Papers

Commentary: Appropriateness of deprivation indices must be ensured

BMJ 1994; 309 doi: https://doi.org/10.1136/bmj.309.6967.1479 (Published 03 December 1994) Cite this as: BMJ 1994;309:1479
  1. Mel Bartley,
  2. David Blane
  1. Nuffield College, Oxford OX1 1NF research officer in public health policy. Academic Department of Psychiatry, Charing Cross and Westminster Medical School, London W6 8RP lecturer in medical sociology.
  1. Correspondence to: Dr Bartley.

    Since 1971, census information has been used to identify localities where a high proportion of households are living under adverse social and economic conditions. The initial intention was that such areas should be provided with extra resources to improve the quality of life of individual people and their families.

    Underlying this exercise was the assumption that improving local facilities (by upgrading housing estates or setting up educational priority areas) could improve residents' opportunity and quality of life without changing individual circumstances such as low wages or unemployment.1 The items from the 1971 census used for the original Department of Environment deprivation index therefore included measures of housing quality and proxy measures of income such as employment status. Other variables such as the numbers of children aged up to 14 years would not in themselves constitute “disadvantage”, but they were included as a measure of the need for services. In general, area deprivation index scores measure the proportion of households within a defined small geographical unit with a combination of circumstances indicating low living standards or a high need for services, or both.

    Several deprivation indices have been developed since the original 1971 Department of Environment measure. The underprivileged area index developed by Jarman was intended as an indicator of the demand for primary medical care. Accordingly, it included items thought to indicate the likelihood of high demand such as the proportion of elderly people living alone and the number of children under 5. Two other well known measures developed by Carstairs and colleagues and by Townsend and colleagues were guided by an interest in the effects of living standards on health. They contain items that might be regarded as good indicators of the amount and stability of household income: housing tenure, car ownership, unemployment, and social class.

    Deprivation indices therefore need to be understood and evaluated in terms of (a) the purpose for which they are being used and (b) the validity of the assumptions about social and economic life that they embody. This is true of the decisions to include or exclude different items. It is also true of the different types of statistical procedures used to combine items together into a summary score. Items are often converted into a standardised Z score, which expresses each one in terms of its mean value in the population and its standard error. If this were not done then items with longer scales would have more weight than those with shorter scales in the overall score. For example, the number of children in a household could vary from, say, 0 to 10, while the number of cars available would lie between 0 and 3. Simply adding these together would give children more weight than cars. Standardisation is intended to avoid this problem, though this is easier for some items than for others. On the other hand, some components may deliberately be weighted more heavily if they are thought to be more important—for example, in the Jarman index “pensioners living alone” is weighted highest, whereas in the Department of the Environment's 1981 index unemployment has double the weight of other variables.

    Information on the relation between mortality and deprivation in individual people can be obtained by linking death certificates to census information. This is what the longitudinal study of the Office of Population Censuses and Surveys makes possible—though only for 1% of the population. As a growing number of people find themselves without a single stable occupation, social class is becoming less satisfactory as a measure of living standards and lifestyle. The use of additional measures of disadvantage, both in the form of linked individual data and in the form of small area statistics, will therefore become increasingly valuable in estimating social variations in health.

    References

    1. 1.
    View Abstract