The influence of individual and area disadvantage over the life course on mortality risk 1996-2001: a multilevel analysis

Chris White, Alison Whitworth and Anne-Marie Jones, Office for National Statistics, and Dick Wiggins, City University London

[Project number 20066]

Spatial variations in health represent an important dimension of health inequality in Britain, having implications for resource allocation at the administrative level. Dorling has undertaken a detailed investigation into regional differences in mortality and found sharp rises in relative risk. A similar pattern was observed in the prevalence of limiting long-term illness, carried out in the wake of the 1991 census.

Geographical differences in health and mortality could be simply due to the personal characteristics, behaviours and exposure histories of the individuals that make up the area population (i.e. compositional effects). Alternatively, they could be due to the ecological and environmental fabric of the area itself (contextual effects). A crucial step in clarifying the balance of these effects is to determine whether the scale of variation between areas goes beyond what would be expected on the basis of individual characteristics alone. Such an understanding can only be achieved through the application of models that take account of the hierarchical nature of the data.

This project aims to demonstrate the use of multilevel modelling techniques in the analysis of longitudinal survey and vital events data, using the statistical packages MLWin and Stata, and contribute methodologically and substantively to the literature on geographical health inequalities.

The objectives of the project are to:

1. Build up multilevel modelling skills capacity within the Health Variations team;
2. Model the effect of individual and county district level disadvantage across the life course on mortality risk 1996-2001, using multilevel logistic regression.
3. Provide a discussion of the advantages and disadvantages of multilevel modelling in the analysis of socio-demographic and mortality data.
4. Assess conformity with multilevel models constructed to measure the relationship between individual and area disadvantage and limiting long-term illness.
5. Assess software performance.
6. Model the effect of individual and ward level disadvantage across the life course on risk of LLTI in 2001, using multilevel logistic regression.
7. Model the effect of individual and ward level disadvantage across the life course on self-reported health in 2001.
8. Explore application of multivariate multilevel models using LLTI and self-reported heath at 2001.

Objectives 6, 7 and 8 will not be tackled until objectives 1-5 have been achieved.

The LS is the only data source available to provide socio-demographic and event histories spanning 30 years. The ability to measure socioeconomic and spatial transitions allows a more sensitive profile of individual disadvantage to be constructed across time, with greater predictive power than assigning a measure of social standing at a point in time. The aims and objectives of this project are inconceivable in the absence of a representative longitudinal data source such as the ONS LS.

Top of page