Thomas Salzmann and Christina Bohk (2008)
Testing Mortality Measured in the "Life Expectancy Survey" for Population Representativeness, Particular Consideration Being Given to Right and Interval Censoring with the Relative Survival Concept*
In: Zeitschrift für Bevölkerungswissenschaft, Vol. 33, 2/2008, p. 121-152, Wiesbaden: VS Verlag für Sozialwissenschaften, ISSN: 0340-2398, DOI: 10.1007/s12523-009-0008-3
Abstract: The Life Expectancy Survey (LES) is one of the few national longitudinal surveys in Germany. The follow-up period for the survival status of the sample population is a maximum of 14 years (1984-1998). The LES affords an opportunity for comprehensive mortality analysis, along with numerous items of personal information on socio-economical, demographic, physical and medical characteristics. However, the observed mortality of the sample population has not yet been adequately tested for representativeness. This becomes necessary if the results that have been obtained for the sample population are to be translated into the general population. The paper at hand aims at closing this gap. By applying the concept of ‘relative survival’, it will be shown that the mortality of the sample population does not differ from the mortality of the general population in Western Germany. To control for the heterogeneous composition of the sample population in the concept of ‘relative survival’, we calculate the expected number of survivors that would occur given the real survival of the general population (Western Germany) during the observation period. Given the two independent survival curves, both (observed and expected) are tested on equal mortality patterns by Hall-Wellner confidence bands. The LES comprises right and interval censored information of the survival status due to the sample design, and these need to be adequately accounted for in line with the ‘relative survival’ concept. To this end, it is necessary to strike a balance between the application of all the available information and the plausibility of the method. A computer program with R was written to ensure implementation.
* peer-reviewed article (original German title: Überprüfung der im Rahmen des "Lebenserwartungssurveys" gemessenen Sterblichkeit auf Bevölkerungsrepräsentativität unter Berücksichtigung rechts- und intervallzensierter Ereignisse mit dem Konzept "Relative Survival")