Artificial intelligence (AI) may enhance detection of diabetic retinopathy and optimize care for diabetic patients; however, its adoption in screening is currently limited.

To understand barriers to adoption, a recent study analyzed data from the TriNetX database—which covers more than 107 million patients in the US—and evaluated the use of Current Procedural Terminology code 92229 for artificial intelligence-based diabetic retinopathy detection from 2019 to 2023. This code was compared with use of codes for remote eye imaging (92227 and 92228), fundus photography (92250), and optical coherence tomography (92134).

The researchers identified several challenges that impeded broader adoption of AI-based imaging, including high costs and integration hurdles; limited awareness and support for implementing AI systems in primary care and ophthalmology settings; and references for established imaging methods, such as fundus photography and optical coherence tomography (OCT).

Among nearly 5 million patients with diabetes, only 0.09% were screened using AI-based imaging for diabetic retinopathy (DR) detection, representing 2.2% of all cases requiring targeted imaging between 2021 and 2023. Traditional remote imaging methods saw a 185.4% increase in usage during the same period, compared to only a 1.0% increase for AI imaging. OCT imaging was the most frequently used modality; it was applied in 80.3% of imaging cases, followed by fundus photography which was used in 35.0% of imaging cases.

“More than 80% of those who received AI imaging were from the South, a region that made up only 40% of other imaging modalities, and almost half of the patients who received AI imaging were Black, compared with approximately a quarter in other imaging modalities,” the researchers noted in their JAMA Ophthalmology article. Further, AI imaging had a higher referral rate to OCT (7.74%) compared to traditional remote imaging (5.53%).

The investigators mentioned FDA-approved systems LumineticsCore and EyeArt that demonstrate high sensitivity (87.2% and 96.0%, respectively) and specificity (90.7% and 88.0%, respectively) in DR screening. “Use of these systems helps increase detection in the primary care setting among patients with diabetes, while optimizing ophthalmic examinations for those with vision-threatening DR,” they wrote.

To enhance the adoption of AI technologies in DR screening, the researchers suggested streamlined workflows to integrate AI systems seamlessly into clinical practice; collaborative efforts between primary care providers and ophthalmologists to ensure early detection and management of vision-threatening DR; and increased patient education and access to advanced imaging modalities.

“This cohort study, while not population based, found that only 4.2% of diabetic patients received ophthalmic imaging for DR over 5 years,” they concluded. “These findings support further evaluation of imaging practices to develop targeted strategies for improving diabetic eye imaging rates and patient outcomes.”

A full list of author disclosures is available in the published research.