Philadelphia Telemedicine Glaucoma Detection and Follow-up Study: Interphysician Agreement for Screening Images

Abstract

Purpose : The Philadelphia Telemedicine Glaucoma Detection and Follow-up Study is an innovative, primary care-based eye screening model to detect, treat, and manage glaucoma in high-risk individuals in urban primary care practices (PCP) and federally qualified health centers (FQHC). The purpose of this study was to evaluate the interphysician agreement in the remote interpretation of fundus photographs and clinical data collected during the initial eye screening visit.

Methods : Our targeted, at-risk population included African-Americans, Hispanics and Asians over age 40; adults of any ethnicity over 65 or over age 40 with a family history of glaucoma or diabetes. Participants underwent a free telemedicine eye screening (Visit 1) which included non-mydriatic fundus photography, medical/ocular history, and intraocular pressure measurements in 7 PCPs and 4 FHQCs. A total of 906 subjects’ images were graded as normal, abnormal, or unreadable. 100 subjects were randomly selected for this study sub-analysis. Three glaucoma specialists, who were blinded to the initial grading and follow-up recommendation, participated in this study. Each glaucoma specialist independently graded each eye based on presence of specific glaucomatous optic nerve findings (disc cupping or asymmetry, neuroretinal rim thinning, peripapillary atrophy, nerve fiber layer loss, disc hemorrhage). Based on the photographs of each eye, and clinical information, each grader then selected a suitable follow up-recommendation. The primary outcome of this study was the agreement of the three readers on the follow-up recommendation per subject. Additionally, the number of raters agreeing on follow-up recommendation and glaucoma findings was computed. A consensus diagnosis was defined if at least two of the three raters agreed.

Results : Results of 100 subjects (mean age 57.0 ± 10.2 years): inter-rater reliability for follow-up recommendations was moderate [Fleiss’ kappa = 0.63; 95% CI: (0.56, 0.70)]. All readers agreed on follow-up recommendation for 64% of subjects. Consensus was reached in 96% of subjects, of which 39 were normal, 41 abnormal, and 16 unreadable. With respect to glaucomatous findings, a consensus diagnosis was achieved for all subjects, with 61% agreement among all raters.

Conclusions : There was substantial agreement in interpretation of integrated data and images obtained through a telemedicine screening program.

This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.