A0176
Title: Doubly robust inference for measuring (un)explained health disparities
Authors: Yan Li - University of Maryland (United States) [presenting]
Abstract: A general framework for statistical inferences is established in measuring (un)explained health disparities between privileged (AG) and marginalized (DG) groups. While the Peters and Belson (PB) method is commonly employed in literature, its reliance on parametric modelling of the outcome can yield misleading results under model misspecification. To address this, an alternative method based on propensity scores (PS) adjusts the empirical distribution of the outcome variable by constructing pseudo-weights that equalize the empirical distributions of the explanatory variables across groups. However, the PS method hinges on the assumption of a valid propensity model to construct pseudo-weights. A rigorous procedure for constructing doubly robust (DR) estimators is developed to measure health disparities. DR estimators use the estimated propensity scores as well as an outcome regression model and remain consistent as long as one of the two models is correctly specified. Variance estimation is discussed under the proposed framework. Results from simulation studies show the robustness and efficiency of the proposed DR estimators as compared to existing PS and PB methods. The proposed method measures the disparity in body mass index in the national health and nutrition examination survey from 1999 to 2004.
