.001
Spots in Vision	2.75	1.20–6.31	0.017
Dry Eye vs. Glaucoma Likelihood ratio Test p < 0.001 Cross-Validated Mean AUC (SD) = 0.93 (0.026) Sensitivity = 89%; Specificity = 87%
	Odds Ratio	95% CI	P value
Light Sensitivity	9.19	2.03–41.68	0.004
Poor Peripheral Vision	0.21	0.059–0.72	0.013
Better Vision in One Eye	0.087	0.010–0.72	0.023
Patches of Vision Missing	0.055	0.009–0.33	0.001
Dry Eye vs. Cataract Likelihood Ratio Test p < 0.001 Cross-Validated Mean AUC (SD) = 0.93 (0.047) Sensitivity = 86%; Specificity = 86%
	Odds Ratio	95% CI	P value
Spots in Visions	4.51	1.52–13.42	0.007
Vision that Varies Across the Week	4.67	1.23–17.72	0.024
Worsening Vision	0.096	0.025–0.37	0.001
Blindness	0.13	0.020–0.80	0.029

AUC area under the receiver operator curve, SD cross-validated AUC standard deviation
^aLikelihood ratio test comparing a model with all symptom frequency variables remaining after backward stepwise selection including all demographic variables, and the nested model with only demographic variables
^bCross-validated mean AUC computed with fivefold split of data in the model of significant frequency variables remaining after backward stepwise selection with all demographic covariates
^cSensitivity and specificity of multivariable model to predict dry eye diagnosis over another study disease with predicted probability set to 50%

ISSN: 1471-2415