Growing up in Scotland: health inequalities in the early years

CHAPTER 2 MEASURING HEALTH INEQUALITIES IN GUS

2.1 Measures used in this report

This report is not designed to be exhaustive. Instead, it provides a snapshot of the potential that the GUS data can contribute to this area. The measures that have been chosen for exploration are to a large extent pragmatic as the space available to explore this topic is limited. However, the choice has been guided by factors of key policy relevance in Scotland.

2.1.1 Health outcomes and risk factors

As noted in the previous chapter, health inequalities can be fairly broadly defined to include differences in: specific health outcomes (such as low birthweight, obesity, long-term conditions, accidents); health related risk factors that impact directly on children (such as poor diet, low levels of physical activity, exposure to tobacco smoke); as well as exposure to wider risks from parental/familial behaviours and environmental circumstances (maternal depression and/or poor physical health, alcohol consumption, limited interaction, limited cognitive stimulation, poor housing, lack of access to greenspace). The longitudinal design of GUS means that for some of these measures it is possible to investigate repeated exposure to risk factors and experience of poor outcomes as well as at single points in time.

One of the most common measures of health status in analyses of adult health is self-assessed general health which has been shown to be associated with long-term morbidity and mortality, as well as being an important marker of health inequality in adulthood (Idler and Benyamini, 1997; Measuring Inequalities in Health Working Group, 2003). However, there are actually very few children within the GUS sample whose health is said to be bad or very bad by their parents (just 6% at 10 months and 7% at 46 months). In part this will be because the main determinant of poor general health is old age while the more extreme negative health outcomes that children can experience have a low prevalence in the population in the first place. There is also a high likelihood that sample attrition and differential response at the first wave will have excluded children with very serious health conditions, for example children who spend a significant amount of time in hospital or those with terminal conditions. Finally, as highlighted in Equally Well, looked after children are at significant risk of negative health outcomes and the GUS sample cannot capture the experiences of this group of children. ³

The following list sets out the measures explored in this report. In line with the WHO and Equally Well definitions of health and health inequalities, the measures span a range of outcomes and risk factors. The list is not exhaustive and a more in depth analysis of this topic could consider a wider range of factors, for example more of the developmental milestones. However, they reflect a broad spread of factors of policy concern and are all likely to be of interest as explanatory variables for later outcomes once the children reach adulthood. The 2010 reports on persistent poverty and maternal mental ill-health are likely to also be of interest to readers of this report as both include analysis of health outcomes (Barnes et al., 2010; Marryat and Martin, 2010).

It is important to stress that although some of the measures are directly associated with mothers, this is not meant to imply that there is no role for fathers. Instead, it is a reflection of the fact that, in order to collect detailed information on the pregnancy and birth of the child, the study sought to interview the child's natural mother at the first sweep of data collection. We acknowledge that this means that important insights about the children's lives will be lost by focusing on one key carer rather than on all the relationships children have, but there is not scope in the study to interview all the children's parents and carers at each sweep.

2.1.2 Inequality measures

Much analysis is limited by what measures have been collected and in this respect GUS is somewhat unusual in having a number of options from which to choose. Starting with the local context, the Scottish Index of Multiple Deprivation will be the primary factor of interest, with the prevalence of selected health outcomes and risk factors compared across quintiles of this variable (these split the sample into five equally sized groups). This is in line with the approach taken by the Scottish Government's long-term monitoring of health inequalities project. To date this has focused primarily on adult health, with the exception of low birthweight, and uses the income and employment domains of the SIMD index, rather than the whole index, as its inequality measure. The whole index includes a health domain (measuring mortality and illness rates, emergency hospital admissions, drug and alcohol related admissions, low birthweight and prescription rates for anxiety, psychosis or depression) which can cause problems when the analysis being conducted also includes some of these measures (for example, analysis of mortality rates by SIMD). However, the use of the full index is less problematic when analysing the child health outcomes selected above as only one of the measures also features in the health domain, birthweight. For this reason, and the fact that this report is considering a range of socio-economic measures, the full index will be used but caution needs to be exercised when considering the association between low birthweight and SIMD.

However, a significant proportion of families living in the most deprived areas are not socially or materially disadvantaged while many families with limited resources live in non-deprived areas. For this reason additional measures of family level deprivation will also need to be explored. The first of which is household income. ⁴ Evidence from the US suggests that household income is a key factor shaping the outcomes of children with long-term conditions with those from low income households having poorer health and worse outcomes in terms of days of schooling lost and overall attainment than children with long-term conditions from wealthier households (Case, Lubotsky and Paxson, 2002). UK evidence also suggests that income is an important marker of health inequality in childhood (Emerson et al., 2006). Income will also be explored using quintiles.

Other important measures of disadvantage include employment status, socio-economic status and mother's educational attainment. All of which have been shown to be associated with child outcomes in many of the GUS reports published to date. The initial analysis for the report looked at socio-economic classification rather than employment status (which just measures whether a person is in work or not) as the former is a clearer marker of structural inequalities within the labour market. However, due to space constraints socio-economic classification will only be referred to briefly in the text and no tables will be shown. Maternal education is considered in the second half of the report, which explores resilience to poor outcomes.

The proportion of children experiencing the health outcome, or risk factor of interest, at each time point will be compared across deprivation and household income quintiles. When looking at area deprivation it is important to be clear that this is an aggregate measure of local circumstances and does necessarily reflect individual experiences (Macintyre, 1997).

This detailed analysis of SIMD and income looks at all levels of these two measures and does not simply contrast children in the highest and lowest deprivation areas or the highest and lowest income quintiles. This makes it possible to identify whether patterns are linear or follow some other form. However, for simplicity a single relative measure of inequality for each factor is also presented (the relative risk) which shows the magnitude of the difference between the most and least disadvantaged groups in the overall measure. ⁵ Relative risk can be a useful illustration of the strength of association between the prevalence of a factor of interest in two groups, but it can be misleading if considered in isolation. For example, if the prevalence of something is 1% in the least disadvantaged group and 5% in the most, then the relative risk between them is 5. If the prevalences were 10% and 20% then the relative risk is lower: 2. A fivefold increase in risk could be of huge clinical or policy significance, but in some circumstances a smaller relative risk might be considered more important if the overall prevalence in question is higher and therefore affects more people. So, factors such as the baseline and overall prevalence, as well as the actual implications of the factor being considered, need to be borne in mind as well.