Can we predict which patients are at risk of having an ungradeable digital image for screening for diabetic retinopathy?

Abstract

We aimed to determine the reasons for, and variables which predicted, ungradeable retinal photographs during screening patients for diabetic retinopathy. Age, duration of diabetes, visual acuity, and HbA1c were recorded. Following dark adaptation, a single 45 degrees nonmydriatic photograph was taken of each fundus. The pupils were then dilated and the photograph repeated. Using slit lamp biomicroscopy, lenticular changes (LOCS III), and fundus appearance were recorded. In ungradeable photographs the fovea could not be visualised in 98% of cases of images from nonmydriatic photography, and in 88% if mydriasis was used. Poor definition in the nonmydriatic image was associated with a subsequent ungradeable mydriatic photograph (P=0.001), however, the positive predictive value was poor (34%). Age, posterior subcapsular cataract, and near vision predicted ungradeable status of nonmydriatic photographs (P<0.001, P=0.004, P=0.006, respectively; regression analysis). Nuclear colour and poor definition of the nonmydriatic photograph predicted ungradeable status of mydriatic photographs (P=0.006 & P=0.001, respectively). Inability to visualise the fovea is the commonest cause of an ungradeable image from digital retinal photography. Age and posterior subcapsular cataract were best predictors of ungradeable status of nonmydriatic fundus photographs. Nuclear colour was the strongest predictor for ungradeable mydriatic photography.