AI Screening for Diabetic Retinopathy Is Cost-Effective for At-Risk Children


At recommended adherence rates, automatic AI screening technology for diabetic retinopathy is a preferred strategy over conventional eye care professional screening.

Risa Wolf, MD

Risa Wolf, MD

A new economic evaluation of diabetic retinopathy screening options suggests autonomous artificial intelligence (AI) screening in pediatric patients with diabetes is effective and cost-saving for families when ≥23% of patients adhered to screening recommendations.

Otherwise, conventional ophthalmologic screening by an eye care professional was considered the preferred strategy in a lower-adherence scenario.

This year, the American Diabetes Association (ADA) has added the US Food and Drug Administration (FDA)-approved AI technology to its standard of care for diabetic retinopathy screening. Thus, patient cost sharing becomes equally important to consider, since a lower financial burden is linked to increased screening adherence and improved outcomes. 

Risa Wolf, MD, from the Department of Pediatrics at Johns Hopkins School of Medicine, and colleagues compared the cost-effectiveness between AI and standard eye care professional screening strategies for pediatric patients using parameter estimates from relevant data published between 1994-2019.

The parameters consisted of out-of-pocket costs for AI screening, ophthalmology visits, and treatment of diabetic retinopathy. They also included the probability of undergoing standard retinal examination, relative odds of undergoing screening, as well as the sensitivity, specificity, and the ability to diagnose both screening strategies.

Main outcomes sought by the investigators were cost-effectiveness based on the proportion of identified true-positive results, as well as patient cost or savings based on the mean patient payment for diabetic retinopathy screening examination.

To better visualize the diagnostic scenarios, Wolf and team constructed a decision tree model to illustrate the assumed branching options for standard ophthalmologic care or AI screening. Each alternative allowed the possibility for pediatric patients to not get screened (non-adherence). If screening did take place for patients who had chosen the AI strategy, and if the patient tested positive, they were then referred to an eye care specialist for further examination.

The investigators noted that the model assumes that these referred patients incur an additional cost from the ophthalmologic examination.

Using their decision tree and the literature-derived parameter estimates, they determined that the expected true-positive proportions for standard screening were 0.006 for type 1 diabetes (T1D) and 0.01 for type 2 diabetes (T2D). According to the investigators, the low true-positive proportion was attributed to the low probability (0.20) of patients keeping their eye care professional appointment.

The expected true-positive proportions for autonomous AI were 0.03 for T1D and 0.04 for T2D.

In terms of cost, they found that for the base-case scenario—where adherence was at 20%—the mean patient payment for the conventional screening was $7.91 for T1D and $8.20 for T2D—versus $8.52 and $10.85, respectively, for autonomous AI screening.

However, when adherence with the eye care professional examination increased above the 20% baseline, the mean patient payment increased as well. At a 23% adherence threshold, the eye care professional examination surpassed the AI option in mean payment.

“The results suggest that DR screening using autonomous AI systems is effective and cost saving for children with diabetes and that use of POC screening with immediate results will increase DR screening rates, thus improving care and providing early identification of diseases,” the investigators wrote.

They concluded that future models for cost-effectiveness and cost savings should analyze the use of such AI systems from the health care system perspective.

The study, “Cost-effectiveness of Autonomous Point-of-Care Diabetic Retinopathy Screening for Pediatric Patients With Diabetes,” was published online in JAMA Ophthalmology.

Related Videos
Roger Goldberg, MD: Impact of Dual Inhibition on Hard Exudates in DME | Image Credit: Bay Area Retina Associates
Deepayan Kar, PhD, MS: A Virtual Reality Approach to Contrast Sensitivity in AMD | Image Credit: LinkedIn
Dolly Chang, MD, PhD | Image Credit: LinkedIn
Daniel Su, MD: | Image Credit: Retina Vitreous-Associates Medical Group
Daniela Ferrara, MD, PhD: Machine Learning Predicts Functional Outcomes in DME | Image Credit: Ophthalmology Innovation Summit
Ashkan Abbey, MD: 3-Year Efficacy of Pegcetacoplan for Geographic Atrophy | Image Credit: Texas Retina AssociatesImage Credit:
A panel of 5 experts on Cushing's syndrome
A panel of 5 experts on Cushing's syndrome
© 2024 MJH Life Sciences

All rights reserved.