Abstract
Purpose:
To assess the levels of agreement in the diagnosis of diabetic retinopathy (DR) between a retina specialist performing indirect dilated stereoscopic examination and optometrists using digital non-mydriatic retinal photography.
Methods:
A retinal photographer used a 45-degree non-mydriatic digital retinal camera to capture macula- and optic disc-centered images from N=469 subjects with diagnosed diabetes. Following photography, subjects were dilated to receive an indirect stereoscopic retinal examination by a retina specialist (RS). The non-mydriatic images were graded by two optometrists (OD1 and OD2) according to the international clinical DR and diabetic macular edema (DME) disease severity scales. Comparisons were made for the five DR severity levels: no DR, mild non-proliferative DR (NPDR), moderate NPDR, severe NPDR, and proliferative DR (PDR); as well as for the presence of DME. Exudates within 500 microns from the fovea were a surrogate for DME. We measured the agreement between digital non-mydriatic retinal photography and the dilated exam using weighted Cohen’s kappa (κ) coefficients. Agreement using κ was also measured for two categories: non-referable DR (no DR and mild NPDR) and referable DR (moderate NPDR and above, and DME).
Results:
The comparison of OD1 vs. RS produced a κ = 0.75±0.06 when comparing all DR levels. The comparison of the refer/non-refer categories of OD1 vs. RS produced a κ = 0.75±0.07. The comparison of OD2 vs. RS for all DR levels produced a κ = 0.79±0.05 and a κ = 0.85±0.06 for the refer/non-refer categories. Similar comparisons for OD1 vs. OD2 produced κ = 0.79±0.05 for all DR categories and a κ = 0.75±0.07 for the refer/non-refer categories.
Conclusions:
Kappa values indicate a high level of agreement between non-mydriatic imaging and dilated exam results. The majority of disagreement occurred when determining mild and moderate NPDR, which can be attributed to the identification of exudates and dot-blot hemorrhages. This study indicates that using non-mydriatic imaging for evaluating diabetic retinopathy compares favorably well with a dilated exam.
Keywords: 499 diabetic retinopathy •
460 clinical (human) or epidemiologic studies: health care delivery/economics/manpower