Decompositions of differences between groups as a population health tool

Date: December 13, 2012
Times: 13h00 to 14h15 Eastern Time (Montréal)
Presenter: Daniel Dutton

Inequality is a major concern of population health researchers. Often inequality is framed as a difference in average outcomes across identifiable groups: men and women, different ethnicities, residents of different regions, or people with different occupations. Quantitative analyses of these groups indicate that there are important individual-level variables, like income, that are associated with differential health outcomes within and across groups. Population-level variables, such as regional differences in policy, opportunities available to societal groups, or culture, which are unmeasured or unobservable, are not quantified in traditional regression analysis of individuals. Individual-level risk factors are generally convincingly portrayed (in a positivist paradigm) as an important target for intervention while unmeasured factors are easily neglected.
If one believes that there is a common exposure operating on all individuals in a population, such as a government policy, then it might be worth exploring the relative importance of individual-level and population-level factors in determining group differences. One way to do this is to decompose the difference between outcomes of two groups. Using such a technique, the expected effect of policies that affect individual-level variables, such as income redistribution, can be estimated. It is difficult to make such estimates in typical regression analysis when comparing across identifiable groups since the difference in group outcomes could be attributed to differences in levels of covariates or differences in the effects of those covariates.
This presentation will outline the reasons for conducting a decomposition exercise, including the history of the use of such techniques; an example of a basic decomposition; and then examples from the literature. By the end of this presentation, the audience will understand how decomposition techniques can be a valuable tool in quantitative population health work.