Evaluation of a Large Urban Teleretinal Diabetic Retinopathy Screening Program in the United States



Diabetes mellitus is a global epidemic, both in the United States and around the world. Diabetic retinopathy is currently the leading cause of blindness amongst working-age Americans. Its devastating consequences are generally preventable with early detection, but unfortunately, most patients are not compliant with annual dilated examinations. In 2013, the Harris Health System in Houston, Texas, United States initiated a tel eretinal screening program to increase the ophthalmic surveillance of diabetic patients.


To evaluate the utility of a teleretinal screening program in detecting diabetic retinopathy and diabetic macular edema in a county population.



  • determine the accuracy of teleretinal (TR) screening in detecting diabetic retinopathy (DR) and diabetic macular edema (DME) and
  • assess the compliance with follow-up of TR patients referred for in-clinic examination


Retrospective, observational clinical study in the Harris Health System (HHS, Houston, TX) where patients with diabetes mellitus were screened via a non-mydriatic fundus camera (DRS, CenterVue, Padova, Italy). Images were manually graded by an eye care provider, and patients were referred for in-clinic exam if the image was consistent with severe nonproliferative diabetic retinopathy (DR), proliferative DR, a non-DR condition, or an ungradable rating. All patients screened and referred for in-clinic exam between 2013-2016 were cross-referenced with patient charts. Information obtained from the charts included in-clinic exam findings and diagnosis as well as patient follow-up status.


80,804 screenings were performed during the study period. Diabetic retinopathy was detected in 42% (33,961) of all screenings while non-DR pathology was detected in 14.6% (11,804). Approximately 5% of all patients were found to have severe nonproliferative DR or worse. Overall, between TR and clinical exam diagnoses, there was moderate agreement (kappa = 0.45, 95% confidence interval [CI] 0.27-0.52) in grading DR. There was agreement within one level of DR severity in 83.0% (95% CI 79.2-86.4%) of patients. The positive predictive value (PPV) for detecting referable-level DR was 71.8% (95% CI 66.3-76.8%). The PPV for detecting center-involving DME was 31.1% (95% CI 24.9-37.9%). When a subset of this population was evaluated, 51.2% (456/891) of patients referred for an in-clinic exam actually attended a clinic appointment. Of those requiring a second follow-up appointment within 4 months, 75% attended this follow-up appointment.


The HHS teleretinal screening program has been successful in improving compliance of diabetic patients with annual eye examinations. TR screening was highly predictive for referable-level DR, but less reliable in detecting non-DR conditions. Fewer than half of referred patients attended a clinic appointment, suggesting there may be additional barriers to ophthalmic care in a county population.




Christina Y. Weng
Houston, United States
Email : christina.weng@bcm.edu
Cell Phone: +15862914400
Work Phone: +15862914400