Purpose
To compare tele-ophthalmic screening for diabetic retinopathy (DR) results determined on a mobile device to results determined with a traditional desktop computer. We sought to evaluate the reliability of detecting DR on screening images using mobile devices, as they are potentially a more efficient method of communicating tele-ophthalmic screening results.
Methods
Over 8 months, 457 adult patients with diabetes underwent tele-ophthalmic screening for DR at 3 primary care clinics in Virginia. Trained clinical staff captured a single 45° macula-centered image of each eye for each patient. Images were transmitted from each clinic to ophthalmologists at the University of Virginia (UVA) using TigerText, a secure, HIPPA-compliant messaging application. The images were read by a UVA ophthalmologist on Samsung Galaxy Note 2 and 3 mobile phones, and a preliminary interpretation was given instantaneously. Up to one week after, images were read again on a desktop computer with 24-inch LCD and a final interpretation was given. Images were graded for image quality and the presence and severity of DR (mild, moderate, or severe NPDR; PDR). Agreement between image results was measured using percent exact agreement and Cohen’s kappa coefficient (k).
Results
435 of the total 457 patients (95%) were able to have both eyes imaged; 870 eyes were assessed in this study. Exact agreement between image quality assessment of preliminary and final interpretations was observed in 745 (86%) eyes (Table 1), giving a k of 0.76 ± 0.02. When comparing severity of DR on a mobile device vs. a desktop computer, exact agreement was observed in 824 (95%) eyes (Table 2) with a k of 0.89 ± 0.02. Both k values are considered excellent agreement.
Conclusions
Excellent agreement of both image quality interpretation and DR severity grades was observed between screening photos interpreted on a mobile phone and on a desktop computer. These results demonstrate that mobile phones can be used as an effective way to detect DR on screening images and communicate results. This allows ophthalmologists to read screening photos virtually anywhere at any time and to provide instantaneous preliminary results to patients who participate in screening programs.