Abstract
Purpose: :
To determine if primary care physicians (PCPs) can accurately detect diabetic retinopathy using fundus photographs from a non–mydriatic camera.
Methods: :
Sixteen primary care physicians from rural/migrant health centers received training on basic retinal anatomy and pathology specifically related to diabetic retinopathy. Patients with diabetes were invited to join the study during routine office visits and one 45 degree photo centered on the fovea was taken through an undilated pupil. One of sixteen PCPs and a retina specialist interpreted the images. Outcome variables included agreement on presence or absence of specific findings in diabetic retinopathy and referral rate. Threshold for referral was defined as the presence of any diabetic retinopathy.
Results: :
1227 patients were recruited into the study. Referral rate for diabetic retinopathy by the PCPs was 10.8% and 10.2% by the retina specialist. With the retina specialist as the gold standard the sensitivity of PCPs detecting diabetic retinopathy was 73.8% with 96.3% specificity. Cases of vitreous hemorrhage were detected with 100% sensitivity but none of the 6 cases with venous beading were correctly identified. Approximately 15% of the photos were of insufficient quality.
Conclusions: :
PCPs can be trained to read fundus images from a non–mydriatic camera and make appropriate referrals for diabetic retinopathy. Continued training of physicians in primary care and improved imaging with digital technology and mydriasis may make this a viable method for diabetic retinopathy screening.
Keywords: diabetic retinopathy • clinical (human) or epidemiologic studies: systems/equipment/techniques • imaging/image analysis: clinical