Use of telemedicine in screening for diabetic retinopathy

A diabetic retinopathy screening program was introduced in September 2000 at a university-affiliated hospital in Montreal to meet the demands of an overburdened health care system. In this article we describe our initial experience with this program.

A Canon CR6-45NM nonmydriatic 45 degrees camera was installed in the outpatient endocrinology clinic of the hospital. All patients who visited the clinic between September 2000 and January 2001 agreed to participate in the program. For each patient, a short questionnaire (age, sex, type of diabetes mellitus, time since diagnosis, prior retinal laser therapy and presence of end-organ involvement) was filled in. Subsequently at least two photographs were obtained of each eye. The photographs were sent via the hospital intranet to a retinal specialist within the hospital. The specialist reviewed the images unchanged and sent a report (grading of the level of diabetic retinopathy, presence or absence of macular hard exudates, quality of the images obtained and follow-up recommendations) via the hospital intranet to the endocrinology service. If the images were judged to be of poor quality and could not be read by the retina specialist, the patient was asked to see his or her regular ophthalmologist.

Images of 830 eyes of 415 patients (211 men and 204 women with a mean age of 57.5 years) were obtained. Most (83.6%) had type 2 diabetes. The average duration of disease was 13.3 years. Macular hard exudates were observed in 50 eyes (6.0%). Nonproliferative diabetic retinopathy was noted in 117 eyes (14.1%) and proliferative diabetic retinopathy in 15 eyes (1.8%). Unexpected findings included epiretinal membranes (in three eyes), macular holes (in two), bilateral cotton-wool spots (in one patient) and central retinal vein occlusion (in one eye). Overall, 10% of the patients were referred to a retina specialist based on the screening photographs. The main difficulty encountered with the screening program was image inconsistency. Overall, 35% of the images graded were felt to be of poor quality, the most common reason being poor exposure (84.4%). The quality of the images improved significantly over the study period (p < 0.01). Images obtained from patients aged 65 years or more were consistently poorer than those obtained from patients less than 65 years of age (p < 0.001).

Despite the imperfections of such a system, our experience was favourable, allowing us to screen large numbers of patients in a cost-effective, reliable manner.