Norman, P. & Riva, M. (2012) Population, Space and Place, 18(5), 483-502. [ONS LS]

Other information:
Abstract:

There is a need in health research to identify whether inequalities are increasing or improving between different geographical areas. Both cross-sectional time-series and longitudinal/cohort studies contribute to our knowledge, with the Office for National Statistics Longitudinal Study (LS) for England and Wales being a major resource. However, any research into geographical change over time can be hampered by boundary change or when the geographical definition for which data are available is not the geography relevant to an analysis.

We develop a method using population-weighted centroids of estimating an LS member's location at a previous time point and then link this to a small-area geography, the 2001 Census Output Areas. This is not so that analyses can be carried out at this scale but so that records can be linked to larger geographies or area classifications. A time-series or longitudinal analysis can then be carried out and geographical trends observed. In terms of reliability, we find that accuracy improves with increasing size of geographical units and when area typologies are used.

In example analyses using a geodemographic classification attached to LS members' records, we find that in a time series of cross-sections, mortality improves across all area types but not to the same extent. A longitudinal analysis indicates that changes in the area types in which people were living lead to steeper health gradients than if people had stayed living in the same type of area. Differences, though, are small, suggesting that, in the main, there is little mobility between area types.

We recommend that longitudinal and cohort studies retain the postcode of each member's address so that ongoing linkages can be made when administrative boundary changes occur and for relevance to application relevant geographies. Our method can be used to enhance previous records and thereby maximise previous investment in the collection of data.

Available online: Population, Space and Place,
Output from project: 0301155