AI-Screening for Diabetic Retinopathy in Primary Care

Non-mydriatic fundus imaging systems and autonomous artificial intelligence (AI) platforms offer an alternative that may reduce the burden on health care systems while maintaining diagnostic accuracy, according to a recent study published in BMJ Open Ophthalmology.

Diabetic retinopathy (DR) is the leading cause of vision loss among working-age adults with diabetes, which affects roughly 1 in 10 Canadians.¹ Additionally, the CDC reports that DR is the leading cause of blindness in American adults of working age, and approximately 4.1 million Americans are affected.² Traditional screening models have family physicians refer all patients with diabetes to ophthalmologists for eye exams. However, this study uses Canadian real-world data to show differences in costs associated with AI platform utilization.¹

EyeArt AI Retinopathy Study

The study employed a comparative cost analysis of real-world AI screening outcomes vs a hypothetical traditional physician-referral scenario. The study population consisted of adults aged 18 or older with type 1 or type 2 diabetes presenting for routine visits at a Thunder Bay, Ontario, clinic, December 2022 to October 23 (n = 202).

The AI system utilized was iCare DRSPlus fundus camera paired with EyeArt by Eyenuk, a Health Canada-approved autonomous platform generating referral recommendations for more-than-mild DR. The cost model included direct costs such as AI exam fees and ophthalmologist consult fees and indirect costs such as patient time, lost income, caregiver time, and commuting. This was based on Ontario Health Insurance Payment (OHIP) rates and Statistics Canada travel-time data.

Diabetic Retinopathy Screening Demographics

A total of 202 participants were screened for DR; the mean age was 70.8 years (SD ± 11.7) and 38.6% female.¹ Most participants (88.1%, n = 179) were Caucasian, and 99% (n = 200) resided in Thunder Bay. The majority (94.5%, n = 189) traveled to appointments by private vehicle. Overall, 29.2% (n = 59) of participants were currently employed, and 35.6% (n = 72) attended appointments with a caregiver; the remaining 64.4% (n = 130) attended alone. Most participants (74.8%; n = 151) reported having type 2 diabetes, 21.8% (n = 44) were uncertain of their diabetes type, and 3.47% (n = 7) reported being diagnosed with type 1 diabetes.

AI Screening Success Rate

Most (93.6%, n = 189) of the AI-based DR screening exams were completed successfully, with only 1.56% (n = 3) participants requiring mydriatic eyedrops for imaging. Of the 202 participants screened, 55 were referred to the ophthalmologist for further review. This included 6.4% (n = 13) of the total 202 participants who were unable to complete their exams using fundus imaging and autonomous AI, and 22.2% (n = 42) had completed exams in which more-than-mild DR was detected.

Michael D. Abramoff, MD, PhD, founder of a separate autonomous AI diagnostic platform for DR (IDx-DR) and professor of ophthalmology and visual sciences at the University of Iowa Carver College of Medicine, described the appeal of point-of-care AI screening in a previous interview with The American Journal of Managed Care^®.³

“Additionally, with the use of AI, we are able to provide an instant diagnosis at the point of care. Clinical staff and patients do not have to wait for results, enabling more comprehensive care plan discussions while the patient is in the office and eliminating the need for callbacks,” he said.

Cost Per Diagnosed DR Case

Notably, the AI screening fee in this model wasn't based on a Canadian billing rate, as none exists.¹ Instead, it was based on US Medicare's CPT code 92229 for autonomous AI retinal analysis, converted to Canadian dollars. The authors note that this conversion introduces uncertainty, which they address with using a wider range of fee estimates.

The cost findings for the base care, per 100 patients, showed the AI-based scenario total direct costs were $C7919.04 ($C5686 for AI screening and $C2233 for ophthalmologist fees) and indirect costs of $C5728.80, resulting in total costs of $C13647.84. In comparison, the total direct costs, per 100 patients, of the traditional physician-based approach were $C8240 (all for ophthalmologist fees) and indirect costs of $C19998.09, resulting in total costs of $C28238.09. The cost per diagnosed case was $C620.36 for the AI approach compared to $C1,283.55 in traditional, showing a 52% reduction.

In some cases, screening and diagnosis of patients with DR also incorporates a retinal optical coherence tomography (OCT) to look for macular edema. Though not needed for all patients, the study assumed the additional scan for up to one-third of patients and added an additional OHIP billing fee of $C35 for the OCT. In this scenario, the total cost per diagnosed DR case was $C634.58 for the AI-based approach and $C1336.05 for the traditional method. The results show a 53% reduction in AI-based cost per case.

Since there is no Canadian reimbursement code yet, the AI fee used in the model was based on a US billing estimate. As a result, the negotiated Canadian rate could come in lower, the authors noted. They modeled a scenario with the IA test fee set 50% lower ($C28.43), on top of the OCT-adjusted assumptions. In this scenario, the total cost per diagnosed DR case was 62% lower ($C505.50) than the traditional physician-based approach, which was the largest gap identified across the study's scenarios.

AI Diagnostic Accuracy Limitations

This analysis is limited by its single-clinic design, with a population skewed older, predominantly Caucasian, and urban, limiting generalizability to remote, Indigenous, or younger working-age populations. The sample size was relatively small (n = 202), and cost estimates relied on several assumptions, including the AI fee's US-to-Canadian conversion and how equipment costs were treated as sunk. Diagnostic accuracy of the AI platform was not independently tested. Additionally, alternative approaches, such as having an ophthalmologist pre-review positive AI results, were not evaluated.

AI-Approached Solution for Ophthalmologist Shortage

This comparative cost analysis is one of the first Canadian real-world cost comparisons of AI-based DR screening and uses a US billing rate. With most of the costs being indirect costs to patients and caregivers, the AI approach provides an alternative with only a small proportion of patients requiring an additional visit for specialist screening.

The authors note the AI screening model's potential to ease documented ophthalmologist shortages and wait lists. They also cite prior findings linking the model to increased specialist productivity and higher patient satisfaction.

References

1. Di Matteo L, Whitestone N, Bhambhwani V. Screening for diabetic retinopathy with artificial intelligence in a primary care setting: a comparative cost analysis. BMJ Open Ophthalmol. 2026;11(3):e002904.
2. Centers for Disease Control and Prevention. About common eye disorders and diseases. Vision and Eye Health. CDC. Reviewed May 15, 2024. Accessed August 7, 2026. https://www.cdc.gov/vision-health/about-eye-disorders/index.html
3. Delmendo I. Artificial intelligence for the diagnosis of diabetic retinopathy: an interview with Michael D. Abramoff, MD, PhD. AJMC^®. Published March 17, 2020. Accessed August 7, 2026. https://www.ajmc.com/view/artificial-intelligence-for-the-diagnosis-of-diabetic-retinopathy-an-interview-with-michael