22 23 The slope order of inequality consisted of wealth quintiles with values of 0.1, 0.3, 0.5, 0.7 and 0.9, that is, the midpoints of each quintile on a scale of zero (least NVP-BEZ235 structure wealthy) to one (most wealthy). The slope order of inequality was modelled as a continuous variable, so that the slope or coefficient of a logit linear regression line across all five quintiles represents the difference in outcome between the hypothetically wealthiest and least wealthy participant. Exponentiating this slope coefficient results in an

OR, which is the ratio of the odds of the outcome in the wealthiest compared with the least wealthy participant. This OR is also known as a relative index of inequality.22 Advantages of this method of quantifying inequality are that it includes all participants, instead of just comparing the highest and lowest quintiles, it accounts for the number of participants in each category and it provides a single overall measure of inequality. We included

all participants in the main cross-sectional analysis in order to compare the distribution of illness burden in the whole population with the distributions of diagnoses and treatments in the whole population. This meant that diagnosis was assessed even in those who did not meet the criteria for ‘illness burden’, and treatment was assessed

even in those with no diagnosis. For the subsidiary analysis using longitudinal data, we estimated the OR of receiving a diagnosis by a subsequent wave only for those who had met the criteria for ‘illness burden’ in a previous wave, and then the likelihood of receiving treatment only for those who had received a diagnosis in a previous wave. This was a subsidiary analysis as the number of participants that could be followed over time in this manner Cilengitide was small, particularly for treatment in angina and depression. Results The whole sample (n=12 765) was composed of participants aged 50 years or more who had responded to at least one wave of ELSA from 2004–2005 until 2010–2011. The response rate in 2004–2005 was 82%.24 25 In wave 5 (2010–2011), self-reported medical diagnosis for all five conditions increased as wealth decreased, for example, in depression from 4% in the wealthiest quintile to 11% in the poorest (table 1). There was little variation between the waves for each of the five conditions (table 2).