Abstract
Purpose :
Recent advancements in imaging technologies have improved the ability of telemedicine programs to detect diabetic retinopathy (DR). We performed a retrospective clinical chart review of the primary-care based Temple University Hospital (TUH) telemedicine program to assess the utility of non-mydriatic one-field fundus photography in generating quality screening images and ability to accurately identify DR.
Methods :
We completed a retrospective chart review of all images acquired from patients aged 18 years and older with Diabetes Mellitus and no prior dilated fundus examination who participated in the telemedicine retinal screening initiative through their TUH primary clinic from March 2016 to May 2017. We investigated how interpretable the photographs were as defined by the reader, how often the photographs generated a diagnosis of DR, and the severity of DR in the patient cohort.
Results :
A total of 1377 screening fundus photographs were acquired (689 patients). Of the 689 total photographs of the right eye, 356 (51.7%) were specified as “good quality,” 114 (16.5%) as “fair,” 208 (30.2%) as “poor,” and 11 (1.6%) were “unspecified” either due to ocular condition or due to an unknown reason. A similar distribution of quality was observed for the 688 photographs of the left eye: 336 (48.8%) “good,” 112 (16.3%) “fair,” 232 (33.7%) “poor,” 8 (1.2%) unspecified DR, and 1 image was unavailable. Of the total 1377 images collected, 928 (67.4%) were gradable and 449 (32.6%) were unable to be assessed due to “poor quality”. Of the 928 gradable fundus photos, 755 (81.4%) were read to have no DR, 56 (6.0%) unspecified DR, 78 (8.4%) mild DR, 16 (1.7%) moderate DR, 19 (2.0%) severe DR, and 4 (0.4%) proliferative DR. The mean and median length of time between the screening visit and the fundus photo interpretation was 55.4 and 23 days, respectively (range, 0 – 418 days).
Conclusions :
Many factors affect nonmydriatic fundus photograph quality including imaging technology, patient cooperation, and user photographer knowledge. This study identified a higher than expected rate of “poor” quality fundus photos which may lead to unnecessary specialist referrals, increasing healthcare costs. Additional inquiry is necessary to better optimize the process of image acquisition to improve fundus photograph quality and accurate DR detection.
This is a 2020 ARVO Annual Meeting abstract.