Comparison of nonmydriatic digital retinal imaging with dilated clinical eye exam for detecting non–diabetes–related pathology in diabetic patients

Abstract

Abstract: : Purpose:Prior studies showed that nonmydriatic digital retinal imaging with Joslin Vision Network (JVN) accurately assesses level of diabetic retinopathy compared with ETDRS 7–standard field photography and retinal specialist clinical exam (CE). This study characterizes JVN's diagnostic accuracy compared to CE for detecting non–diabetes–related ocular pathology in diabetic (DM) patients. Methods:JVN imaging was performed on 952 DM patients new to Joslin Diabetes Center's outpatient clinic from Jan 1–Jul 31, 2003. Of these, 295 (31%) had a subsequent dilated eye CE by a Joslin retinal specialist. Retrospective chart review compared CE and JVN findings for all available charts (n=260). Results:Of 260 pts, 43.5% were female, 69.2% had type 2 DM, mean patient age was 50.0±14.9yrs and mean DM duration was 10.6±9.6yrs. Median interval between JVN and CE was 24 days. JVN identified non–DM ocular pathology in 151/520 total eyes (29%). JVN agreement with CE for presence of lesions included retinal emboli (100%, 3/3), asteroid hyalosis (100%, 1/1), renal/hypertensive retinopathy (88%, 14/16), epiretinal membrane (ERM) (71%, 5/7), macular RPE/drusen changes (57%, 12/21), chorio–retinal (CR) atrophy/scar (57%, 4/7), cup/disc asymmetry (50%, 4/8), choroidal lesions (46%, 6/13) and cataract (38%, 35/91). Agreement for either presence or absence of these lesions was 100%, 100%, 95%, 99%, 91%, 99%, 92%, 97% & 81% respectively. Sensitivity ranged from 60% (C/D asymmetry) to 100% (emboli, asteroid hyalosis). Specificity ranged from 93.3% (macular RPE changes) to 100% (emboli, asteroid hyalosis). Kappa values ranged from fair agreement (Κ=0.2–0.4) for C/D asymmetry & macular RPE changes to substantial agreement (Κ>0.6) for asteroid hyalosis, CR atrophy/scar & ERM. Sensitivity, specificity, positive, and negative predictive values for detecting any of the lesions noted here was 0.64, 0.98, 0.67 & 0.98 respectively. Eight eyes (1.5%) had clinically significant lesions outside JVN evaluated regions, including iris rubeosis (3), lattice (3) & operculated retinal hole (2). Conclusions:The sensitivity and specificity of JVN nonmydriatic digital retinal imaging for identifying non–diabetes–related ocular pathology in DM patients is substantial. However, important lesions may be missed when these lie outside retinal fields imaged by JVN.

Keywords: diabetes • diabetic retinopathy • clinical (human) or epidemiologic studies: health care delivery/economics/manpower