Methods for Measuring Racial and Ethnic Disparities in Health Care with Observational Data: Defining a Counterfactual for Causal Inference

It has been 13 years since the Institute of Medicine (IOM) released Unequal Treatment, the seminal report identifying significant health care disparities in the U.S. In this report, the IOM committee defined health care disparity as all differences except those due to clinical appropriateness, need, and patient preferences. While the IOM definition permeates health services research, it is equally relevant for social work research.  This is particular true as the field of social work takes on the grand challenge of closing the health gap and ensuring health equity for all. 

The purpose of this workshop is to present statistical methods that are used to explain racial and ethnic disparities in areas of health care ranging from mental health to developmental disability.  The workshop panel includes experts currently engaged in research using the methods to be discussed. 

Panelists will present a comprehensive overview of counterfactual methods to identify and quantify racial and ethnic disparities in service use.  Counterfactual analysis is the dominant causal paradigm in recent literature in statistics. This type of analysis provides a framework for better understanding racial and ethnic disparities in health care, and for rigorously conceptualizing scientific questions that are relevant in the field of disparities.

In typical gold standard randomized control trials (RCT), the "treatment" is randomized to a set of respondents. The causal effect can then be identified as a difference in outcomes between the treatment and control groups. The underlying assumption is that the randomization allows for the approximation of the difference between the outcome if an individual received the treatment and the outcome if the same individual did not receive the treatment. The reason that this is a counterfactual is that an individual cannot simultaneously be in both the treatment and control group.

Race/ethnicity is not "manipulable" in this way and cannot be randomized.  Nonetheless when identifying service use disparities, the counterfactual framework helps to guide adjustment for allowable or "just" differences but not un-allowable or "unjust" differences.  The panelists will describe commonly used methods for measuring health care disparities in a regression context including: 1) reduced covariate method; 2) propensity score method; 3) rank and replace method; 4) residual direct effect method.  The panelists will discuss the benefits and limitations of each method as well as concordance with the IOM definition of disparity. 

In order to design interventions and policy to eliminate racial and ethnic disparities in health care, the counterfactual framework demands the formalization of pathways leading race and ethnicity to outcomes.  Pathways may not simply exist at the individual level, but likely occur through the complex interaction of individuals and their environments. Panelists will demonstrate how these methods can be used to “unpack” the race variable and quantify the contribution of pathways.  Panelists will present empirical applications of these methods.