{"slug": "medicare-pays-for-the-ai-eye-scan-nobody-pays-for-what-comes-next", "title": "Medicare Pays For the AI Eye Scan — Nobody Pays For What Comes Next", "summary": "Medicare's 2026 national rate for an AI diabetic eye exam is $46.76 in a doctor's office and $60.27 in a hospital outpatient clinic, but the payment covers only the retinal photo and nothing for the staff who take it, the follow-up calls, or getting a positive patient to an eye specialist, according to an analysis of CMS fee schedules and published studies. Use of the technology remains minimal: a JAMA Ophthalmology study of 62 health systems found fewer than 1 in 1,000 patients with diabetes had an AI eye exam from 2021 to 2023, and Medicare data show about 5,000 people in traditional Medicare had an AI eye exam billed by a doctor's office in 2024, up from about 400 in 2021 — still fewer than 1 in 1,500 Medicare patients with diabetes. The FDA authorized the first autonomous AI retinal-reading system in 2018 and four such systems are now authorized, yet close to half of patients who screen positive in U.S. primary-care diabetic eye-screening programs never reach an eye doctor within the year, a gap that matters for the estimated 9.6 million Americans with diabetic retinopathy, the leading cause of blindness in working-age adults.", "body_md": "In U.S. diabetic eye-screening programs run through primary care, close to half of patients who screen positive [never make it](https://pubmed.ncbi.nlm.nih.gov/39710903/) to an eye doctor within the year. The scan was the easy part.\n\nThat gap matters more now that the scan can be automated. In 2018, the Food and Drug Administration [authorized](https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/denovo.cfm?ID=DEN180001) the first AI system that can read a retinal photo and decide, without a doctor, whether a person with diabetes needs to see an eye specialist. Four such systems are now authorized. Medicare began paying for the AI exam in 2021 and [set a national rate](https://www.govinfo.gov/content/pkg/FR-2021-11-19/html/2021-23972.htm) in 2022. The main quality score that insurers are graded on [counts an AI exam](https://www.cms.gov/files/document/2026-quality-rating-system-measure-technical-specifications.pdf). And the American Diabetes Association’s [2026 Standards of Care](https://doi.org/10.2337/dc26-S012) call AI screening appropriate, as long as patients who need follow-up can get it.\n\nUse is still tiny. In a [JAMA Ophthalmology study](https://pubmed.ncbi.nlm.nih.gov/39480408/) of 62 health systems, fewer than 1 in 1,000 patients with diabetes had an AI eye exam from 2021 to 2023, and use barely moved. Medicare’s [own data](https://data.cms.gov/provider-summary-by-type-of-service/medicare-physician-other-practitioners/medicare-physician-other-practitioners-by-geography-and-service) show growth from almost nothing: about 5,000 people in traditional Medicare had an AI eye exam billed by a doctor’s office in 2024, up from about 400 in 2021. That is still fewer than 1 in 1,500 Medicare patients with diabetes.\n\nThose counts miss hospital clinics, community health centers, private Medicare Advantage plans and programs that screen without billing. They also show where the technology ended up: [optometrists billed](https://pubmed.ncbi.nlm.nih.gov/42554455/) most of the AI exams Medicare paid for in 2023, not the primary-care offices the tools were built for.\n\nLow AI use isn’t the same as low screening; many patients still get a regular eye exam. The real problem is the people who are never screened, and those who screen positive and stop there. Diabetic retinopathy, damage to the retina caused by diabetes, affects an estimated [9.6 million Americans](https://www.cdc.gov/vision-health-data/prevalence-estimates/dr-prevalence.html) and is the [leading cause of blindness](https://www.cdc.gov/diabetes/diabetes-complications/diabetes-and-vision-loss.html) in working-age adults.\n\n**Medicare pays for the photo, not the result**\n\nMedicare’s 2026 national rate for an AI eye exam is [$46.76](https://www.cms.gov/medicare/payment/fee-schedules/physician/pfs-relative-value-files) in a doctor’s office and $60.27 in a hospital outpatient clinic. It includes nothing for a doctor’s time, which makes sense, because no doctor reads the image. The trouble is what else the payment leaves out.\n\nIt covers the photo. It doesn’t cover the medical assistant who takes it, the patient whose photo is too blurry to read, getting the result into the medical record, or the calls it takes to get a positive patient into an eye clinic. Medicare does pay for some care coordination, but nothing ties that money to making sure a patient with a positive screen is seen. Patients in traditional Medicare also [generally pay 20%](https://www.medicare.gov/coverage/eye-exams-for-diabetes) of the cost once they’ve met their annual deductible, about $9.\n\nPayment isn’t the only barrier, but it shapes all the others.\n\n**The scan is the easy part**\n\nDay-to-day logistics are where programs stall. Without eye drops to widen the pupils, five academic programs [reported](https://pubmed.ncbi.nlm.nih.gov/41140908/) that only 49% to 75% of patients produced a photo clear enough to read. At one academic medical center, [just 31%](https://pubmed.ncbi.nlm.nih.gov/35970206/) of patients 70 and older did; adding the drops raised it to 85%. Few primary-care clinics use the drops, so most send patients with unreadable photos for a full eye exam anyway, and for them the AI saves nothing. Add staff turnover, finding a spare room and getting results into the medical record, and many clinics decide it isn’t worth it.