Panel Highlights Gaps in Race, Ethnicity Data Collection
The pandemic has put a spotlight on the fact that the collection of race/ethnicity, language, sexual orientation and gender identity data is often limited in federal and state health programs, as well as in commercial insurance. An expert panel hosted by the Alliance for Health Policy recently discussed barriers that lead to limitations in health data collection and analysis to guide policy making.
Gathering this demographic data is essential for having a complete picture of the status of disparities, for understanding how different factors drive disparities, for guiding resources and efforts to advance equity and for measuring progress and establishing accountability toward achieving equity, as well as for identifying best practices or strategies to advance equity, said Samantha Artiga, vice president and director of the racial equity and health policy program at the Kaiser Family Foundation. “When we lack data or when the data have quality issues, it is much harder to achieve all these goals,” she said.
Despite the existence of some federal standards for gathering this type of data, there are gaps and limitations in the racial and ethnic health data that are available today, Artiga said. “These gaps and limitations have existed for a long time, but I think have really been amplified and gained increased attention throughout the COVID-19 pandemic, when the lack of comprehensive race and ethnicity data was recognized as a hindrance for both understanding disparities and addressing them.”
Artiga noted that some datasets simply don't have race/ethnicity data available. “When race/ethnicity data is available, there often is insufficient data for smaller population groups, particularly American Indian and Alaskan Native people, as well as Native Hawaiian and other Pacific Islander people,” she said. “We often also lack data for subgroups within those broad racial and ethnic categories.”
The five minimum racial/ethnic categories required by federal programs such as Medicaid are very broad and within them there's a significant diversity of populations who may have very different health and healthcare experiences. “We've also seen inconsistency in racial and ethnic classifications, so those federal minimum standards apply to federal datasets, but not to other datasets,” she said.
At the state level, she explained, some states have not reported vaccination by race/ethnicity over the course of the pandemic. “We saw a lot of variation across states in their different categorizations and reporting of racial and ethnic data. Many states did not report data for the smaller population groups, including American Indian, Alaskan Native and Native Hawaiian and other Pacific Islander people. And very few states reported data in a way that allowed for intersectional analysis.”
Although the focus on health equity is crucial, it can be pretty difficult to make these changes in data collection, said Niall Brennan, chief analytics and privacy officer at Clarify Health. “The aspirations of where we want to go is running smack into a brick wall of reality and as a data person and a data nerd, it's both personally and professionally very frustrating not to have that information,” he said. “Our data collection mechanisms aren't really optimized for this type of data. We have universal claims and administrative data infrastructure. Many of those were designed 20, 30, 40, 50 years ago when the collection of this information wasn't considered important or it wasn't prioritized. And then we've got this massive, rapidly evolving, emerging set of more modern data systems — all our investments in clinical health records. There's been incredible progress, but a huge rate-limiting factor is both an inability for the data to be truly interoperable and flow, and also, to a certain extent — almost too much data in electronic health records and sort of separating the wheat from the chaff and figuring out what's important.”
Prior to Clarify, Brennan was president and CEO of the Health Care Cost Institute, where he oversaw HCCI’s overall research agenda, highlighting trends in U.S. healthcare spending and the factors behind those trends. Before that, he served as the chief data officer at the Centers for Medicare & Medicaid Services.
Another reason why this type of data collection is so hard involves trust, he said. A lot of people are unsure or afraid of how this information might be used, even if people's intentions are really pure. “You can’t change what you can’t measure, but a lot of the folks that we're trying to capture this more granular information on are people who have been repeatedly let down over many generations by the healthcare system and by society in general,” he said. “When you start to make these well-intentioned efforts to try to harvest the information, you can run into some resistance because people are not sure of how it's going to be used.”
Elizabeth Lukanen, M.P.H., deputy director of the State Health Access Data Assistance Center (SHADAC), a health policy research program at the University of Minnesota School of Public Health, said that states recognize they have quality problems with their data. There are four states right now that have unusable data, according to CMS’ assessment, and a large handful where it's really unusable, she said. “What a lot of states are doing is really committing to improving their race/ethnicity data in Medicaid and we work with them to try to make these data better.”
“As we have been working with states to try to think about how to improve these data, we've come up with two streams of activities. One is to enhance the existing data that they have,” Lukanen said. “We've been working with states to leverage other data that they might have available to them to sort of triangulate the data within the Medicaid program. This often means identifying individuals in other data systems they have access to.”
The second approach is more of a statistical methodology where you're filling gaps in your data using other information about the individual. “If my data were missing, it would look at my surname and maybe where I live and try to give some indication of what my race might be,” she explained. “You can imagine there are concerns with this methodology. but it is one way that states can use to fill in some of the gaps. Finally, we always recommend that states validate their data against other sources of data on race/ethnicity, and probably the best source is census data.”
Irene Dankwa-Mullan, M.D., M.P.H., is chief health equity officer and deputy chief health officer at Merative (formerly IBM Watson Health). She noted that sometimes our current practices reinforce norms of homogeneity within Black, Hispanic, and indigenous communities, “but we know that there are within-group differences, or population group differences, or different features or risk attributes that may contribute to their outcomes or their risk or resilience.”
Dankwa-Mullan said we may also be missing the influence of environmental, occupational, work, place exposure, and life course exposures on health outcomes. She also used the term “data empathy” to describe the knowledge or experience about the people or the places or the factors that actually make up the data. A lack of knowledge of those data sources or an inability to recognize them “results in our inability to optimize algorithms or models that go into decision-making processes.”
