Review of life expectancy model on the LS

Brian Johnson, Office for National Statistics

[Project number 20090]

Historically the LS has been the primary data source for the calculation of life expectancy by social class, since it allows the attribution of a classification to individual members which can then be used to study differences in mortality prospectively. This avoids the numerator-denominator bias which affects measures using death registrations and population at the same or similar time points.

Published analyses by Hattersley (1999) and Donkin et al (2002) established a methodology for estimating life expectancy by social class using the LS, which was recently adapted to produce updated estimates for the Pensions Commission.

This work has highlighted two main issues requiring review, to inform the methodology for further work in this area using the LS. One is the attribution of social class which is the subject of a separate LS project. The other is the calculation of denominators or person years at risk and estimation of numbers "lost to follow up" as a result of unrecorded emigration. Non-enumeration at a census and non-matching owing to data discrepancies between different sources reduce the value of the use of census as a means of active follow-up. This has become a more urgent consideration in mortality studies because the level of underenumeration in the 2001 Census was around 6%. Rules for inclusion and exclusion of members from the study and for censoring of contributions to person years at risk must be derived in a way that minimises bias in prospective follow-up.

This study will test the validity of the existing algorithms which apply to the estimation of numbers lost to follow up from the study over four census points and three decades. It will aim to provide improved rules for calculating population at risk in studies requiring prospective follow-up using the LS.

The LS data will provide the basis for the work, which aims to provide a transparent ONS methodology for future prospective studies of which there are several planned.