Health Disparities
Estimating causal effects of hypothetical interventions to reduce health inequality Lauren Zalla* Lauren Zalla Zalla Zalla Zalla Zalla Johns Hopkins University
Modern estimation of causal effects has focused on health outcomes, rather than on health inequality. We propose a framework for estimating the effects of hypothetical interventions on health inequality, as well as on population-average health, using observational data. First, to build intuition, we define three causal mechanisms through which an intervention can affect inequality: by targeting a determinant of health that has a different average causal effect on the outcome across social groups; by altering the distribution of a determinant of health across social groups; and/or by allocating the intervention according to modifiers of the average causal effect of the determinant of health that are differentially distributed across groups. Second, building on prior work, we define measures of the population intervention effect of a hypothetical intervention on absolute and relative inequality in a health outcome. These measures can be estimated from observational data under the same set of assumptions needed to estimate population-average causal effects. Our proposed measures incorporate an allocation rule that defines eligibility for the intervention and thus permits comparison of different intervention strategies (e.g., dynamic or stochastic interventions). We illustrate our proposed measures using simple examples and a real-world data example comparing the effects of different treatment strategies on the mortality inequality between non-Hispanic Black and non-Hispanic White people in HIV care in the United States. Finally, we discuss the limitations of our proposed approach and key considerations for its application. Our focus is on developing a deeper understanding of how public health and policy interventions shape health inequality and providing a framework for estimating effects on inequality using existing data and tools.