To achieve health equity, Medicaid programs have increasingly incorporated the concept as a priority in every aspect of their strategies and functions, but data collection challenges continue to hamper monitoring efforts. As one hurdle, health equity remains a goal with no common definition and this ambiguity filters down from ideal to implementation very tangibly when collecting data on marginalized populations.
“Challenges around race and ethnicity data changed from one of accuracy to one of completeness,” said Elizabeth Lukanen, the deputy director at State Health Access Data Assistance Center (SHADAC). “There is a lot of missing data on race and ethnicity because reporting is voluntary, and the federal reporting guidance is not strenuous.”
As a result, definitions of race and ethnicity used by states vary widely. During the self-reported enrollment process, some state programs use the 2 question/5 race option-format definitions put forth by the federal Office of Management and Budget in 1997; others use the 2 question/14 race option-format used by the U.S. Department of Health and Human Services under which falls the Centers for Medicare and Medicaid Services (CMS). There are 62 variations of race/ethnicity definitions across states, including one that has disaggregated to 37 different categories in one question. According to Lukanen, not all of these definitions and approaches to data collection comport with the best practices recommended by US Census, which suggests combining race/ethnicity questions to better elicit responses, for instance.
Moreover, the quality of the collected data, as measured by the Medicaid’s Data Quality Atlas, differs drastically across states. The publication of the Atlas is a culmination of efforts by CMS and other stakeholders to improve monitoring of the quality and usability of administrative records submitted to the national Medicaid data system known as the Transformed Medicaid Statistical Information System (T-MSIS). It ranks the states according to how much data is missing, how valid and reliable the data are, and how closely their submissions compare to the U.S. Census Bureau’s American Community Survey.
“We have couple of states where all the records are missing race and ethnicity information across the board.” said Carol Irvin, a senior fellow at Mathematica who works closely with CMS on the T-MSIS matters. “We have states where the rate of missing information is very low, but also other states where it is very high.”
For example, California is classified a “low concern” state as less than 10 percent of the race and ethnicity records are missing and the distribution of the race and ethnicity group is comparable to that in the American Community Survey. On the other hand, New York is categorized as a “high concern” state for having a considerable amount of missing data and underreporting of the Hispanic group. There are endeavors under way to refine the quality of state reporting, such as scoring the state T-MSIS submissions and imputing missing data.
But what are we collecting and why? The foundation of the modern American social support system was laid during Roosevelt’s New Deal era and expanded upon with the Johnson administration’s Great Society initiatives. While health and social service benefits affected tidal change in American society, significantly expanding government and extending new or richer benefits to certain populations, they coincided with the Jim Crow era that legally codified racism, and the Civil Rights era during which those opposed to racial equality sought to couch discrimination and oppression in less overt ways.
“We don’t think about how racism plays a part in how our public health system reacts or determining who is covered under our public insurance systems,” said Ruqaiijah Yearby, the executive director and co-founder of the Institute for Healing Justice and Equity. “Law structures the systems in a way that disadvantages racial and ethnic minorities, and we need to think about the ways the racism is tied to data.”
Dr. Yearby described how the racial bigotry which has existed throughout our nation’s history underpins and scaffolds the laws, programs, and systems we use today, and structural racism remains embedded in ways that we are only now as a nation beginning to examine and recognize, if not reckon and recompense. For instance, Yearby highlighted the gaps in legal protection for the vulnerable groups, pointing at the early funding and policy decisions that lead to disproportionate harms in settings like nursing homes where there are persistent racial and ethnic disparities in both access to and the quality of nursing home care.
Further, the valid mistrust by those who have been historically disenfranchised of how data have been collected and used, which affects willingness to supply such information, stems from centuries of the discrimination and mistreatment inherent in our systems. Therefore, “we really need to have the communities lead and drive the change and think about what they think is necessary to get in terms of data to support them,” said Yearby.
“Big data has both opportunities and challenges for addressing equity for Medicaid populations,” said Susan Goold from the University of Michigan. According to Goold, the paradox of data collection remains that, for researchers to draw generalizable conclusions and control for confounders in race/ethnicity data, a large enough sample size must be available. But the populations for which we have the least data, and at meaningful and intersectional levels, are also the smallest and are not even themselves homogenous. Moreover, many equity questions cannot be approached using the data elements typically collected. For example, more information is needed about member experiences both in and outside of the health care system, yet not much is available.
Ultimately, “we know that measuring things is important, but it’s not always possible to measure things that are important and not everything that you can measure is important,” said Goold.
The lack of satisfactory data with which to measure health equity reinforces the need to include communities in discussions about what data is needed, how the elements are defined, and how it will be used.
For example, administrative data can be supplemented with survey or focus group input to provide context to enrollment and utilization data. “The data aren’t born in a lab or a hard drive. You can actually ask people in a systematic and rigorous way about what they are experiencing and what they need,” said Anne Schwartz from the Medicaid and CHIP Payment and Access Commission (MACPAC). “Valuing what the people who are being served want is important to have as a part of the picture,” she said.
Members can be asked in systematic and meaningful ways about what they value, and their input incorporated as policies and programs are developed and evaluated. In particular, mixed method approaches to data collection can be employed to yield richer data that provides context to the information we now make available about the Medicaid program.
While the needs of federal, state and local agencies must be balanced and coordinated, community and policy priorities can in part drive data collection efforts. States can:
- Analyze how best to use the information they already have, and to define what they need, by evaluating the information’s usefulness. Populations, programs, and services have changed significantly over time, yet many data elements remain as legacy to outdated and sometimes racist systems which prioritized the needs of certain populations while excluding others, while other data is of poor quality or is not yet collected.
- Include enrollees and community members to gather their perspectives as new processes and procedures are developed to increase participation and more accurately reflect their needs.
- Use a single question for race/ethnicity that includes options for ‘mixed race’, ‘unknown’, and ‘choose not to answer’ to increase accuracy and completeness.
- Require an answer to the race/ethnicity question (by including “Don’t Know” and “Choose Not to Answer” options) during enrollment and re-enrollment before registration completion to minimize missing data.
- Train enrollment navigators and specialists to increase frequency and accuracy of completing race/ethnicity data.
- Use financial and other incentives to encourage MCOs to ensure complete and accurate data, for example by requiring Healthcare Effectiveness Data and Information Set (HEDIS) submissions by race/ethnicity as a starting point.
At the end of the session, Schwartz stressed that the limitations in existing approaches to data collection should not discourage the work to improve them. “Data don’t have to be perfect. We can continue to do this work while we improve it, and we can walk and chew gum at the same time. We can’t wait until all the improvements have been made before we do the work. So, let’s go forth and do the work, we will get better as we do it.”
This blog post highlights quotes and learnings from the panel "Improving Health Equity in Medicaid: Data Needs, Challenges, and Opportunities" 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 M. Kenney of the Urban Institute, Chima Ndumele of Yale University, and Kosali Simon of Indiana University.