\n\nThen comes the step no one is specifically paid to finish. In two U.S. screening programs where eye photos were read remotely, in [Houston](https://pubmed.ncbi.nlm.nih.gov/39710903/) and [New York](https://pubmed.ncbi.nlm.nih.gov/37006661/), only 55% to 59% of patients with a positive result reached an eye specialist. In a [county hospital system](https://pubmed.ncbi.nlm.nih.gov/32576560/), about half of patients with sight-threatening disease were seen within a year, and only 5% within the recommended time. Specialists are scarce too: federal projections show the number of ophthalmologists [falling 12% by 2035 while demand rises 24%](https://pubmed.ncbi.nlm.nih.gov/37739231/), and rural areas can meet less than a third of their need. Sending unreadable and milder cases to optometrists would ease the load; serious disease still needs an ophthalmologist.\n\nThe quality score doesn’t help. The main measure insurers are graded on, which feeds the star ratings Medicare gives private plans, gives full credit for a completed screen whatever it finds. That year, a positive result that never reaches an eye doctor scores the same as a clean one.\n\nAI helps when someone owns what happens next. A [2026 review](https://pubmed.ncbi.nlm.nih.gov/41998218/) of six studies found AI screening nearly doubled the share of patients who followed up with an eye doctor, with the biggest gains where the referral process itself was redesigned.\n\n**Three fixes**\n\n**1. Pay for the follow-up, not just the photo.** Medicare should let its existing care-coordination payments cover chasing a positive screen until the patient sees an eye doctor, and test a payment released only once that visit takes place. Private Medicare Advantage plans can do this today, and make the screen free to patients.\n\n**2. Grade the follow-up, not just the screen.** Insurers are scored on whether patients got screened, not on whether positive patients got care. This year Medicare [kept the eye-exam score](https://www.federalregister.gov/documents/2026/04/06/2026-06600/medicare-program-contract-year-2027-and-certain-contract-year-2026-policy-and-technical-changes-to) and said it would consider adding follow-up. It should, and start now: new quality scores take years to build.\n\n**3. Make someone responsible for every positive screen.** Today a positive result belongs to no one. The insurer or doctor group paid to keep the patient healthy should own it until the patient is seen. Some of this needs no new rules: clinics can book the eye appointment before the patient leaves and require the result to come back. England’s national program, which follows each patient from invitation to referral, [screens about 79%](https://www.gov.uk/government/publications/diabetic-eye-screening-2022-to-2023-data) of the people it invites, though even it struggles to get urgent cases seen on time.\n\n**The next wave will hit the same wall**\n\nDiabetic eye screening is the first AI that works without a doctor to get a billing code, a quality score and a medical guideline behind it, which makes it a preview. Medicare’s [proposed 2027 rule for hospital outpatient care](https://www.federalregister.gov/documents/2026/07/07/2026-13656/medicare-program-hospital-outpatient-prospective-payment-and-ambulatory-surgical-center-payment) would put it and dozens of other software tools in a new payment category, which Medicare calls a temporary step toward approaches “that may better align payment with clinical outcomes.” In plain terms: paying for results. That is the right goal. For patients, the question isn’t whether software gets paid for. It’s whether the people it flags get care. Keep paying for the photo instead of the result, and the next wave of AI will repeat this story: authorized, paid for and largely unused.\n\n*Photo: Jay_Zynism, Getty Images*\n\n[Dr. Ashkan Eliasy](https://www.linkedin.com/in/ashkaneliasy/) is Managing Director of [Nuvisy Strategic Advisory](https://nuvisy.com/), advising medtech and AI-health companies and investors. He previously served as an investment principal at a global ophthalmic device company, co-founded medtech and fintech startups and holds a PhD and an MBA. Disclosure: he advises companies in ophthalmic technology.\n\nThis post appears through the **MedCity Influencers** program. Anyone can publish their perspective on business and innovation in healthcare on MedCity News through MedCity Influencers. [Click here to find out how](https://medcitynews.com/medcity-influencers/?rf=1&__hstc=212719371.8259b40d019fae28e136d397caf529df.1712764039103.1715704787503.1716390589114.5&__hssc=212719371.8.1716396490210&__hsfp=4136036889).", "url": "https://wpnews.pro/news/medicare-pays-for-the-ai-eye-scan-nobody-pays-for-what-comes-next", "canonical_source": "https://medcitynews.com/2026/11/medicare-pays-for-the-ai-eye-scan-nobody-pays-for-what-comes-next/", "published_at": "2026-11-17 14:18:00+00:00", "updated_at": "2026-11-17 19:51:43.766192+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-policy", "computer-vision"], "entities": ["Medicare", "Food and Drug Administration", "American Diabetes Association", "JAMA Ophthalmology", "Centers for Medicare & Medicaid Services"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/medicare-pays-for-the-ai-eye-scan-nobody-pays-for-what-comes-next", "markdown": "https://wpnews.pro/news/medicare-pays-for-the-ai-eye-scan-nobody-pays-for-what-comes-next.md", "text": "https://wpnews.pro/news/medicare-pays-for-the-ai-eye-scan-nobody-pays-for-what-comes-next.txt", "jsonld": "https://wpnews.pro/news/medicare-pays-for-the-ai-eye-scan-nobody-pays-for-what-comes-next.jsonld"}}