Comparison of Population-Based Census versus Birth History for the Estimation of Under-5 Mortality in Niger

The WHO guidelines on mass distribution of azithromycin for child survival recommend monitoring of mortality to evaluate effectiveness. Trials that contributed evidence to these guidelines used a population-based census to monitor vital status, requiring census workers to visit each household biannually (twice yearly). Birth history is an alternative to the census approach that may be more feasible because it decreases the time and labor needed for mortality monitoring. This study aimed to compare the population-based census (reference standard) and birth history (index test) approaches to estimating mortality among children 1 to 59 months old using data from the Macrolides Oraux pour Réduire les Décès avec un Oeil sur la Résistance (MORDOR) trial. Sixteen communities that received 5 years of biannual census in the MORDOR trial were selected randomly also to receive birth history surveys. The census approach recorded more participants and households than birth history, with correlations more than 0.94 for each. The correlation between number of deaths in each community was 0.84 (95% CI, 0.59-0.94). A comparison of the mortality incidence rate estimated from the census against the under-5 mortality rate estimated from the birth history resulted in a correlation of 0.60 (95% CI, 0.15-0.84). Of the 47% of children who were linked individually to compare vital status from each method, the death status of children had a sensitivity of 80% (95% CI, 73-89) and a specificity of 98% (95% CI, 98-99), comparing birth history to census. Overall birth histories were found to be a reasonable alternative to biannual census for tracking vital status.