Accounting for differences in population health between the regions of the United Kingdom: a new measurement framework for ordinal data

The paper investigates the association between regional health outcomes and socioeconomic characteristics in the United Kingdom (UK), based on a recently proposed measure of the degree to which the populations of different regions occupy well-defined strata in the national health distribution. The headcount index of health stratification is well-defined even if only ordinal health data are available and has a straightforward interpretation as the population-weighted mean difference in the probabilities that the healthier of any randomly chosen pair of individuals will be from the region with the better rather than the worse population health. The paper provides alternative aggregate decompositions of the index based on the construction of counterfactual distributions using indirect and direct standardisation techniques, with the indirect aproach also providing the basis for a detailed decomposition of the composition effect. The empirical study shows that health stratification is largely due to differences in the socioeconomic and demographic composition of regions rather than in regional health outcomes conditional upon individual-level sociodemographic characteristics, with age, ethnicity and qualifications all more important factors than income.