Evaluation of diabetic retinopathy screening programme in general out-patient clinics in Hong Kong

Abstract

Objective The main objective is to evaluate the current diabetic retinopathy (DR) screening by optometrists under Risk Assessment and Management Programme (RAMP) in general outpatient clinics (GOPC), and to compare it to conventional screening by clinical examination. The secondary objective is to predict the prevalence of DR and maculopathy (DMac) in the study cohort, and identify risk factors especially the role of nephropathy.

Methodology Ophthalmologist’s re-grading of the digital fundus photos previously screened by optometrists according to RAMP protocol was used as the gold standard to evaluate the current screening programme. Accuracy of optometrist screening was calculated by percentage of agreement and Kappa coefficient. Fundus photo grading by ophthalmologist and optometrist was compared to clinical examination findings in eye clinics. Sensitivities and specificities were calculated, and plotted on ROC curve for comparison of the two methods of screening. Prevalences of DR and DMac were estimated from the gold standard grading, and their correlation with other factors also screened by RAMP was identified using chi-square test and logistic regression.

Results There was an overall over-grading of disease by optometrists. The overall inter-observer agreement in diagnosis was 81.2%, and the overall kappa coefficient was 0.65 (p<0.001), which reached substantial strength of agreement. Use of mydriatics reduced the percentage of ungradable photos by at least 4 times. The overall agreement of clinical examination with ophthalmologist-photo-grading was 69.2%, and that with optometrist was 60.9%, and the respective Kappa coefficients were 0.31 (p<0.001) and 0.25 (p<0.001). The areas under curve (AUC) on ROC curve were larger for optometrist photo screening (DR=0.85 and DMac=0.80) than clinical examination (DR=0.52 and DMac=0.54). The prevalences of DR and DMac were 19.4% and 3.3% respectively. Duration of DM was the only common significant predictor of referable DR and DMac by chi-square test. 15-year of disease significantly increased the risk of more advanced DR and DMac. Nephropathy was only significant for DR but not DMac. Moderate renal dysfunction as indicated by decreased excretory glomerular filtration rate (eGFR) < 60 mL/min/1.73m2 and presence of macroalbuminuria significantly worsened DR staging, by over 5 times (p=0.02). Family history of diabetes mellitus (DM), on the other hand, increased the risk of DMac only, by 5 times (p=0.01).

Conclusion The current optometrists’ screening is reasonably valid and reliable, although there is room for improvement. None-the-less, the higher false-positive rates than false-negative rates for referable disease suggests that it is safer than otherwise. It is a better method than clinical examination. Duration of DM and presence of nephropathy, especially macroalbuminuria which predates decrease in eGFR, and family history of DM can predict more advanced DR and/or DMac development.