Validity of mortality follow-up in SHARE

Anne Schulz, Deutsches Zentrum für Neurodegenerative Erkrankungen e.V. and Universität Rostock
Gabriele Doblhammer-Reiter, University of Rostock

This paper studies the validity of the first mortality-follow-up in SHARE using longitudinal data from the first two waves of SHARE. Analyses are performed for all countries together (Austria, Germany, Netherlands, Sweden, Spain, Italy, France, Denmark, and Belgium) as well as country-specific. SHARE’s first wave is a selection of “healthy” persons since almost all respondents are sampled from private households. Furthermore, the current dataset does not provide information on date of death for all deaths between the two waves. Therefore SHARE’s mortality follow-up is assumed to underestimate real mortality. Age-specific death rates in SHARE are calculated by using survival analysis methods. In a second step observed age-specific death rates are weighted by a constant factor to correct for missing deaths. Those death rates are compared with expected values based on data from the Human Mortality Database (HMD). Additionally and in order to explain differences between SHARE and HMD, a meta-regression is carried out to predict life expectancy in SHARE-countries. SHARE underestimates mortality systematically. Observed death rates stay within confidence intervals of expected death rates until the age of 70 years. Underestimation increases with age. Meta-regression cannot explain differences between life expectancy in SHARE and HMD when controlling for attrition, non-response and composition of sampled population with regard to health status. However, the proportion of the observed to the expected number of deaths seems to be the driving factor for the differences between life expectancy in SHARE and HMD. The current version of the mortality follow-up does underestimate real mortality because of missing deaths and missing institutionalized population in the sample.