How One Health System Is Embedding Diabetic Retinopathy Screening Into Primary Care

Some hospitals are finding it difficult to scale AI across their enterprise — with this process often complicated by things like inadequate technology infrastructures, varying needs across specialties, and the need for strong data governance.

When it comes to deploying AI at hospitals, it’s best to start small, iterate, and then scale thoughtfully, said Dr. Kathleen Provanzana, medical director at OhioHealth, during a webinar this week.

“The process begins with action — and don’t let perfect get in the way of good,” she declared.

Her Central Ohio-based health system operates 16 hospitals, hundreds of ambulatory sites, and about 100 primary care offices. It is currently scaling an AI solution that it began piloting in 2022: Digital Diagnostics’ AI system that autonomously diagnoses diabetic retinopathy. The system, called the LumineticsCore platform, uses FDA-cleared AI models to detect diabetic retinopathy from retinal images taken during a primary care visit.

The AI delivers same-day diagnostic results without the need for a specialist, allowing for immediate follow-up care.

When OhioHealth began its partnership with Digital Diagnostics, the system had about 45,000 diabetic patients in its primary care population. Now that figure has risen to more than 50,000 — and 25-30% of these patients have diabetic retinopathy, Dr. Provanzana said.

Diabetic retinopathy is a diabetes-related eye disease that damages blood vessels in the retina and can lead to vision loss if not detected and treated early.

When it began piloting Digital Diagnostics’ solution, OhioHealth deployed 10 diagnostic cameras. The health system’s main goal was to improve its diabetic retinopathy screening rate, which was low at 35% in 2021, Dr. Provanzana noted. She also noted additional aims of achieving same-day diagnoses, reducing burden on ophthalmologists and increasing clinic efficiency.

“I would also share that it reduces some of the burden on our referral staff. Now that we can screen a patient in the office, we don’t have to do a referral to an external eye care provider unless the patient unfortunately screens [positive],” she explained.

The platform’s pre-visit planning features flag diabetic patients who are due for eye exams, helping primary care staff to easily prepare orders. Diagnostic results are also delivered immediately in the EHR and provider inbox, and positive results include a referral hyperlink with next steps, Dr. Provanzana said.

Throughout the pilot process, OhioHealth learned how much site selection matters — Dr. Provanzana pointed out that the health system now knows to prioritize primary care offices with high volumes and underserved populations.

She also highlighted the importance of provider buy-in.

“We have some staff who have seen the ravages of diabetes — affecting patients or loved ones, robbing them of their sight — and they became champions in their offices. We had to have providers who really saw the value of screening in primary care. We had some providers that were very much tied to the gold standard of a full ophthalmologic exam — and there’s no harm in that, certainly. But if your patient’s not getting to that ophthalmologist, then what good are we doing by holding on to that idea?” Dr. Provanzana explained.

OhioHealth now has about 40 of Digital Diagnostics’ cameras located across its footprint, she stated.

One reason that OhioHealth is scaling its use of this AI tool is because it fits well into existing workflows, Dr. Provanzana noted.

Overall, OhioHealth saw a 162% increase in diabetic eye exams within 12 months of adopting Digital Diagnostics’ solution — as well as a 1.7 increase in positive diagnoses year-over-year.

Once integrated into staff members’ workflows, the AI-based screening becomes routine — just like checking a vital sign, Dr. Provanzana declared.

Photo: Flickr user Rakesh Rocky