Health Disparities
Measuring Policy Effects on Health Inequities Hidden by Compositional Change Colleen A. Reynolds* Colleen Reynolds Reynolds Reynolds Departments of Public Health and Epidemiology, Erasmus University Medical Center
Background
Reducing socioeconomic inequalities in health is a major public health priority. However, when policies affect socioeconomic status, estimating changes in health inequalities becomes more challenging because the composition of the groups being compared changes.
For example, imagine we want to understand how a policy aimed at increasing educational attainment affects educational inequalities in life expectancy. Educational attainment is an intercurrent event between exposure and outcome. Stratifying on this variable would obscure the effect of the policy on life expectancy through educational attainment, the primary mechanism.
Methods
Utilizing the principal stratification framework, we propose a new estimand relevant to policy evaluation: the effect of the policy on health inequalities across the distribution of education that would have existed in the absence of the policy. Using the example described above, we apply this estimator on simulated data and outline the interpretation and policy relevance.
Results
In our simulation we observe a 3.3-year increase in overall life expectancy. When we stratify on observed educational attainment, the policy is not associated with a change in educational inequalities in life expectancy. When we stratify on educational level that would have been attained in the absence of intervention, the policy is associated with a 5-year reduction in educational inequalities in life expectancy.
Conclusion
We demonstrate that our approach can capture changes in health inequalities that are obscured by compositional changes in the disparity population, overcoming post-treatment bias that results from simple stratification on the disparity population of interest This approach has broad potential applications to scenarios where the exposure of interest alters the composition of the disparity population, and offers a new methodological tool for health inequalities research.
