Telemedicine and Retinal Imaging for Improving Diabetic Retinopathy Evaluation | Diabetic Retinopathy | JAMA Internal Medicine | JAMA Network
[Skip to Navigation]
Figure. Percentage of patients with a documented retinal examination over a 12-month period in a University of North Carolina primary care clinic.

Figure. Percentage of patients with a documented retinal examination over a 12-month period in a University of North Carolina primary care clinic.

Research Letter
Nov 26, 2012

Telemedicine and Retinal Imaging for Improving Diabetic Retinopathy Evaluation

Author Affiliations

Author Affiliations: Department of Ophthalmology (Drs Garg, Jani, and King) and UNC Kidney Center (Dr Kshirsagar), University of North Carolina, Chapel Hill; and Hamilton Eye Institute, University of Tennessee, Memphis (Dr Chaum).

Arch Intern Med. 2012;172(21):1677-1680. doi:10.1001/archinternmed.2012.4372

Early detection of diabetic retinopathy (DR) is crucial to prevent blindness. Timely intervention with laser therapy can reduce the risk of severe vision loss by more than 90%.1 Despite well-accepted national and international guidelines, on average, less than 50% of patients with diabetes mellitus (DM) undergo screening for retinopathy in the United States.2,3 Barriers to care include socioeconomic factors, geography, lack of patient education, and cultural barriers among minorities.

Telemedicine is an emerging strategy for improving DR screening through retinal imaging with remote expert interpretation. Introducing this technology at the point of care of the primary care physician could substantially reduce barriers and improve early detection of retinopathy.4,5 Outside of the Department of Veterans Affairs (VA) system,6,7 effectiveness data for telemedicine screening in the United States is limited.

The purpose of this study was to evaluate the impact of telemedicine on DR screening in a primary care setting with a diverse patient and payer mix. Predictors of DR were determined by analyzing clinical patient characteristics obtained at the time of image acquisition.


Institutional review board approval was obtained. The DR screening frequency at a University of North Carolina primary care clinic was determined prior to and 12 months after the implementation of the TRIAD ocular telemedicine network.8 Patients who did not have a documented retinal examination within the prior 12 months were identified. A staff member obtained retinal photographs of both eyes with a nonmydriatic fundus camera along with patient demographic and clinical information. The images were securely sent over a Health Insurance Portability and Accountability Act (HIPAA)-compliant, web-based network to a single retina specialist (S.G.), who remotely classified the retinal images according to severity of retinopathy. The diagnostic and management report was transmitted to the original primary care provider within 48 hours and incorporated into the patient's electronic medical records. Based on the degree of DR, patients were either scheduled for repeated follow-up photographs in the primary care clinic or referred to an ophthalmologist for further management. Logistic regression analysis was used to determine predictors of retinopathy.


A total of 1002 patients with DM were evaluated for retinopathy within the telemedicine network in 1 year. Of these, 869 patients had no retinopathy. The mean age was 57 years, mean hemoglobin A1c (HbA1c) level was 7.6%, mean total cholesterol was 184 mg/dL (to convert to millimoles per liter, multiply by 0.0259), and mean duration of DM was 9.0 years. Twenty-two percent of patients had a smoking history, 78% had hypertension, 10% had coronary artery disease, 8% had a history of stroke, 9% had a history of myocardial infarction, and 6% had kidney disease. While the distribution of black and white patients was similar (48% vs 49%, respectively), the prevalence of DR was significantly higher among black (71%) vs white (27%) patients and was also more severe: 11.5% of black patients had mild nonproliferative retinopathy and 4.3% had moderate retinopathy vs 3.6% of white patients with mild nonproliferative retinopathy and 1.8% with moderate retinopathy.

The Figure shows the percentage of diabetic patients undergoing retinal evaluation during the observation period. The screening frequency improved from 32% to 71%. Race (nonwhite) was strongly associated with DR (odds ratio, 2.2; 95% CI, 1.2-4.1) along with HbA1c and duration of DM. The odds ratio of DR for every 1-point increase in HbA1c level was 1.5 (95% CI, 1.3-1.7) and for every 5-years' duration of DM was 1.5 (95% CI, 1.3-1.7).


We found that point-of-care retinal imaging with remote interpretation in patients with type 1 and 2 DM improved the frequency of retinal screening from 32% to 71% in only 12 months.

Visual impairment from DR results in loss of personal independence and economic productivity and decreased quality of life. As the prevalence of DM is projected to increase from 25 million Americans to 125 million Americans by the year 2050,9 the number of patients requiring annual retinal evaluation will far exceed the capacity of ophthalmologists.

Primary care physicians are at the frontline of this epidemic and already play a critical role in primary prevention of retinopathy with the management of serum glucose and lipid levels and blood pressure. Telemedicine potentially allows primary care physicians to manage the screening and monitoring of this potentially blinding disease. Specifically, they can distinguish patients who only require surveillance with retinal photography from those who need urgent referral. Such a paradigm could lead to better use of physician and patient resources. In our group, for example, most patients (89%) did not have retinopathy and therefore did not need referral to an ophthalmologist for DR screening.

