Inter-observer Agreement in Fundus Photography for Diabetic Retinopathy Screening in Primary Health Care

Abstract

Background: To assess the agreement between a group of trained non-physician ophthalmic careproviders as primary grader, and vitreoretina specialist as secondary grader, in the interpretation ofsingle-field 45o, non-mydriatic digital fundus image for diabetic retinopathy (DR) screening in primaryhealth care.Methods: This is an analytic, retrospective study. A total of 117 fundus images of diabetic patientswere graded by primary grader, followed by secondary grader using the modified National ScreeningCommittee (NSC) as a guideline. Kappa statistics and Intraclass Correlation (ICC) were used for thereliability assessment of photograph quality, DR presence and classification, presence of maculopathyand photocoagulation, severity level, and presence of vision-threatening diabetic retinopathy (VTDR).Diagnostic values of fundus image interpretation by primary grader in detecting DR using secondarygrader as the reference standard were also assessed.Results: Agreement for DR classification and its severity level was good (ICC 0.654; 0.642). Fairagreement was found for presence of maculopathy and VTDR (κ 0.368; ICC 0.595). Slight agreementwas found for photograph quality (κ 0.170). The percentage agreement in identifying DR was 29.9%.Sensitivity and specificity of fundus image interpretation by primary grader in DR and VTDR detectionwere 52.24% and 71.43%, 54.55% and 95.12%, consecutively.Conclusion: The reliability of primary grader for interpreting single-field fundus photography in DRscreening has not reached the ideal result in this study (κ more than 0.6 in all variables). Acomprehensive training with continuing education is essential for non-physician personnel, to ensurethe reliability of DR screening program using single-field fundus photography.Keywords: diabetic retinopathy screening, interobserver agreement, kappa value, intraclass correlation.