There are very few population-based data on childhood blindness in the Horn of Africa. In Sekoru district, Ethiopia, we performed a whole-of-population cross-sectional survey using the key informant method to determine the magnitude and causes of childhood blindness (presenting visual acuity <3/60 in the better eye) and severe visual impairment (SVI; presenting visual acuity <6/60 but ≥3/60 in the better eye). Forty-two key informants were trained to identify blind/SVI children aged <16 years. Identified children were examined by an ophthalmologist. Causes of blindness/SVI were established by clinical examination. Key informants visited 94% of 22,666 households, identifying 112 children, of whom 36 (32%) were confirmed to be blind/SVI by the ophthalmologist. Twenty (56%) of the 36 were male. The mean age was 10.7 years. The district prevalence of childhood blindness/SVI was 0.062% (95% CI 0.042-0.082%). Lens-related abnormalities (mainly congenital cataract) were the commonest causes. Avoidable causes of blindness accounted for 89% of cases. Paediatric ophthalmic services are required here, as elsewhere in rural Ethiopia. Assessment by key informants has an acceptable positive predictive value in this setting. This method warrants further evaluation as a tool for measuring the prevalence of other rare conditions in rural and remote populations.