Data indicate that only about one-half of people with diabetes are properly screened for diabetic retinopathy (DR), and many patients at high risk for vision loss do not receive treatment. “Annual DR screening has been recommended by guidelines for years, but many Americans have limited access to these screenings,” says Yi Zhang, MD, PhD.  “A more practical alternative is to perform DR screening when caring for patients at primary care offices and then referring those who need further eye care to ophthalmologists or retina specialists. By bringing DR screening to primary care, we may be able to identify this treatable condition earlier and improve visual outcomes.”

For a study published in BMC Ophthalmology, Dr. Zhang, Jeffrey D. Henderer, MD, and colleagues evaluated how often positive retinopathy screening exams via telehealth resulted in a completed ophthalmology appointment among patients at the Lewis Katz School of Medicine at Temple University. Data from a 15-month period was reviewed to investigate how many patients were screened, the interpretability of remote fundus photographs, how often these photographs led to a DR diagnosis, and how many patients followed up for an office exam, if indicated.

Telemedicine DR Screening Practical, Efficient But Needs Refining

“We found that 32.6% of remote fundus images were graded as uninterpretable, which was higher than what we expected and adversely affected the quality of our screening,” Dr. Zhang says. “This finding suggests that it’s critical to provide staff sufficient training and timely feedback to shorten their learning curve with remote fundus photography and ensure that they are proficient in this capability.”

Among all screening exams, slightly more than one-half triggered a request for a referral to ophthalmology. Roughly 55% of the total referrals generated were due to an inability to assess for retinopathy in at least one eye (Figure). “We also found that only 9.5% of patients who needed a referral eventually received an eye exam at our offices,” Dr. Zhang says. “This reminds us that a successful telemedicine screening program must close the communication gap between screening and timely follow-up my eye care specialists. Mere identification of DR, or an uninterpretable photo, is not enough.”

Dr. Zhang says that the study highlights that primary care-based telemedicine is a practical, efficient way to carry out DR screening, especially among at-risk urban communities. “That said, our current technique needs improvement,” he says. “Currently, the mainstay of DR screening is an in-person visit at an eye care provider office, but this is inconvenient for patients, inefficient for providers, and costly for the healthcare system. Our study showed that more than one-half of patients did not have any DR, so these people did not require a more costly ophthalmology eye exam. A primary care-base telemedicine DR screening can be a cost-efficient alternative to in-person screening.”

Artificial Intelligence-Assisted Telemedicine DR Screening & Beyond

Recently, artificial intelligence (AI)-assisted telemedicine has become a trend in population-based disease screenings, according to Dr. Zhang. “In stage 2 of our project, we’re adding an AI system to our DR screening to potentially overcome issues with poor photo quality and follow-up,” he says. “We want to take advantage of AI to help speed image interpretation and potentially identify other common eye diseases, such as glaucoma and age-related macular degeneration. To achieve this goal, AI and clinicians will both review the initial remote fundus images in parallel. This will hopefully provide the best of both worlds—the speed and efficiency of AI and the ability to screen for other eye diseases. If only diabetes is present, patients will have AI-only reads for the next 2 years, at which time another AI and human read will occur. This protocol will give us a glimpse into the efficiency of our primary care-based telemedicine system to screen for other eye diseases.”

“In addition, we’re trying to refine criteria for which patients need an annual remote fundus photograph,” says Dr. Henderer.  “Since most patients with diabetes do not have DR, determining criteria that are predictive of having DR may help us better target patients for appropriate DR screening.  Length of time with diabetes and glucose control are predictive of retinopathy, but we want to determine if other biomarkers can be combined with this information to better identify higher risk patients.”