Abstract
Purpose :
Retinopathy of prematurity (ROP) examinations have traditionally been performed using binocular indirect ophthalmoscopy (BIO). Telemedicine using digital fundus imaging (DFI) is an emerging strategy that may improve the delivery of ophthalmic care. Traditional telemedicine studies have assessed accuracy of telemedicine compared to BIO as a gold standard. However, there is little published literature that directly compares the accuracy of DFI vs. BIO using a consensus reference standard.
Methods :
We developed a web-based platform to allow storage and interpretation of wide-angle fundus photos of infants undergoing ROP examinations. As part of a prospective cohort study, infants who met screening criteria at 7 participating centers were examined directly via BIO by experienced clinicians and indirectly via DFI grading by 3 trained graders independently. These examination findings were combined into a consensus reference standard diagnosis for each imaging session for zone, stage, plus and category. We calculated the agreement of both DFI and BIO with the gold standard using percent agreement and weighted kappa statistics.
Results :
A total of 1553 eye examinations were classified using both DFI and BIO. On average, BIO was slightly more accurate than DFI in identifying zone (91 vs 89%, p<0.01), stage (88 vs 75%, p<0.01), and overall disease category (84 vs 77%, p<0.01), whereas DFI was slightly more accurate than BIO in diagnosing plus disease (92 vs 88%, p<0.01). There was significant inter-grader variability in both groups (Table 1). BIO was more sensitive in identifying stage 3 disease (85 vs 73%, p<0.01; n=136), but not zone I (78 vs 78%; n=165), plus disease (74 vs 79%, p=0.41; n=50), or any referral-warranted ROP (RWROP, defined as Stage 3, Zone I, or Plus disease; 84 vs 79%, p=0.10; n=251).
Conclusions :
This study compared the accuracy and sensitivity of DFI to BIO in ROP examination. Key findings were: 1) Accuracy of ROP diagnosis varies between graders, both using DFI and BIO; 2) There was no significant difference in the sensitivity of BIO and DFI to detect RWROP; 3) BIO was slightly more accurate in identifying zone, stage, and category of ROP, while DFI was slightly more accurate in identifying plus disease. These results show that neither DFI nor BIO is completely accurate, and that inter-grader diagnostic variability is high using current criteria regardless of exam method.
This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.