Objective
The objective of this study is to determine if the autonomous AI system (IDx-DR, Digital Diagnostics) is accurate in detecting and diagnosing diabetic retinopathy from retinal images of youth with diabetes. Using this information, we hope the FDA will expand their clearance for use of this device in pediatric diabetes to screen for diabetic retinopathy, which will allow more widespread use of this screening system.
Background Rationale
Diabetic retinopathy (DR) is a complication of diabetes that affects up to 4-15% of children with diabetes. Early detection of DR before symptoms occur prevents vision loss, yet only 35-72% of youth undergo recommended regular DR screening, and youth from minority and low-income households are less likely to complete the recommended screening. Barriers to DR screening in this age group include time off from work and missed school and transportation access. Use of nonmydriatic fundus cameras, cameras that take pictures of the retina, at regular diabetes care visits has been shown to improve screening rates. These retinal images can be interpreted remotely or in a deferred manner, with results provided to the patient within 1-2 weeks. In 2018, based on a preregistered clinical trial, FDA De Novo cleared the first autonomous artificial intelligence (AI) system for diagnosing diabetic retinopathy and macular edema (IDx-DR, Digital Diagnostics, Coralville, Iowa) in adults (aged 21 and up). This AI system interprets retinal images on its own and provides an immediate result to the patient. In a pilot study using this same autonomous AI system, we demonstrated safety, diagnostic accuracy and equity in the pediatric population. Implementation of autonomous AI systems for diabetic retinopathy screening at the point of care have the potential to improve access to screening, facilitate early detection of diabetic retinopathy and prevent vision loss in youth with diabetes, while saving patients and their caregivers cost and time associated with seeing another subspecialist.
Description of Project
Diabetic retinopathy is a complication of diabetes that leads to blindness. Screening for diabetic retinopathy is recommended for children and adolescents with type 1 diabetes, and generally involves a referral to an eye care provider for a dilated eye exam requiring time off from work and school, and eye dilation that lasts for several hours. In general screening rates are low, especially for underserved minority and low-income youth. In 2018, the FDA approved the first autonomous artificial intelligence system for diagnosing diabetic retinopathy using a non-mydriatic fundus camera at the point of care, meaning that retinal pictures can be taken without eye dilation and the system will determine if there is diabetic retinopathy or not. This point of care artificial intelligence system is safe and effective in adults, but has not been FDA cleared for use in children. In this study, we aim to expand upon already funded clinical trials to obtain FDA approval for this system’s use in youth with diabetes. Standard use of this AI system can improve screening rates in children with diabetes and save patients and their caregivers the cost and time that it would require to see a separate eye doctor.
Anticipated Outcome
Through this study, we hope to obtain FDA clearance for use of an autonomous artificial intelligence system for point of care diabetic retinopathy screening and results that will improve access to screening, allow early detection of diabetic retinopathy that can ultimately prevent vision loss, and save patients and their caregivers cost and time associated with seeing another specialty provider. We hope that improved screening access and early detection of diabetic retinopathy can help mitigate disparities in outcomes for racial minority underserved youth with type 1 diabetes.
Relevance to T1D
Type 1 diabetes is a chronic disease requiring complex daily management. In addition to the daily burden of care, patients with type 1 diabetes are recommended to see their diabetes care provider every 3 months. These visits generally take up to 1-2 hours as they often involve multidisciplinary care with nurse/CDE, dietitian, diabetes provider and psychology. In addition, patients are recommended to undergo diabetic eye exams at least every 2 years, which generally requires another specialty care visit. Through this study, we hope to obtain FDA clearance for use of an autonomous artificial intelligence system for point of care diabetic retinopathy screening and results that will improve access to screening, allow early detection of diabetic retinopathy that can ultimately prevent vision loss, and save patients and their caregivers cost and time associated with seeing another specialty provider. Furthermore, there are significant disparities in glycemic outcomes and rates of diabetes complications for racial minority and low-income youth, and we anticipate that improved screening access and early detection of diabetic retinopathy can help mitigate disparities in outcomes for underserved youth.