Eyes on Site

Transforming retinal screening for diabetic patients


Diabetic retinopathy (DR) is the most common cause of vision loss among working-age adults. Treatment of late-stage DR can cost as much as $28,000 per patient, while early-stage treatment costs are minimal. 

No one has to lose vision to diabetic retinopathy — those who do usually don't know that they have it or find out too late. Early screening and treatment prevent vision loss in 90 percent of cases. However, because it is asymptomatic until later stages, many DR patients are unaware of the disease. 

Traditional eye exams require an office visit to an eye doctor, take considerable time, and require pupillary dilation, which causes hours of disruptive blurred vision. Because of this time consuming and negative experience, some patients are reluctant to complete screening.

In 2015, only 22 percent of Penn Medicine patients met recommended screening standards for diabetic eye care. However, 8,900 of these patients came to Penn Medicine for other services. These "on-site "health care interactions represented a missed opportunity to assist non-adherent patients in getting sight-saving diabetic eye exams.


Eyes on Site (EOS) is a retinal screening model that makes it easy for patients to meet recommended screening standards for diabetic eye care.

Rather than requiring a standalone visit for screening, EOS is deployed in Penn Medicine locations with high volumes of diabetic patients, such as endocrinology and primary care. Patients are offered free, rapid retinal screening while they are on-site for other appointments. Most notably, EOS leverages non-mydriatic cameras - which do not require pupil dilation - to conduct screening.

Images are interpreted remotely, and findings are quickly communicated back to patients and providers.


EOS offers a superior retinal screening experience and drives better patient outcomes.

We saw impressive results in our pilot population.

  • Increased screening rates: We engaged non-adherent patients in diabetic eye screenings. Of the 125 patients screened, 47 percent did not know when their last screening was or reported never completing one. 
  • Improved patient experience: EOS screenings are 18x faster than traditional retinal screenings, and they don't require pupil dilation, which means patients can drive themselves home afterward.
  • Earlier diagnosis: We discovered previously undiagnosed diseases within pilot participants, with 57 percent requiring a referral, either for baseline or detected eye disease, and 11 percent presenting with evidence of DR.

The EOS model enables earlier diagnosis and treatment for DR, thereby reducing poor patient outcomes and decreasing the need for costly late-stage diabetic eye disease treatments. If EOS screened 890 patients yearly, we predict 356 cases of undiagnosed disease, amounting to more than $290,000 in cost avoidance.


When the COVID-19 pandemic reached Penn Medicine, the EOS model provided a foundation for the development of the Telemedicine Enhancement Pathway (TEP). TEP expedites in-person ophthalmology care through the use of advanced retinal imaging and eye testing. After on-site visits, ophthalmologists review studies and complete the anterior portion of eye exams via telemedicine. TEP reduces the time patients spend in the office from up to three hours down to only 30 minutes. It also enables for safe social distancing for patients whose medical conditions require a full in-person exam.

Phase 2: It does work

Thomasine Gorry, MD, MGA
Eydie Miller, MD
Tomas Aleman, MD

Innovation leads

Davis Hermann, MiD
Roy Rosin, MBA


Innovation Accelerator Program

Innovation Methods

Frequently, our work involves reimagining processes to achieve better results.
The concept of deletion is about isolating part of a process that is problematic and solving for its absence. Imagine solving for airport security without the security line or collecting tolls without tollbooths.
With this technique, you're not allowed to improve the flawed part of a process. You must delete it and then, making sure you understand what essential benefits it provided, if any, introduce alternative approaches.

The EOS model was conceived of, in part, by using the deletion method.

We visualized the patient journey for the traditional eye exam and removed some of the drivers of negative patient experience. For example, if we took away pupil dilation, how could we solve for its absence?

High fidelity learning can come from low fidelity deployment.
Mini-pilots will allow you to learn by doing, usually by deploying a fake back end. You might try a new intervention with ten patients over two days in one clinic, using manual processes for what might ultimately be automated.
Running a "pop-up" novel clinic or offering a different path to a handful of patients will enable you to learn what works and what doesn't more quickly. And, limiting the scope can help you gain buy-in from stakeholders to get your solution out into the world with users and test safely.

To test the proposed EOS model, we ran time-limited mini-pilots at three clinics.

These pilots helped us to quickly learn what did and didn’t work about the EOS model. We were also able to test risky assumptions. For example, we demonstrated that we could capture adequate retinal images in various settings without pupil dilation.


Aha moments during pilot testing

During initial testing of the EOS model, we had several a-ha moments. For example, we observed that front desk staff was as influential as the physician when recommending using the new technology and that technology only underscored that relationships matter. Embedding the technology-based screening in a care environment is what really makes this model work.

Thomasine Gorry, MD, MGA


Pitch Day 2016