Effective and accurate screening for diabetic retinopathy using 60° mydriatric fundus camera

Abstract

Objectives. To establish whether an experienced endocrinologist could screen accurately for diabetic retinopathy using mydriatic 60° fundus photographs compared with a reference standard, viz. the combined highest scores of two experienced ophthalmologists. Design. Retrospective review of 60° colour transparency photographs taken over a 6-year period. Retinopathy was graded in a standardised way. Setting. Patients attending the diabetic clinic at Johannesburg Hospital, South Africa. Subjects. Fifteen hundred and seventeen patients (2 446 eyes) formed the basis for the study. Patients were included if there was more than 50% readability of the fundus photographs. Outcome measures. Outcome measures were prevalence of any retinopathy and presence of referable (severe) retinopathy. Inter-observer agreement was measured using the kappa statistic, and sensitivity and specificity of the screener were evaluated. Results. The prevalence of retinopathy at the clinic was approximately 30%, but only about 12% was severe enough to warrant referral to the ophthalmology outpatient department. The endocrinologist was very accurate in determining cases requiring referral; there was 97% agreement with the reference standard, viz. the combined highest score of two experienced ophthalmologists (gold standard). Correlation on the determination of any retinopathy was less accurate (80% agreement), mostly owing to the endocrinologist reporting more isolated microaneurysms than the ophthalmologists. The screening method used gave a sensitivity of 83% and specificity of 99% which are within recommended standards. Conclusions. The screening strategy using a mydriatic fundus camera at the diabetic clinic was found to be effective and accurate and greatly reduced the number of possible referrals to the ophthalmology outpatient department.