1 Further information on the design, development and future of the project is available from the study website:
3 Acute illness was initially considered for exploration. However, changes to the question wording across the sweeps meant that the measure was not consistent across all years. Also, the question in GUS measures how many different conditions children have had, rather than how many illness episodes have been experienced, so it was not an ideal measure for this report's purposes.
4 Household income only started to be included routinely in social surveys within the last decade so its use in analyses of health inequality is less extensive than is the case for area level measures which date back to the 1980s, or social class, which has been measured for many decades. For more discussion of the measurement of income in surveys, and how it is measured in GUS, see the report on persistent poverty (Barnes et al., 2010).
5 Note that the relative risks presented here do not estimate the difference between the two categories that happen to have the highest and lowest prevalence for the outcome. As a measure of inequality it compares the most and least disadvantaged groups according to the underlying classification, regardless of the pattern in the data.
6 See, for example, the Food Standards Agency's advice on this:
7 The concerns parents could choose were: his/her language is developing slowly; it is hard for other people to understand him/her; he/she doesn't seem to understand other people; he/she pronounces words poorly; he/she doesn't hear well; he/she stutters; other concerns.
8 The SF-12 questionnaire measures health related quality of life and covers the impact of physical, emotional and psychological symptoms on people's physical functioning and ability to carry out normal activities. It is a shortened version of the widely used SF-36 questionnaire. See: http://www.sf-36.org/tools/sf12.shtml
9 Unlike the tables in Chapter 3, these tables compare the number of negative outcomes children experienced according to each resilience measure, rather than the other way round, so the relative risk associated with each factor cannot be calculated.
10 The statistical analysis and approach used in this report represents one of many available techniques capable of exploring this data. Other analytical approaches may produce different results from those reported here.
11 The confidence intervals in the rest of the table are in line with what might be expected with a sample size such as this. GUS analyses typically involve the whole sample which, with around 4,000 cases, usually result in much more precise estimates. This analysis was not designed with the intention of estimating the strength of association between factors with a high degree of precision, it was intended as an exploration of resilience as a concept in explaining the avoidance of negative outcomes. For this reason, the discussion focuses on the factors identified rather than on the estimated effect sizes.
12 The two scale measures did not have comparison groups, instead their odds ratios are an estimate of the increase in odds associated with a one unit change in the underlying values in the scale.
13 Note that the fact that the significant association was between medium and low levels, rather than high and low levels is likely to be due to the small sample size for the high satisfaction category. The key point is that the relationship is telling us something about people with low satisfaction levels relative to other groups, rather than anything specific about having medium satisfaction levels.