Primary Submission Category: Causal Inference and Bias/Discrimination
The Observational Target Trial: A Conceptual Model for Measuring Disparity
Authors: John Jackson, Yea-Jen Hsu, Raquel Greer, Romsai Boonyasai, Chanelle Howe,
Presenting Author: John Jackson*
We present a conceptual model for measuring disparity using an observational target trial (an inception cohort with minimal intervention). First, we discuss disparity definitions in public health and medicine and how they relate to a descriptive measure of disparity. Second, we outline the key elements of the target trial and provide inverse probability weighting and g-computation estimators to emulate it. Third, we discuss non-random selection into the eligible population and its contribution to a disparity measure from a normative and moral perspective. Fourth, for investigators who wish to do so, we extend our target trial model and its emulation to remove the contribution of non-random selection from disparity (via a stochastic intervention on all or some of the variables that establish eligibility) under various causal structures. We demonstrate our methods using electronic medical records to measure racial disparities in hypertension control in a regional health system.