Accuracy of single-field nonmydriatic digital fundus image in screening for diabetic retinopathy.

Objective: To determine the accuracy, the sensitivity and the specificity of a single-field nonmydriatic digital fundus image interpreted by an endocrinologist for diabetic retinopathy (DR) screening.

Material And Method: Two hundred and forty-eight diabetic patients who attended the Diabetic Center BMA Medical College and Vajira Hospital between May 2007 and March 2008 were included in the present study. The fundus images of all patients, which would include optic nerve and macular area, were captured by a digital camera without any mydriatic agent. After image taking, the patients were subsequently examined for any evidence of diabetic retinopathy by an experienced ophthalmologist. The fundus images were later interpreted by a trained endocrinologist and would be compared with the findings from the ophthalmologist, which were used as a gold standard

Results: The prevalence of DR was 24.2% of the population or 22.8% of the 495 eyes studied. Ninety-three fundus images were considered low quality for interpretation and were excluded from the analysis. From the remaining 402 eyes (155 patients), the Kappa value of the endocrinologist's interpretation and the ophthalmologist's findings was 0.48. The accuracy for screening DR by the image capture was 80.6% (95% confidence interval [CI], 76.4-84.3) while the sensitivity and specificity were 65.6% (95% CI, 60.9-70.2) and 84.9% (95% CI, 81.4-88.4), respectively. Positive predictive value and negative predictive value were 55.7% (95% CI, 50.8-60.5) and 89.5% (95% CI, 86.5-92.5), respectively

Conclusion: Single-field nonmydriatic digital fundus image is a convenient screening tool for a diagnosis of diabetic retinopathy. The test could be achieved by a trained endocrinologist who could practically serve the patients in one visit at diabetic clinics. A referral to an ophthalmologist is still recommended in any cases with abnormal findings, or those with questionable findings, and those with poor quality photographs when diabetic retinopathy could not be definitely excluded.