DP24-081 The Glaucoma and Retinopathy Screening Study

Massachusetts Eye And Ear Infirmary, Boston MA

Investigators

Abstract

Project Summary Glaucoma is the second leading cause of blindness in the United States, with over half of those with glaucoma unaware that they have it. Underserved minority populations are at even higher risk of being undiagnosed, leading to severe health inequities. Glaucoma is asymptomatic mainly in the early stages when vision loss and disability can be prevented and, therefore, is only recognized by someone with the disease after permanent severe vision loss is noticed. Glaucoma is an excellent candidate for early detection since effective and timely treatment can arrest or delay loss of visual function. The relatively low prevalence of glaucoma, the high cost, and the low specificity of existing diagnostic approaches, such as clinician evaluation, have limited widespread early detection, which has been successful, for example, for diabetic eye disease. Recent improvements in the accuracy of autonomous AI diagnosis of glaucoma, in combination with improved ease of use and accuracy of virtual perimetry, offer novel technological breakthroughs that can be applied to screening for glaucoma. Building on an increasingly widespread infrastructure of autonomous AI for diabetic eye disease (that uses the same patient images), this novel glaucoma screening strategy could significantly change this equation for the better, especially for currently underserved populations that are at the highest risk for glaucoma. Implementing such a program would likely be cost-effective and improve health equity. We hypothesize that people in high-risk communities with diabetes undergoing point-of-care, real-time autonomous AI for diabetic eye disease (DED) are an ideal at-risk population to also screen for glaucoma, using a combination of fundus photograph-based autonomous AI and virtual perimetry for suspect patients. This efficient system using AI diagnostics as the initial screening followed by confirmatory screening with perimetry evaluation will reduce over-referral, allowing for a scalable model with the potential to reach all individuals undergoing screening for DED as well as others without diabetes who could be screened for glaucoma using our approach. We will add screening for glaucoma to DED screening in four United Against Racism primary care clinics in the Boston area with strong representation of at-risk populations. We will screen 2000 individuals over this study and, assuming a 10% prevalence of glaucoma will identify 200 individuals with glaucoma. We will determine optimal strategies for adding this screening to eye screening programs nationally and will determine the cost- effectiveness of the screening by comparing outcomes of those screened for glaucoma to outcomes of those undergoing DED screening without being screened for glaucoma. The model will be scalable.