“Making progress on eliminating racial and other disparities in health and health care requires really good measures and interpretations of health care disparities” stated Dr. Lisa Clemans-Cope, a panelist at AcademyHealth’s December meeting “Harnessing Medicaid to Improve Health Equity: A Research and Policy Agenda.” Clemans-Cope presented research from a collaboration with colleagues Dr. Stacey McMorrow and Dr. Bowen Garrett and funded by the Robert Wood Johnson Foundation.
Quantifying and interpreting measures of disparities in health care is challenging, particularly with respect to measuring the role of systemic racism. The Institute of Medicine (IOM) offers a principled approach to defining disparities in health care access and use when evaluating health care system performance. Under the IOM principles, racial and ethnic disparities in health care outcomes for reasons outside of clinical need and/or patient preferences are considered inequitable and unjustifiable. Thus, implementation of IOM principles urges researchers to use statistical adjustments that control for patient characteristics related to clinical need (such as age, gender, health status, patient preferences, etc.) and not those related to socioeconomic status (SES) in their analyses. Despite these guidelines, many published articles either ignore or choose not to implement these principles, resulting in estimates that may be ill-defined, more subject to misinterpretation, and that may under- or overstate actual racial disparities in care.
Where Analyses Go Wrong
To implement IOM’s guidance, researchers must first define the disparities being measured as they relate to the study’s goals. Absent these clear definitions, researchers can risk making inaccurate policy inferences. Often in these types of analyses, researchers will use statistical adjusters, such as neighborhood deprivation indices that include SES variables like income, housing, and transportation access, to control for extraneous factors and pinpoint the disparity in question. Another common statistical adjuster is health status. This factor, which is often driven by structural racism, further underscores why researchers should articulate a clear conceptual framework and methodological approach. Adjustments that include SES and other socio-demographic factors control for these variables across the population being studied. As a result, they not only understate the impact these factors have on disparate outcomes, but they can also erase or diminish health disparities caused by hospital-level racial segregation, discriminatory health care, or other systemic inequalities.
Five Recommendations to Improve Measuring and Interpreting Racial Disparities in Health Care
To combat these challenges, Clemans-Cope and colleagues McMorrow and Garrett at the Urban Institute’s Health Policy Center offer a set of five recommendations for producing and interpreting estimates of health care disparities.
- Include a clear definition of health care disparities consistent with the goals of the study. Definitions should also be transparent about what they are measuring. For example, is the study evaluating unequal treatment in a specific health care system or quantifying the broader inequities in society? Often, researchers only define disparities implicitly where the disparity is equal to whatever remains once the statistical adjustments are made. Introducing clear definitions of disparities and measures allows users of the analysis to easily understand where disparities lie and what contributes to them.
- Provide a clear conceptual framework to interpret estimates of health care disparities. Even in more recent papers, Clemans-Cope, McMorrow and Garrett emphasize that clear frameworks are often absent: “The more explicit researchers can be about the present policies and mechanisms left behind by lingering effects of discriminatory past policies and practices that lead to disparities, the better we can interpret the estimates of disparities, we can test the mechanisms with the data, and we can develop policies that reduce disparities.”
- Discuss data limitations in implementing a given definition of disparity. Data limitations of measuring race and ethnicity used in disparity estimates are well-documented and understood. Clearly stating these limitations allows readers and fellow researchers to understand where inferences were made or where there may be ambiguity in results.
- Use the full set of covariates in regression models for estimating disparities and include the appropriate components of those in that disparity measure. This helps reduce omitted variable bias and facilitates calculation of comparable disparities across studies.
- Account for the sources of disparities and associated factors to move beyond documenting disparities and towards estimating models that identify structural drivers. This last piece is crucial, underscored Clemans-Cope: “To reduce disparities, the mechanism generating them needs to be better understood, especially the factors that are amenable to modification by public policy and other actions.”
This post highlights quotes and learnings from the panel "Methodological Challenges Associated with Improving Health Equity in Medicaid" presented at the meeting "Harnessing Medicaid to Improve Health Equity: A Research and Policy Agenda" on Dec. 1 and 2, 2021. This meeting was co-hosted by Julie Donohue of the University of Pittsburgh, Susan Kennedy of AcademyHealth, Genevieve Kenney of the Urban Institute, Chima Ndumele of Yale University, and Kosali Simon of Indiana University.