A team of ophthalmology researchers at Stanford University, working with a colleague from New York University School of Medicine, has found that despite promise for detecting diabetic retinopathy, few patients are given AI-assisted screening tests. In their study, published in the journal JAMA Ophthalmology, the group analyzed patient data on millions of diabetics and the treatments they received.

In diabetic retinopathy, high blood sugar levels damage blood vessels in the retina, which can lead to retinal cell death due to an inadequate supply of oxygen. Treatments for the condition include surgery, laser therapy and anti-VEGF drugs and are considered to be effective in preventing vision loss, though the earlier it is caught the better the outcome.

Prior research has shown that despite the high risk of vision loss, less than two-thirds of diabetes patients have an annual eye exam. Currently, testing during such exams includes conducting a dilated fundus exam, which involves taking pictures of the retina which are then inspected for diabetic retinopathy.

Several years ago, researchers found that AI applications trained to spot the signs of diabetic retinopathy are just as good, and in some cases, better than humans at assessing eye scans. They showed so much promise that the FDA approved them for general use in 2021. In this new effort, the research team wondered if ophthalmologists have begun using the new technology.

To find out, the researchers accessed the TriNetX federated database, which includes data from 62 health care organizations in the U.S. Included in the data was whether a patient had type I or II diabetes and when and where they had had eye exams—and if so, which kind. The team looked at data from 2019 to 2023 for 4,959,890 diabetic patients.

The research team found that AI-based tests were used on only 0.09% of eye exams, even as use of traditional methods increased by 185.4% during the same period. They also noted that the vast majority of patients who received the AI tests lived in the South and approximately half were Black. They conclude that the adoption of AI-based diabetic retinopathy is making little headway in the medical community.

© 2024 Science X Network