Abstract
Purpose :
To characterize discrepancies in retinopathy of prematurity (ROP) status between results of clinical examinations and gradings of retinal images from the “Telemedicine approaches to evaluating acute-phase ROP - e-ROP” Study.
Methods :
A secondary analysis of an observational cohort study of infants with birth weights of <1251g. Consensus review by four ROP experts of discrepancies [in stage 3, zone 1 ROP or plus disease which are components of referral-warranted ROP (RW-ROP)] between results of clinical examination and remote evaluation of images by trained readers. For discrepancy categories with >100 cases, 40 cases were randomly selected. A total of 188 image sets were reviewed.
Results :
Among the 5350 image set and clinical exam pairs where RW-ROP status could be determined for both, there were 161 discrepant cases where image grading did not detect RW-ROP noted on clinical examinations (G-/E+) and 854 discrepant cases where grading noted RW-ROP when the examination did not (G+/E-).
Among G-/E+ cases, 56.3% (18/32) of consensus reviews agreed with clinical exam that ROP was present in zone I and 45.0% (18/40) agreed stage 3 ROP was present. Consensus reviewers noted trained readers were less likely to detect stage 3 ROP than the clinicians, as, when the clinician noted stage 3 ROP, only stage 2 ROP was noted on review in 32.5% of 40 image sets evaluated and 22.5% no ROP was noted. Review noted only 5% (1/20) agreement with clinical exam on presence of plus disease in G-/E+ cases.
Among G+/E- cases , 90.0% (36/40) of consensus reviews agreed with trained readers that zone I ROP was present when the clinical exam did not, 57.5% (23/40) agreed with stage 3 ROP, and 25% (4/16) for plus disease.
Conclusions :
This report highlights strengths and limitations of both remote evaluation of fundus images and bedside clinical examination of infants at risk for ROP. Despite extensive training, trained readers were less likely to detect stage 3 ROP than the clinicians. However, for more quantitative evaluations including zone of ROP and plus disease, reader gradings were superior to clinical assessments. These findings are of importance as ROP telemedicine becomes more widespread and highlight the need for more quantitative approaches.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.