BACKGROUND/OBJECTIVES: Late presentation of congenital cataract in the developing world has led to poor outcomes such that cataract is the leading cause of childhood blindness. Our hypothesis was that, sensitivity of red-reflex testing is greater than sensitivity of torchlight examination. We aimed to compare sensitivity of new red reflex screening tools and assess the feasibility of Arclight red reflex screening in the community. SUBJECT/METHODS: We compared the diagnostic accuracy of four different screening tools for cataract and retinoblastoma performed by ophthalmic nurses, using a clinic based enriched sample of 41 positives and 60 negatives. We then conducted a separate feasibility study, training non-specialist community nurses. Following the training, community nurses examined 2827 children <5 years with Arclight who were attending their clinics for growth monitoring and immunisation. FINDINGS: Diagnostic accuracy study: estimated sensitivities were 97.6% for Catcam, 92.7% for Arclight, 90.2% for PEEK retina and 7.3% for torchlight. Estimated specificities were above 90% for Catcam, Arclight and torchlight and 87% for PEEK retina. Feasibility study: twenty-four out of 2728 children screened failed community screening, seven were true positive (six cataract, one retinoblastoma). Prevalence of bilateral cataract was 1.5/1000 (95% CI: 0.40-3.75 per 1000). CONCLUSIONS: Arclight and CatCam have higher sensitivity than torchlight, are easy to learn and use by primary health care nurses. Red reflex testing should be recommended in the WHO guidelines instead of torchlight examination to help early detection of potential blinding causes including congenital cataract and retinoblastoma.