Advanced Journal of Environmental Science and Technology

ISSN 2756-3251

Advanced Journal of Environmental Science and Technology ISSN 2756-3251, Vol. 17 (11), pp. 001-004, November, 2026. 

Research Article

Artificial Intelligence for Early Detection of Diabetic Retinopathy in Primary Care: A Pragmatic Implementation Study

Amara Okafor¹, Daniel Reyes², Hannah Kim¹, Tunde Bakare³

¹Department of Primary Care and Population Health, University of Lagos Teaching Hospital, Lagos, Nigeria
²Institute of Biomedical Engineering, University of Toronto, Toronto, Canada
³National Eye Centre, Kaduna, Nigeria

Corresponding Author: Amara Okafor — [email protected]

Abstract

Background: Diabetic retinopathy remains a leading cause of preventable vision loss, yet screening coverage in primary care is low because specialist grading is slow, scarce, and unevenly distributed. This pragmatic implementation study evaluated whether an artificial intelligence (AI) triage system, embedded in routine primary-care visits, could safely increase early detection without overwhelming referral pathways.

Methods: Twelve primary-care clinics enrolled 4,812 adults with diabetes over 18 months. Trained nurses captured two 45° retinal photographs per eye. The AI system assessed image quality, graded referable diabetic retinopathy, and returned an immediate triage recommendation: routine rescreening, routine referral, or urgent referral. All AI-positive and a random 10% sample of AI-negative cases were independently graded by two masked ophthalmologists. Primary outcomes were sensitivity and specificity for referable disease; secondary outcomes were image gradability, time to referral, and clinic workflow impact.

Results: Of 4,812 participants, 4,658 (96.8%) had gradable images. Referable diabetic retinopathy was confirmed in 612 participants (13.1%). The AI system achieved 93.2% sensitivity and 91.5% specificity for referable disease, with 89.4% referral accuracy. Median time from screening to specialist appointment fell from 42 days in the pre-implementation period to 11 days for urgent referrals. Clinics reported no increase in average consultation length after the first four weeks.

Conclusions: AI-assisted triage in primary care can detect referable diabetic retinopathy early, safely, and at scale, provided that image-quality feedback, clear referral thresholds, and human oversight are built into the workflow. The approach is promising for settings where ophthalmology capacity is limited, but prospective monitoring for performance drift, equity, and cost remains essential.

Keywords: diabetic retinopathy; artificial intelligence; primary care; screening; implementation science; eye health

Manuscript Timeline: Received 12 August, 2026; Revised 28 September, 2026; Accepted 30 October, 2026; Published 21 November, 2026.

Citation: Okafor, A., Reyes, D., Kim, H., and Bakare, T. (2026). Artificial Intelligence for Early Detection of Diabetic Retinopathy in Primary Care: A Pragmatic Implementation Study. Advanced Journal of Environmental Science and Technology, 17(11), pp. 001-004.