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

Minority and underserved youth are less likely to undergo diabetic retinopathy screening, yet more likely to have diabetic retinopathy. We hypothesize that autonomous artificial intelligence-based diabetic eye exams could mitigate screening disparities in minority youth. This prospective study enrolled 397 White and non-White youth with type 1 (72%) and type 2 (28%) diabetes at diabetes care visits. Participants underwent autonomous artificial-intelligence diabetic eye exams and optical coherence tomography imaging. The primary outcome is equivalence in screening completion between White and non-White youth. The secondary outcome is diagnostic accuracy of AI against prognostic standard. At baseline, participants in the White cohort were more likely to have completed a prior diabetic eye exam (79.8% v 61.8%, p < 0.0001). Following the artificial-intelligence intervention, 99.5% in the White cohort and 98% in the non-White cohort completed the autonomous artificial-intelligence diabetic eye exam, demonstrating equivalent screening completion after implementation. When compared to standard grading, AI demonstrated sensitivity of 100% (95% CI: 67.6%, 100%), and specificity of 76% (95% CI: 71.4%, 79.9%) for detecting more than mild diabetic retinopathy. Implementation of autonomous AI at the point of care increases access to and completion of diabetic eye exams, achieving equivalent screening completion among White and non-White youth, suggesting the potential to reduce disparities in diabetic retinopathy screening. Study Registration: NCT05463289. This study demonstrates that autonomous AI screening for diabetic retinopathy can achieve equivalent screening completion among white and non-white youth with T1D, suggesting the potential to reduce disparities in diabetic retinopathy screening.