An AI eye exam increased screening among young people with diabetes

IDx-DR, now sold as LumineticsCore. [1]

The locked diagnostic algorithms within IDx-DR. The trial does not name a separate model family or release. [1]

The paper and supplement do not identify the installed software release. [1]

Digital Diagnostics, United States. [1]

The software checked image quality and assessed photographs for diabetic eye disease. [1]

The device used two retinal views and allowed up to three attempts to obtain suitable images. The trial does not name the camera model. [1]

Two color retinal photographs centered on the fovea and optic disc were assessed. The output indicated disease, no disease, or insufficient image quality. [1]

The software produced its result without a clinician interpreting the photographs first. The research team nevertheless obtained specialist overreads for safety. [1]

Clinic staff photographed the retina during the diabetes visit. Participants received the result immediately and were referred for an eye-care visit when indicated. [1]

A board-certified retina specialist reviewed every AI image for safety. An independent monitoring board oversaw the study. [1]

IDx-DR assessed retinal photographs and returned a screening result during the diabetes visit. The trial compared this on-site service with referral elsewhere. A retina specialist also reviewed the images for safety. A coauthor had financial interests in the manufacturer, although he did not have access to the underlying data.

IDx-DR made an eye exam available at the diabetes visit. The trial tests this entire on-site service against referral elsewhere. [1]

AIBD-0067-outcome-01 [1]

More participants completed screening when an AI eye exam was offered during their diabetes visit. [1]

Improved Physical Health

Observed in a randomized trial. [1]

People.

Young people with type 1 or type 2 diabetes who were due for a diabetic eye exam. [1]

All 81 AI participants and 18 of 82 controls completed screening. The 25 positive AI screens were not all confirmed diagnoses. [1]

The trial included a racially diverse population, and 47% of participants had Medicaid coverage. It was not large enough to establish equal effects across every subgroup. [1]

Screening and referral attendance were followed for six months. [1]

Yes

The AI produced positive screening results that required further assessment. Its reported specificity was 78.9% against specialist review. A positive screen was not a confirmed diagnosis. [1]

NCT05131451 [1]

An unmasked randomized trial, with allocation concealed until assignment. Randomization used blocks of four or six and was stratified by clinic. [1]

The primary outcome was documented completion of a diabetic eye exam within six months. Researchers also tracked referral attendance and participant satisfaction. [1]

The trial randomized 164 participants at two Baltimore clinics. After one control was excluded, the analysis included 81 AI participants and 82 controls. [1]

Completion of a diabetic eye exam within six months, with referral attendance as a secondary outcome. [1]

Scripted education and referral to an eye-care provider, with a printed handout. [1]

Screening completion was 100% with AI and 22% with referral, a difference of 78 percentage points (95% CI, 69–87). [1]

Randomization supports the effect of offering an immediate AI exam rather than a referral. It does not separate the algorithm from convenience and access at the clinic. [1]

The trial measured access to screening, not later vision. The referral comparison used 25 AI-positive participants versus the entire control group, so those secondary denominators are not equivalent. [1]

Participants enrolled from 24 November 2021 to 6 June 2022. Follow-up ended on 6 December 2022. An AI exam generally took five to ten minutes. [1]

The protocol and supplement describe the procedure. De-identified data require an approved proposal and data-use agreement. The commercial diagnostic code is not published. [1]

A gift card for study participation. This was a transfer, not value that AI created. [1]

US$25 per participant, plus a parking pass whose value was not reported. [1]

U.S. dollars during the 2021–2022 study. No inflation adjustment was reported. [1]

Per participant for study participation. [1]

The authors report that participants received the gift card and parking pass. [1]

Study participants received compensation. The article does not identify the precise paying account. [1]

The US$25 is the face value of a gift card. It is not net benefit from screening. [1]

The article does not report the full cost of the service, specialist review, staff time, or parking passes. [1]

No comparison of financial outcomes between the randomized groups was reported. [1]

The source states the gift card’s face value. It does not value health outcomes. [1]

No financial return or payback period was estimated. [1]

Johns Hopkins conducted the trial at Johns Hopkins Hospital and Mount Washington Pediatric Hospital. Digital Diagnostics supplied the diagnostic product. [1]

Both clinics were in Baltimore, Maryland, United States. The manufacturer was also based in the United States. [1]

The reviewed clinical sources identify the organizations involved, but do not establish their complete ownership or control arrangements. [1]

The service required a retinal camera, trained clinic staff, IDx-DR, specialist safety review, and a referral route. [1]

The reviewed sources do not document international transfers of patient data or quantify cross-border financial flows. [1]

Not assessed

The tested service combined an immediate eye exam, clinic staff, follow-up referrals, and specialist safety review for young participants. [1]

The reviewed sources do not report an independent repetition of this clinical result. [1]

The study followed eye-exam completion and referral attendance for six months. It did not report later vision outcomes.

LumineticsCore is commercially available. The trial’s pediatric use was investigational because the FDA indication covered adults aged 22 or older. [1]

Johns Hopkins evaluated AI screening in young people below the device’s FDA-indicated age. [1]

The study received IRB approval, an investigational device exemption, and independent safety monitoring. [1]

Staff used the device in the trial and arranged retina-specialist review of every AI image. [1]

The prespecified primary outcome was completion of an eye exam within six months. [1]

1 retained sources.

Original citations and the passages that support each finding. [1]

English. No translation was used. [1]

Academic investigators conducted the trial. One coauthor was an investor, director, and consultant of Digital Diagnostics, but had no access to the underlying data. [1]

The original studies and their supporting materials are linked to the findings that they support. [1]

Access and version details are listed for each reviewed source. [1]

An on-site IDx-DR exam increased screening completion from 22% to 100%. [1]

The trial flow diagram supplies the analyzed counts. Its supplement records specialist safety review. [1]

A peer-reviewed randomized trial by academic investigators, with a coauthor who held financial interests in the manufacturer. [1]

The paper does not establish later vision outcomes, the device’s exact software release, or a separate causal effect of the algorithm. [1]

The benefit described here, for the beneficiary groups shown.

Depth 2 for improved access to completed screening.

Depth 2 for improved access to completed screening.

The depth describes the reported result. Evidence strength is assessed separately.

AIBD’s five-level depth rubric, reviewed on 25 October 2026.

AIBD, 25 October 2026.

A null grade would omit the measured improvement. Grades 3–5 would require a lasting capability or health benefit that this trial did not establish.

A randomized academic evaluation with a manufacturer-affiliated coauthor. The outline is retained under the current independence rule.

A randomized trial of on-site screening versus referral. [1]

Not assessed

Healthcare - diabetic retinopathy screening. Autonomous AI screening in pediatric diabetes care.

IDx-DR/LumineticsCore is built to autonomously detect more-than-mild diabetic retinopathy from fundus images at the point of care. Increasing screening completion is its intended benefit.

Enrollment began on 24 November 2021, and follow-up ended on 6 December 2022. [1]

Investigational use in a pediatric diabetes trial. [1]

SHA-256: f6eae30ef3c1…

NCT05131451 [1]

Not assessed

Not assessed

7 revisions. Latest: 25 October 2026.