Decomposing geographic disparities in health and health behaviors to inform community interventions

INTRODUCTION: Good understanding of the sources of community-level disparities in health and health behaviors is critical to properly formulate community-based interventions. We developed a method to quantify relative contributions of demographic, socioeconomic and contextual factors to community-level disparities. METHOD: Using a mixed effects model, the new method decomposes the effects of significant risk factors on health outcomes into demographic, socioeconomic and contextual disparity indices. The indices are defined as the sum of the products of community average levels minus the expected state-level value of the relevant risk factors multiplied by the magnitude of their respective regression coefficients. The method is illustrated using town/city-level disparities in hypertension prevalence in Massachusetts.

RESULT: Using state average prevalence rate of 26% as a benchmark, the ratio of community-specific prevalence rate to state average rate ranged from 0.4 to 1.34. In logarithmic scale, demographic factors (sex, age, race/ethnicity), socioeconomic (education, income, employment) and contextual (e.g., per capita income, density of housing units) factors contribute 56% (range: 34%-70%), 15% (range: 8%-24%) and 25% (range: 15% to 36%), respectively, to the variation in community-level prevalence estimates. For a given community, domain specific disparity indices were frequently in opposite directions. CONCLUSION: The effects of a particular domain for town-level disparities can easily be overlooked when only overall disparities are calculated. Use of domain-specific disparity indices will help epidemiologists understand the relative importance of demographic, socioeconomic and environmental factors to the observed community-level health disparities. Such information can be used to formulate and prioritize community-based interventions.