OBJECTIVES: The study was conducted to determine whether trained key informants (KI) could identify children with impairments. DESIGN: Trained KI identified children with defined impairments/epilepsy who were then examined by a medical team at a nearby assessment centre (Key Informant Methodology: KIM). A population-based household randomised sample survey was also conducted for comparing the prevalence estimates. SETTING: Three districts in North Bangladesh. PARTICIPANTS: Study population of approximately 258 000 children aged 0-<18 years, within which 3910 children were identified by KI, 94.8% of whom attended assessment camps. In the household survey, 8120 children were examined, of whom 119 were identified with an impairment/epilepsy. MAIN OUTCOME MEASURES: Prevalence estimates of severe visual impairment (SVI), moderate/severe hearing impairment (HI), substantial physical impairment (PI) and epilepsy. RESULTS: Overall prevalence estimates of impairments, including presumed HI, showed significant differences comparing KIM (9.0/1000 (95% CI 8.7 to 9.4)) with the household survey (14.7/1000 (95% CI 12.0 to 17.3)). Good agreement was observed for SVI (KIM 0.7/1000 children: survey 0.5/1000), PI (KIM 6.2/1000 children: survey 8.0/1000) and epilepsy (KIM 1.5/1000 children: survey 2.2/1000). Prevalence estimates for HI were much lower using KIM (2/1000) compared to the survey (6.4/1000). Excluding HI, overall prevalence estimates were similar (KIM: 7.5/1000 children (95% CI 7.2 to 7.8) survey: 8.4/1000 (95% CI 6.4 to 10.4)). CONCLUSIONS: KIM offers a low cost and relatively rapid way to identify children with SVI, PI and epilepsy in Bangladesh. HI is underestimated using KIM, requiring further research.