Measurement of Inequalities
Different measures can give information about different aspects of inequalities. Some measures concentrate on the extremes of deprivation, whilst others include inequalities across the scale – taking into account the whole population. Absolute and relative measures can give quite different interpretations of inequalities.
Information about different measures of inequality and their calculation was based on work done by the Scottish Public Health Observatory, available at: http://www.scotpho.org.uk/home/Publications/scotphoreports/pub_measuringinequalities.asp
The approach adopted in this report uses a combination of measures, with the aim of giving a fuller understanding of the inequalities concerned. Detailed descriptions of the measures are given below, and are also explained in the following PowerPoint slide show which includes a worked example:
Absolute range: How big is the gap?
This measure describes the absolute difference between the extremes of deprivation. It has the advantage that it is intuitive and straightforward to explain, but the disadvantage that, because it focuses only on the extremes of deprivation, it does not take account of patterns of inequalities observed across the intermediate groups.
Relative Index of Inequalities (RII): How steep is the inequalities gradient?
The RII describes the gradient of health observed across the deprivation scale, relative to the mean. This ensures that RII is meaningfully comparable between indicators. Comparisons between indicators should be made with some caution, in particular where indicators are based on different measures (for example, relative inequality in an age-standardised rate compared to relative inequality in scores in a survey question).
RII has the advantage that it is based on data about the whole population, rather than just the extremes of deprivation. However, it does assume a linear relationship between the health indicator and deprivation.
The value of RII tells you the magnitude of inequality in relation to the mean, in the linear model describing the data: absolute range = mean * RII
Interpretation of RII
- An RII of zero would indicate no inequality, although there may be variations between deprivaiton deciles.
- If the indicator follows a linear relationship with deprivation, then the natural upper bound on RII is +/- 2.
- RII may exceed 2, in particular in cases where the relationship with deprivation is not linear.
- A useful interpretation of RII is to consider the average change in the health outcome with each increasing deprivation decile.
- For example, say for a mortality indicator the mean mortality rate is 500 per 100,000 and RII = 1.50. This could be interpreted as, with each increasing decile, mortality increases by 1.50 * 500 / 10. i.e. the mortality rate increases by approximately 75 per 100,000 with each increasing deprivation decile.
While there is evidence of non-linearity for some indicators in some years, the expert group who designed the indicators concluded that the linear methodology should be retained due to the complexity of non-linear methods, and the need of consistent reporting and general understanding.
Scale: How big is the problem?
The aim of this measure is to give insight into the underlying scale of the problem and to put it in context, for example by presenting numbers involved and past trends at Scotland level.
The Short Life Technical Advisory Group addressed the precise way in which deprivation should be defined for this work. The group agreed that the ideal would be to use individually linked records of health and socio-economic indicators, but acknowledged that these are not yet available. The preferred interim approach was to use the latest available versions of the Scottish Index of Multiple Deprivation (SIMD) income and employment domains. The reasoning behind this being that income / poverty / employment are felt to be the best indicators of deprivation for health inequalities analysis and because the possibility of being able to update these domains on a regular, perhaps annual basis, is being investigated.
In order to combine the SIMD income and employment domains, each domain was exponentially transformed to reduce averaging effects. Exponential transformation gives greater weighting to the most deprived ranking, so combining a datazone ranked most deprived with a datazone ranked least deprived would give a combined ranking skewed towards the deprived end of the scale. This is the method used to create the SIMD.
The income and employment domains have been given equal weighting when combined in the Income-Employment Index.
In line with the recommendations of the Short Life Technical Advisory Group, the Income-Employment Index deciles are population based. Datazone based deciles are produced by ranking the 6,505 datazones in Scotland according to their deprivation score and then dividing them into deciles based on number of datazones (so that those datazones ranked from 1 to 651 are in decile 1 and so on). Population basing the deciles uses the same approach but also takes into account the population sizes involved. The 6,505 datazones are ranked according to their deprivation score alongside a cumulative total of datazone populations. The cut-off for decile 1 is the point at which 10% of the population has been included, rounded to the nearest whole datazone. Population basing the deciles ensures that they contain equally sized populations, which is the best proxy to individual level indicators of deprivation available when using an area based measure. Equally sized populations in the deciles are also important for the types of inequalities analyses presented in this report.