Estimating effects of hypothetical public health interventions on health disparities: A standardization-based simulation approach.

Causal inference research typically estimates effects on population average health rather than impacts on health disparities. We present a standardization-based approach to simulate how altering population distributions of exposures might change health disparities, with explicit consideration of two mechanisms: differential exposure distributions and heterogeneous exposure effects across social groups. Using data from the Health and Retirement Study (n=11,322) and the English Longitudinal Study of Ageing (n=5,179), we examined how hypothetical interventions reducing social isolation and loneliness might affect socioeconomic and racial disparities in cognitive functioning among older adults. Social isolation was longitudinally associated with lower average cognitive functioning in both HRS and ELSA, with evidence of effect heterogeneity in the HRS sample, where the adverse association was stronger among individuals with lower education, Black race, and paradoxically, higher income/wealth. Simulations demonstrated that reducing social isolation would narrow cognitive disparities across income, wealth, education, and race. Interventions specifically targeting exposure disparities achieved greater reductions than uniform approaches. This framework enables researchers to move beyond estimating average treatment effects to quantifying how different intervention strategies might reduce health disparities between social groups.