Validation of smartphone-based screening for retinopathy of prematurity in a low-resource setting

Abstract

To study the validity of smartphone-based screening to detect moderate-to-severe levels of retinopathy of prematurity (ROP) in a low-resource setting. In this observational validation study, all new patients at a single center who met screening criteria for ROP (birth weight of ≤1700g and gestational age of ≤35weeks) were examined by a pediatric ophthalmologist using indirect ophthalmoscopy. At the first ROP examination, a trained photographer captured fundus images using a smartphone fitted on an adapter, the Paxos Scope. The photographs were graded by two retina specialists masked to results of the gold standard examination. A total of 100 children (200 eyes) were included. ROP was detected in 27.4% by indirect ophthalmoscopy, 16% of whom had moderate-to-severe ROP, defined as stage 3 in zone I or II, stage 2 in zone I or II, or pre-plus or plus disease. Sensitivity of fundus images reviewed by grader 1 against the gold standard in detecting moderate-to-severe ROP was 87.5%, and specificity was 82.1%. The positive predictive value was 48.3 %; the negative predictive value, 97.2%. The sensitivity of fundus images reviewed by grader 2 was 87.5%; the specificity, 81.6%. Positive predictive value was 47.5%; negative predictive value, 97.2%. The measured κ coefficient for intergrader agreement was 0.94. Our study is one of the few studies on smartphone-based telescreening for ROP. The results suggest that smartphones may have potential as a screening tool for ROP in low-resource settings.