Autonomous AI eye exam closes diabetic retinopathy screening gap in minority youth

Autonomous artificial intelligence-based diabetic eye exams nearly eliminated a persistent screening gap when delivered directly at pediatric diabetes care visits, according to a prospective study of 397 young people with diabetes published in Nature Communications. Before the intervention, White youth were substantially more likely than non-White youth to have ever completed a diabetic eye exam, with 79.8% of the White cohort reporting prior screening compared with 61.8% of the non-White cohort, a statistically significant difference. After the point-of-care AI exam was offered, completion rose to 99.5% among White participants and 98% among non-White participants, achieving the study's primary endpoint of equivalence in screening completion between the two groups.

The findings address a twofold disadvantage documented in the study's framing: minority and underserved youth with diabetes are less likely to undergo diabetic retinopathy screening, yet more likely to have the condition. The ACCESS2 study hypothesized that autonomous AI-based diabetic eye exams could mitigate these screening disparities, and tested that hypothesis by embedding the exam directly into routine diabetes care visits rather than relying on a separate referral pathway. The study was registered as NCT05463289.

Participants were young people with type 1 diabetes, who made up 72% of the cohort, and type 2 diabetes, who accounted for 28%. All enrolled participants underwent the autonomous AI diabetic eye exam along with optical coherence tomography imaging during their diabetes care visits. Diagnostic performance was assessed against a prognostic standard using standard grading as the comparison.

For detecting more than mild diabetic retinopathy, the AI achieved a sensitivity of 100%, with a 95% confidence interval ranging from 67.6% to 100%, meaning no cases in this cohort were missed. Specificity was 76%, with a 95% confidence interval of 71.4% to 79.9%, indicating that roughly a quarter of eyes without the condition were flagged as positive. The equivalence in screening completion was demonstrated against a prespecified criterion rather than assumed.

The primary outcome compared completion rates between White and non-White youth after the AI intervention, and the near-identical rates of 99.5% and 98% satisfied it. This contrasts sharply with the baseline disparity in prior screening history that the researchers documented before any exposure to the AI exam. Even so, the authors describe the findings as suggesting the potential to reduce disparities in diabetic retinopathy screening, rather than proving a sustained reduction in retinopathy outcomes over time, and the results come from the specific study population and setting reported.

Subject of Research: Autonomous AI diabetic retinopathy screening to reduce disparities in youth with diabetes

Article Title: ACCESS2 study: the potential for autonomous AI to mitigate disparities in diabetic retinopathy screening in youth

Article References: Wolf, R. M., Channa, R., Liu, T. Y. A., Brown, E. A., Bromberger, L., Parimi, N., Zehra, A., Patel, D., Pan, B., Kim, Y. C., Shehadeh, A., Lehmann, H. P., & Abramoff, M. D. (2026). ACCESS2 study: the potential for autonomous AI to mitigate disparities in diabetic retinopathy screening in youth. Nature Communications. https://doi.org/10.1038/s41467-026-78842-w

Image Credits: AI Generated

DOI: 10.1038/s41467-026-78842-w

Keywords: autonomous AI, diabetic retinopathy, screening disparities, youth, type 1 diabetes, type 2 diabetes, point-of-care diagnostics, optical coherence tomography, health equity, ACCESS2 study, Nature Communications, prospective study