Telemedicine screening at the point of care of the primary care physician represents a potential paradigm shift in the management of DM, can improve screening, and may ultimately prevent vision-threatening DR.

Back to top
Article Information

Correspondence: Dr Garg, Department of Ophthalmology, University of North Carolina, Chapel Hill, 5132 Bioinformatics Bldg, CB #7040, Chapel Hill, NC 27599-7040 (

Published Online: October 1, 2012. doi:10.1001/archinternmed.2012.4372

Author Contributions:Study concept and design: Garg, Kshirsagar, and Chaum. Acquisition of data: Garg and Chaum. Analysis and interpretation of data: Garg, Jani, Kshirsagar, and King. Drafting of the manuscript: Garg, Jani, Kshirsagar, and King. Critical revision of the manuscript for important intellectual content: Garg, Kshirsagar, and Chaum. Statistical analysis: King. Obtained funding: Chaum and Garg. Administrative, technical, and material support: Garg, Jani, and Chaum. Study supervision: Garg and Kshirsagar.

Financial Disclosure: Dr Chaum is the principal investigator of a National of Institutes of Health grant “Automated Screening for Diabetic Retinopathy by Content” (R01-EY017065), under which some of the technology used in the TRIAD telemedical network described in this Research Letter has been developed. Dr Chaum is an author on a patent application describing technology that is, in part, being developed and implemented in the TRIAD network. The intellectual property of this patent (when issued) belongs to the University of Tennessee. Dr Chaum is a cofounder and equity holder of Hubble Telemedical, a start-up company that has licensed the technology developed under the aforementioned grant for use in the commercial space under a transparent conflict management plan in place through the Office of Research and Office of Finance at the University of Tennessee Health Science Center.

Funding/Support: This study received funding from the National Institutes of Health (grant R01-EY017065); Prevent Blindness North Carolina; Research to Prevent Blindness, New York, New York; and the Plough Foundation, Memphis, Tennessee.

Previous Presentation: This study was presented as a poster at the Association for Research in Vision and Ophthalmology International Conference; May 10, 2012; Ft Lauderdale, Florida.

Additional Contribution: Sam Weir, MD, Family Medicine, University of North Carolina, contributed to the implementation of telemedicine in the primary care clinic and did not receive compensation.

Early Treatment Diabetic Retinopathy Study Research Group.  Early photocoagulation for diabetic retinopathy: ETDRS report number 9.  Ophthalmology. 1991;98(5):(suppl)  766-7852062512PubMedGoogle Scholar
Brechner RJ, Cowie CC, Howie LJ, Herman WH, Will JC, Harris MI. Ophthalmic examination among adults with diagnosed diabetes mellitus.  JAMA. 1993;270(14):1714-17188411502PubMedGoogle ScholarCrossref
Lee PP, Feldman ZW, Ostermann J, Brown DS, Sloan FA. Longitudinal rates of annual eye examinations of persons with diabetes and chronic eye diseases.  Ophthalmology. 2003;110(10):1952-195914522771PubMedGoogle ScholarCrossref
Ogunyemi O, Terrien E, Eccles A,  et al.  Teleretinal screening for diabetic retinopathy in six Los Angeles urban safety-net clinics: initial findings.  AMIA Annu Symp Proc. 2011;2011:1027-103522195163PubMedGoogle Scholar
Zimmer-Galler I, Zeimer R. Results of implementation of the DigiScope for diabetic retinopathy assessment in the primary care environment.  Telemed J E Health. 2006;12(2):89-9816620162PubMedGoogle ScholarCrossref
Cavallerano AA, Conlin PR. Teleretinal imaging to screen for diabetic retinopathy in the Veterans Health Administration.  J Diabetes Sci Technol. 2008;2(1):33-3919885175PubMedGoogle Scholar
Kerr EA, Gerzoff RB, Krein SL,  et al.  Diabetes care quality in the Veterans Affairs Health Care System and commercial managed care: the TRIAD study.  Ann Intern Med. 2004;141(4):272-28115313743PubMedGoogle Scholar
Li Y, Karnowski TP, Tobin KW,  et al.  A health insurance portability and accountability act-compliant ocular telehealth network for the remote diagnosis and management of diabetic retinopathy.  Telemed J E Health. 2011;17(8):627-63421819244PubMedGoogle ScholarCrossref
Boyle JP, Thompson TJ, Gregg EW, Barker LE, Williamson DF. Projection of the year 2050 burden of diabetes in the US adult population: dynamic modeling of incidence, mortality, and prediabetes prevalence.  Popul Health Metr. 2010;8:2920969750PubMedGoogle ScholarCrossref