Health status balancing weights for estimation of health care disparities

Propensity score (PS) weighting is widely used to make causal or unconfounded comparisons between groups. Particularly, the balancing weights approach unifies existing PS weighting methods and allows one to identify target populations clearly. Few studies have applied PS weighting to generate estimates that are concordant with the Institute of Medicine (IOM) definition of racial disparities. This PS weighting approach aims at estimating racial disparity conditioned on balancing the health status distributions between groups for a target population. Despite these attempts, however, no study has extensively examined the balancing weights in implementing the IOM definition. This article presents the balancing weights based on the health status PS applied to the IOM definition of racial disparity. Particularly, we propose using the absolute standardized difference to assess the degree to which specific balancing weights satisfy the IOM definition. We consider hybrid health status balancing weights, which are equivalent to the inverse probability weights and overlap weights as special cases. We propose a data-adaptive selection of the tuning parameter for the hybrid weights to minimize the bias of disparity estimates due to the alterations of the distributions of socioeconomic status variables by weighting. In our simulation study, the hybrid weights were shown to perform well in implementing the IOM definition. The practical utility of the proposed methods is illustrated in a study of bladder cancer treatment disparity.