November 29, 2023
Most epidemiology studies treat race and ethnicity as confounding variables, which may mask and perpetuate systemic racism, according to new research from epidemiologists at the UNC Gillings School of Global Public Health.
Their new study, published in Epidemiology Reviews, found that in 2020 and 2021, three out of every five studies in the American Journal of Epidemiology (AJE) and Epidemiology that collected race data used poor practices for considering the complex connection between systemic racism and health effects.
Practices that do not examine the context of race data beyond treating it as a confounding variable can suppress our understanding of racism’s influence on health inequities and perpetuate structural racism. The study lays out a list of some best practices for using race and ethnicity in statistical analyses.
What is a confounding variable?
When studying the exposures and behaviors that might influence health, epidemiologists can encounter additional factors that skew their data. For example, when studying the association between physical inactivity and cardiovascular disease, factors such as genetic history, socioeconomic status and age may distort the association. In this scenario, genetics, socioeconomics and age are examples of confounding variables, or “confounders.”
To address this issue, researchers learn ways to adjust their studies so that the results remove the influence of these confounding variables. In theory, by “adjusting” or “controlling” for confounders, the results of a study should give a clearer picture of the direct relationship between the behavior being studied and its effect on health.
But in practice, this process is not so simple.
The study’s senior author, Shabbar Ranapurwala, PhD, who is an assistant professor of epidemiology at the Gillings School, said his own experiences of adjusting for race in studies helped him realize the importance of re-examining research practices that could mask inequities.
“In an earlier study, colleagues and I found that states with 40 or more restrictive firearm policies had 20-40% fewer police shooting deaths (after adjusting for race among other things). This finding by itself was important, and we could have stopped there,” Ranapurwala explained. “However, upon examining effect modification by race, we found that, in fact, this effect was even stronger for people who didn’t identify as Black (52% reduction), and among Black Americans, the rate of police shooting deaths actually increased by 53% in states with 40 or more restrictive firearm policies.”
In using a practice that more closely examined race data, Ranapurwala said that a “favorable effect of restrictive firearm policies was nullified (and then some) by systemic racism. The overall effect of 20-40% reduction hides the disparity and does not provide the true effect for any group.
“This made me realize that this happens all too often,” he continued, “including in some of my own prior studies where I adjusted for race without examining racial disparities. Without unmasking the systemic racism, we can’t do anything to solve it.”
Is race a confounding variable?
Research has found that race is a concept defined by society, not biology. It stems not from physiology but from complex and often convoluted associations with place, culture and skin color, as well as a conflated understanding of race and ethnicity. As our understanding of structural racism evolves, many researchers have called for a similar evolution in the practices that consider racism’s impact on health outcomes. This includes moving away from the practice of treating race as a biological variable similar to things like genetics and age.
Researchers in this new study say this is important because the discriminatory effects of systemic racism cannot be adjusted away. Instead, they must be examined directly so as not to perpetuate discrimination and limit the ability to develop interventions that address systemic inequities.
Ranapurwala said that this is where effect measure modification (EMM) comes into play. Effect measure modification is an epidemiological method used to focus on the associations between exposures and outcomes of interest within subgroups of people separated by an effect modifier, such as race and ethnicity, age, or gender.
“If we find disparities due to systemic discrimination, we must report them, explore why pathways cause them and think about how they can be mitigated,” he said.
Adjusting for race in post-COVID America
The new study examined 192 research publications from 2020 and 2021 in either AJE or Epidemiology that utilized race data, 160 of which were conducted in the United States. Both 2020 and 2021 were significant in U.S. history, as the COVID-19 pandemic and the murder of George Floyd heightened American awareness about racism’s impact on public health.
The researchers analyzed each study’s use of race data based on the extent to which their practices revealed or masked issues of systemic bias. According to the study, good practices utilized EMM (with or without confounding) while also reporting disparities and discussing mechanisms that led to disparities. Such practices help to shed light on population-level disparities, understand their drivers and guide the solutions that can strengthen health equity.
Conversely, poor practice utilized confounding only and reported neither disparities nor mechanisms or reported disparities only.
Among the 192 studies where the authors used race and ethnicity, only 22.9% – fewer than one in four – exhibited good practices, while 63.5% – three out of every five – used poor practices. In addition, 13.5% percent of studies used practices that were categorized as neither good nor poor, meaning they used data on race for descriptive purposes only.
The research team also found that many of the studies used categorizations for racialized groups that either excluded some groups or collapsed groups with different lived experiences into larger groupings. The most common of these collapsed groupings was “Black, white, Hispanic or other.”
“Collapsing categories of race is a popular practice in epidemiological studies, so this finding was expected,” said doctoral student Monica Martinez, MPH, who is the study’s first author. “However, the frequency with which the ‘other’ group is often ignored, and the frequency with which studies collapsed race categories without providing any reason, did surprise me. These seemingly routine practices may further mask systemic racism and hide racial health disparities.”
The road to equity in analysis
“The fact that researchers use race in their studies indicates an understanding that there are structural disparities in who experiences certain exposures and health outcomes,” the study authors wrote, “and yet, the analytic methods do not reflect this understanding.”
The researchers, therefore, made a series of recommendations based on these findings that they hope can help peers in epidemiology leverage that understanding while improving communication, analysis and interpretation of race in their research.
The 13 recommendations can be found in detail in the full study. Broadly, they provide guidance on how to describe the context behind race and socioeconomic data, avoid collapsing race categories, effectively implement EMM, and promote inclusion while being explicit about how systemic racism drives observable health inequities.
“We hope that journals adopt these recommendations across public health research, just like the National Institutes of Health adopted inclusion policies for women and minoritized people in 1994,” Ranapurwala said. “Journal policies are one of the best ways to ensure that science which unknowingly or knowingly masks systemic discrimination – be it from racism, sexism, ageism or any other -ism – does not impede our ability to protect public health.”
The research team is currently working to complete another, similar systematic review of injury and violence prevention research that began before this published paper. They are also pursuing ongoing work to examine other metrics and methods used in public health and epidemiology that can potentially harm public health by masking systemic discrimination.
Contact the UNC Gillings School of Global Public Health communications team at email@example.com.
March 4, 2024 James Swenberg, DVM, DACVP, PhD, Kenan Distinguished Professor at the UNC Gillings School of Global Public Health in the Department of Environmental Sciences and Engineering, died October 5, 2023. There will be a Scientific Symposium to honor him and his work on March 22 from 3–5 p.m. in 133 Rosenau Hall at the Gillings School.