The feasibility and reliability of a remote teleophthalmology reading center for improving access to care

Abstract

Purpose : Early detection is key in the management and long-term outcomes of numerous ophthalmic disorders, such as glaucoma and diabetic retinopathy, many of which can present insidiously with symptoms unnoticed prior to advanced disease. Teleophthalmologic screening presents a major opportunity to reach populations that traditionally lack access to routine care. The goal of this study is to evaluate the feasibility and reliability of a remote international reading center using ocular imaging.

Methods : 50 patients (mean age 58.5 years, 48/50 female) presented for comprehensive eye exams from an ophthalmologist (Reference Standard, or “Ref”) in Venezuela, and fundus photography and optical coherence tomography (OCT) images were sent to two outside ophthalmologists (Reader 1 and Reader 2, or “R1” and “R2”) blinded to the in-person exam. Features assessed included cup/disc ratio (CDR) on fundus photos and retinal nerve fiber layer thinning on OCT. Fundus findings were compared for reliability and accuracy, with the in-person exam treated as the best available assessment; while OCT findings were compared between R1 and R2 for reliability.

Results : While agreement on exact CDR between the readers and Ref was relatively low (17.6-29.4%), CDR findings were fairly well correlated (r 0.614-0.649, p<0.01), with moderate reliability among all evaluators (intraclass correlation 0.665 for right eye, 0.698 for left eye; Table 1). Agreement was higher for CDR estimate on fundus photo (r 0.865-0.848, p<0.01). When assessing for CDR > 0.5, accuracy improved (76.5-83.3% agreement) and moderate interrater reliability was observed. On remote OCT assessment, percent agreement between readers were 72% (right eye) and 64% (left eye), with fair reliability (Cohen’s kappa coefficient 0.323 for right eye, 0.236 for left eye; Table 2).

Conclusions : Although CDR agreement rates between in-person fundoscopic exam and remote fundus photography reading were low, findings were well correlated and improved when assessing for the clinically relevant distinction of CDR > 0.5, suggesting that remote screening may be reliable and accurate in assessing for abnormal CDR. Lower reliability of remote OCT assessment was driven by a high disagreement rate between these readers, reflecting the need for consensus, standardization, and further evaluation with a larger pool of readers in a sample better representing the general population.

This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.