Remote Diabetic Retinopathy Screening: Ready for Prime Time?

Experts debate widespread implementation of telemedicine-based strategy. Cheryl Guttman Krader reports.

Default banner image for Remote Diabetic Retinopathy Screening: Ready for Prime Time?
Photo of Cheryl Guttman Krader

Experts debate widespread implementation of telemedicine-based strategy. Cheryl Guttman Krader reports. Leading retina specialists agree a telemedicine-based approach holds tremendous promise for increasing diabetic retinopathy (DR) screening. However, their opinions diverge about its readiness for large-scale expansion. In a debate held at the AAO in New Orleans, Jennifer I Lim MD, FARVO, FASRS took up the argument in favour of telemedicine screening for DR being ready for expansion now. She argued it addresses the growing burden of screening for DR and is efficient, effective, and accurate. The discrepancy between the growing population of patients in need of DR screening and the number of eyecare providers available to provide the service combined with the low rate at which patients with diabetes receive standard-of-care ophthalmic screening establishes the rationale for a telemedicine approach, she asserted. “Telemedicine will allow us to address the shortage of eyecare providers, and specifically to address the mismatch between the numbers of providers serving patients in rural areas or who are socioeconomically disadvantaged, which leads to disparities in care,” Dr Lim said. She cited evidence demonstrating the effectiveness of telemedicine DR screening programmes for increasing the screening rate. For example, data from the Appalachian Diabetic Eye Network, a primary care-embedded programme established to extend DR screening access for underserved populations in Kentucky, showed the screening rate increased from 29.9% in the year before the programme started to 47.7% after only one year. Other analyses showed more than 75% of the acquired screening images were gradable, and a significant proportion of those (22%) showed DR, including a need for specialist referral in 15.2%. “Other studies also show telemedicine screening for DR conducted at primary care offices can be very effective. That is important because only 30% to 50% of diabetics see an eyecare provider. However, most diabetics visit their primary care provider at least once per year,” Dr Lim said. More recently artificial intelligence (AI) screening has become available that allows ophthalmologists to treat patients and not screen. Two autonomous AI systems are approved by the US FDA, and both have proven to have very high sensitivity, high specificity, and high imageability for detecting both more than mild DR and vision-threatening retinopathy, Dr Lim stated. “Recognising the accuracy of these systems, the Centers for Medicare and Medicaid Services in the US granted final approval for a CPT code (billing code) for AI detection of DR. This is further evidence that telemedicine screening for DR is ready to go. In fact, it has already gone,” Dr Lim said. OBSTACLES ON THE PATH TO WIDESPREAD ADOPTION Taking the opposing side, Christina Y Weng MD, MBA agreed that telemedicine DR screening has tremendous potential. But then she outlined several obstacles necessary to overcome before it can become the primary method for screening. She began with a case example illustrating the limited accuracy of a telemedicine screening programme for detecting diabetic macular oedema (DME). “DME affects over half a million people in the US, and telemedicine- based detection of DME is simply not as robust as detection of DR. While we could integrate OCT into teleretina screening, it would drive up cost, which is already a challenge in this space.” Dr Weng continued, “Care must also be taken when interpreting the reported sensitivity rates for telemedicine DME screening as they are often compared to reading centre fundus photographs as the reference rather than OCT, which is the real gold standard.” Economic considerations are also limiting the expansion of telemedicine screening for DR. Dr Weng cited cost analyses from her group and others that estimated the cost of non-AI-based teleretina screening to be between $25 and $50 per patient. “Data for AI-based screening are more limited because of [how new] this approach [is to the industry]. However, a camera for one of the available autonomous AI systems costs about $18,000 and $34 to $50 per reading.” Furthermore, reimbursement rates in the United States for teleophthalmology do not cover or barely cover the associated costs, she said. As another concern for clinicians, the medicolegal landscape of teleretina screening is ambiguous. “There are not a lot of precedents, and medicolegal protection will become trickier with AI incorporation,” Dr Weng said. Lastly, telemedicine DR screening success relies on in-clinic follow-up compliance from individuals who have disease detected outside ophthalmology. Available data indicate this is an important weak link. “Looking at our data for a four-year period, we were shocked to find that the attendance rate of screened patients who were referred for an in-clinic exam was only 52%. We need to improve the transition between telemedicine and in-person care before the telemedicine approach becomes our primary way of DR screening.” The debate took place at the AAO 2021 Retina Subspecialty Day in New Orleans, Louisiana, USA. Jennifer I Lim MD, FARVO, FASRS, is Professor and the Marion Schenk Chair, Vice Chair, and Director of the Retina Service, Department of Ophthalmology, University of Illinois, Chicago, USA. jennylim@uic.edu Christina Y Weng MD, MBA, is an Associate Professor of Ophthalmology, Baylor College of Medicine, Houston, Texas, USA. christina.weng@bcm.edu