Changing effect of the numerator–denominator bias in unlinked data on mortality differentials by education: evidence from Estonia, 2000–2015

Abstract

BackgroundThis study highlights changing disagreement between census and death record information in the reporting of the education of the deceased and shows how these reporting differences influence a range of mortality inequality estimates.MethodsThis study uses a census-linked mortality data set for Estonia for the periods 2000–2003 and 2012–2015. The information on the education of the deceased was drawn from both the censuses and death records. Range-type, Gini-type and regression-based measures were applied to measure absolute and relative mortality inequality according to the two types of data on the education of the deceased.ResultsThe study found a small effect of the numerator–denominator bias on unlinked mortality estimates for the period 2000–2003. The effect of this bias became sizeable in the period 2012–2015: in high education group, mortality was overestimated by 23–28%, whereas the middle education group showed notable underestimation of mortality. The same effect was small for the lowest education group. These biases led to substantial distortions in range-type inequality measures, whereas unlinked and linked Gini-type measures showed somewhat closer agreement.ConclusionsThe changing distortions in the unlinked estimates reported in this study warn that this type of evidence cannot be readily used for monitoring changes in mortality inequalities.

The dose-response relationship between socioeconomic deprivation and alcohol-attributable mortality risk—a systematic review and meta-analysis

“…This can lead to the so-called “numerator-denominator bias” where the information on SES for the denominator is based on self-report in census or survey data, whereas the SES of the deceased (numerator) is assessed at the time of death via proxy informants which is generally considered to generate lower-quality data [ 48 , 52 , 53 ]. While there is some evidence for the bias leading to an underestimation of the socioeconomic inequalities [ 54 ], other studies found no systematic or the opposite effect [ 55 , 56 ].…”

The Importance of Education for Middle- and Old-Age Mortality in Chile: Estimates From Panel Data Linked to Death Records

“…To do this, we benefit from a nationally representative longitudinal household survey linked with death records in Chile. On a cross-sectional analysis, the information provided on death certificates, reported by proxy informants, is of lower quality than those for the population from household surveys, leading to biased mortality rates for subgroups (Jasilionis & Leinsalu, 2021). In the longitudinal data we use in the current work, education reporting comes from a single source (the household survey), eliminating numerator-denominator bias.…”

Educational gradients in all-cause mortality in two cohorts in the Czech Republic during the early stage of the postcommunist transition

“…However, as most of available studies used unlinked routine data, the numerator-denominator bias makes them less reliable. 16 In this paper, we analysed data from two Czech cohorts using identical protocol established in 1985 and 1992 to assess changes in educational inequalities using individual-level (ie, linked) data. We hypothesised that the educational gradient would become steeper in the later cohort and this would be apparent in both men and women.…”

“…In most reports, the educational inequalities increased very quickly, over only a few years. However, as most of available studies used unlinked routine data, the numerator-denominator bias makes them less reliable 16…”

“…However, as most of available studies used unlinked routine data, the numerator-denominator bias makes them less reliable. 16 …”