Point-of-care retinal imaging is effective at identifying diabetic retinopathy. However, appropriate follow-up, referral, and access to specialized eye care remains a barrier for the treatment of identified eye disease. We tested the hypothesis that rapid results and immediate referrals afforded by use of an automated diabetic retinopathy screening platform with AI technology (EyeArt) would result in greater patient follow up with ophthalmology specialists as compared to “traditional” tele-retinal screening (images acquired in clinic and transmitted electronically for interpretation by a vitreoretinal specialist) . All images were acquired during a patient’s regularly scheduled diabetes clinic visit. With traditional tele-retinal screening, the average turnaround time to patient notification of findings was days. Among the 357 patients that underwent traditional screening, 42 (12%) were found to have referable (more-than-mild) retinopathy, and of these, 32 (76%) followed up in ophthalmology clinic. In contrast, the fully autonomous diabetic retinopathy screening platform using AI technology provided an immediate assessment on the presence of retinopathy (within 60 seconds of photo acquisition) . Patients were given this point-of-care summary of their eye-exam findings. When indicated, an ophthalmology referral was ordered and patients were provided with contact information for the ophthalmology office. Of the 81 patients screened with AI technology, 22 (27%) had referable retinopathy detected, 15 (68%) of whom followed up in ophthalmology clinic. The higher detection rate with AI screening raises the question of “overcalling” the presence of disease. We conclude that although AI screening offers immediate results allowing for immediate ophthalmology referral, there remains a substantial barrier to ophthalmology follow up after a positive retinopathy screen identified by either screening method.

Disclosure

M.Wahl: None. R.Kiani: None. J.Neuberger: None. B.Wilson: None. M.Hartnett: None. S.Fisher: None.

Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at http://www.diabetesjournals.org/content/license.