The importance of screening for diabetic retinopathy has been established, but the best method for screening has not yet been determined. We report on a trial of assessment of digital photographs by telemedicine compared with standard retinal photographs of the same fields and clinical examination by ophthalmologists. A total of 129 diabetic inpatients were screened for diabetic retinopathy by slit-lamp biomicroscopy performed by an ophthalmologist and by two-field 50 degrees non-stereo digital fundus photographs assessed by six screening centers that received the images by electronic mail. Conventional 35-mm transparencies of the same fields as the digital photographs were assessed by a retinal specialist and served as the reference method for detection of diabetic retinopathy. Slit-lamp biomicroscopy was the reference method for the detection of macular edema. The prevalence of any form of diabetic retinopathy was 30% (n = 35); of sight-threatening retinopathy including macular edema, the prevalence was 6% (n = 7). The assessment of digital images by the six screening centers resulted in a median sensitivity of 85% and a median specificity of 90% for the detection of moderate nonproliferative or sight-threatening diabetic retinopathy. Clinically significant macular edema (n = 4) was correctly identified in 15 of the 24 grading reports. An additional seven reports referred the patients for further investigation because of concurrent diabetic retinopathy. Telescreening for diabetic retinopathy by an assessment of two-field 50 degrees non-stereo digital images is a valid screening method. Although detection of clinically significant macular edema using biomicroscopy is superior to digital or standard non-stereo photographs, only few patients with sight-threatening diabetic retinopathy are missed.