The American Diabetes Association (ADA) recently released its latest Standards of Medical Care in Diabetes publication, which normally wouldn’t raise any eyebrows in the optometry world—except for this year. The ADA’s 2020 standards finally make a formal leap of faith with artificial intelligence (AI), stating that “systems that detect more than mild diabetic retinopathy and diabetic macular edema (DME) authorized for use by the FDA represent an alternative to traditional screening approaches.”¹

New Game in Town

The FDA cleared the first AI system for diabetic retinopathy and diabetic macular edema screening in April 2018, the IDx-DR.² The device was cleared as a stand-alone tool that “provides a screening decision without the need for a clinician to also interpret the image or results,” according to the FDA release. Such autonomy “makes it usable by health care providers who may not normally be involved in eye care.”²

But the system isn’t perfect, according to A. Paul Chous, OD, whose practice focuses on diabetes eye care and education. “It had a 13% false-negative rate for detecting ‘more than mild DR,’ does not evaluate the peripheral retina whatsoever, and has no sensitivity data reported for the detection of DME, the leading cause of vision loss due to diabetes and the detection of which multiple raster SD-OCT is the gold standard,” he points out.

Some health clinics—including grocery store chain Albertsons’ CarePortMD—have already embraced the technology, but with some caveats.³ In these retail settings, the IDx-DR system will require a prescription (only those at risk are screened) and is performed by a registered nurse, according to a press release.³

Play at Your Own Risk

While the inclusion of AI in the standards is a significant change, the ADA included a word or two of caution when considering implementing the device: “the benefits and optimal utilization of this type of screening have yet to be fully determined. Artificial intelligence systems should not be used for patients with known retinopathy, prior retinopathy treatment, or symptoms of vision impairment.”² The standards further emphasize that remote image reading, including use of AI, is not a substitute for a dilated eye examination by an optometrist or ophthalmologist, and that patients with previously diagnosed DR of any severity should see an eye care provider, explains Dr. Chous.

“This makes sense, given that the algorithm is not designed to assess change in individual patients’ DR status, as well as the fact that patients with diabetes are more likely to experience a number of ocular conditions apart from DR, including cataract, ocular surface disease and glaucoma,” he adds.

For most clinicians, implementation of the technology still raises a host of questions, including concern that patients with early signs of DR, such as peripheral lesions, given the green light by the AI system will forego a clinical exam. “What about the 30% of patients who have subclinical DME only detected by OCT and who are significantly more likely to develop significant DME over time?” Dr. Chous adds. 

Nonetheless, Dr. Chous is convinced AI is here to stay, and it will improve. “It may very well help more patients be screened for DR and save vision, especially in underserved and some rural communities. It is not designed, however, to detect DR at its earliest stages when prevention and patient education are key. Nor will it replace knowledgeable providers when DR progresses and clinical acumen, experience and more sophisticated/comprehensive examination and technology is requisite.”