Estimates that cannot fail visibly: a current debate on measuring child mortality as household surveys contract

Global health has an argument on its hands that it has not yet had openly. For most of the countries that carry the burden of child deaths, the published under-five mortality rate is not a measurement. It is a model output built largely from birth histories collected in household surveys, because only about a tenth of deaths are registered in the World Health Organization African region. That input stream contracted sharply in 2025, when United States funding ended for the survey programme that had supplied about half of the child mortality data points for sub-Saharan Africa; interim funding now sustains only selected services. Three positions can be discerned in response, two of them largely implicit: that the estimation model will absorb the loss, that the contraction is a reason to accelerate civil registration, and that the survey stream must be restored. I argue that none can be settled on present information, because the evidence behind each estimate is published apart from the estimate and from the products that use it, so the figures look the same whether or not new evidence has arrived. I propose three steps in order of priority: print the age, number and type of the data sources beside every country estimate; direct scarce survey funding first to countries whose evidence is oldest and whose burden is largest; and make sample-based mortality registration a nationally owned and financed function of the state, as the bridge to complete civil registration.