A systematic global assessment of the completeness and quality of household death reporting in censuses and surveys since 2000

Abstract

Many censuses and surveys in low- and middle-income countries ask questions about deaths in the household to fill the evidence gap about mortality. This study undertakes the first published systematic assessment of the completeness and quality of these data. For 82 censuses from 56 countries and 26 surveys from 21 countries since 2000 we calculated completeness of household death reporting using deaths estimated by the United Nations World Population Prospects (UN WPP) and Global Burden of Disease (GBD) as the denominator. The median completeness of reported household deaths in censuses was 89% (inter-quartile range (IQR) 66–102%) and surveys 96% (IQR 80–124%). Completeness was similar for males and females and substantially lower where date of death was asked (census median 73%, IQR 53–91%) than not asked (census median 93%; IQR 74–110%); these differences remained after controlling for other covariates in a linear regression. The ratio of reported household to estimated deaths was higher in younger ages but age-invariant where date of death was asked. In conclusion, household death data in censuses and surveys have major completeness and quality issues. Where date of death was not asked, there appears to be considerable reporting of deaths that occurred outside of the reference period.