The agency set a national payment amount for AI-based screenings for diabetic retinopathy in its 2022 Medicare Physician Fee Schedule final rule. It would also allow the screenings to count toward quality measures.
The Centers for Medicare and Medicaid Services (CMS) will cover an autonomous screening tool and allow it to count toward quality measures. As part of its 2022 Medicare Physician Fee Schedule, CMS set a national payment amount for using AI to detect diabetic retinopathy.
It also included these screenings as a quality measure, saying that eye exams must be performed by an ophthalmologist or optometrist, or that retinal images must be read by a system that provides an AI interpretation.
Diabetic retinopathy is a condition where high blood sugar levels can cause damage to the blood vessels in the retina. It can go undetected until its later stages, when it starts to affect patients’ vision.
Because of this, annual eye exams are recommended for patients with Type 1 or Type 2 diabetes. But in practice, that doesn’t always happen. One study of hospitalized patients with diabetes found that less than half of them had an eye exam in the last year.
Some commenters contended that quality measures should only apply when the software system is FDA approved, or that equally considering a comprehensive eye exam and an AI-based screening to denote care quality is not appropriate.
In response, CMS wrote, “We believe that allowing AI will increase the accessibility and therefore occurrence of yearly eye exams, leading to quality outcomes in this patient population.”
The decision was also cheered by two companies that make AI-based screening systems for diabetic retinopathy, Digital Diagnostics and Eyenuk.
“These final regulations will allow eye exams to be administered at patients’ point-of-care where they receive diabetes management eliminating many barriers to care and increasing access,” Digital Diagnostics Founder and Chairman Dr. Michael D Abramoff said in a news release.
Digital Diagnostics (formerly known as IDx) got a de novo clearance from the Food and Drug Administration in 2018 for its software to automatically detect diabetic retinopathy. Images taken with a retinal camera are uploaded and processed by its algorithm, which screens for moderate- to severe- retinopathy. It would either indicate to a physician that diabetic retinopathy is present or that the screening was negative.
Notably, the device was cleared to make screening decisions on its own, meaning it could be used by primary care physicians and other providers that don’t specialize in eye care.
While physicians in a Nature editorial touted the clearance as an important milestone, they also raised up a few shortcomings with these automated tools. For instance, they wrote that the study that served as the basis for the device’s FDA decision was still relatively small in terms of diagnostic accuracy, and that because it is only designed to detect diabetic retinopathy, it might miss other conditions that would be detected in an eye exam, such as glaucoma or macular degeneration.
Eyenuk, for its part, got 510(k) clearance for its system last year.
