Telemedicine is increasingly used for disease monitoring and management of chronic medical and mental disorders, but also has screening and diagnostic applications such as teleradiology and teledermatology. Indeed, "therapy at a distance" will complement office-based care in the 21st century. Another screening application is teleophthalmology, in which digital photography with telemedicine links has proven cost-effective for retinal disorders, including diabetic retinopathy and retinopathy of prematurity. 1,2 In a recent study in JAMA Ophthalmology, Chasan et al 3 complemented prior cost-effectiveness analyses of teleretinal screening programs by examining both the accuracy of diabetic teleretinal screening as well as the subsequent eye care use and resources required in such a program. Regarding accuracy, screening tests that have good sensitivity and specificity are likely to be more cost-effective. Regarding resource use, implementation of screening generates more referrals, testing, and procedures. The cost-effectiveness of any screening program is therefore contingent on reasonable test accuracy coupled with a sufficient supply of resources to meet the increased demand.Chasan et al 3 conducted the study in community-based clinics of a single Veterans Affairs (VA) medical center. Of 1935 patients who underwent diabetic retinal screening, 465 (24%) had an abnormal finding on the retinal photograph and were referred to the VA eye clinic, of whom 326 were seen by an eye specialist and had a confirmatory diagnosis. The most common reasons for referral were nonmacular diabetic retinopathy (43.2%), nerve-related disease (30.8%), lens or media opacity (19.1%), age-related macular degeneration (12.9%), and diabetic macular edema (5.6%). The percentage of agreement between retinal screening and the ophthalmic examination for all diagnoses was 90.4%, with a sensitivity of 73.6%. Sensitivity may have been somewhat inflated for 2 reasons. First, only patients with abnormal retinal photographs were referred, thus greatly increasing the number of cases with disease and minimizing the number of cases without disease. This verification bias (also known as "workup bias" or "referral bias") is a type of measurement bias in which the results of a diagnostic test affect whether the gold standard procedure is used to verify the test result. Second, it is not clear that the clinician performing the ophthalmic examina- EXPOSURES Clinical medical records were reviewed for a 2-year period after patients were referred from teleretinal screening. The following information was collected for analysis: patient demographics, referral and confirmatory diagnoses, ophthalmology clinic visits, diagnostic procedures, surgical procedures, medications, and spectacle prescriptions.
MAIN OUTCOMES AND MEASURESThe accuracy between referring and final diagnoses and the eye care resources that were used in the care of referred patients.
RESULTSThe most common referral diagnoses were nonmacular diabetic retinopathy (43.2%), nerve-related disease (30.8%...