# Improving biostatistical models to decrease health disparities

August 15, 2024 – Briana Stephenson, assistant professor of biostatistics at Harvard T.H. Chan School of Public Health, develops biostatistical models to better understand population health disparities. In this Q&A, she shares what motivates her work and the broad applications of her research.

Q: How did you become interested in biostatistics and public health?

A: I’ve always loved math, but I also was really interested in the life sciences. Around 7th grade when I was first exposed to life science, my teacher told me, “You’re going to be a doctor.” So I went down that route and stayed pretty rigid. Every decision that I made through school and college was with the intention of becoming a physician.

When I got to college, I was a math major and pre-med. While the topic of health and medicine interested me, I found myself more captivated in my math courses. I loved learning how mathematical theories and principles could be applied to solve interesting problems, from understanding traffic flow on the Massachusetts Turnpike to finding disease causes in vulnerable populations. However, I stayed on the pre-med track. It was not until my first semester of medical school that I found myself at a crossroad.

At medical school, while I wanted to improve the health of my community, my scope of impact was limited to the individual patient level. As part of my practical training, I worked at a children’s sports clinic, where I saw a lot of repetition of injuries and treatment plans. Over time, I began to notice patterns in the types of injuries being presented, based on the patient’s age, neighborhood, and other demographics. My attention began to shift to identifying the source of these clustered cases to better strategize interventions to reduce the number of injuries within these patient groups.

I took a seminar in epidemiology and biostatistics and learned how I could use math to quantify and explain trends at the population level, and predict the impact that interventions could have on the community if implemented. That seminar is what ultimately motivated my pivot from medical school to public health.

Q: Could you briefly explain what the field of biostatistics is?

A: In general, statistics is trying to make sense of data—building the quantitative story behind different problems that we see in the world, and using models to identify different trends. Biostatistics is is a subfield of statistics applied to biological sciences, including medicine.

Biostatistics distinguishes itself from statistics by the fact that it is an interdisciplinary and applied discipline. Both statistics and biostatistics are engaged in the development of methods and theory, but the motivation for our developed methods in biostatistics is rooted in their application. We don’t want our work to just sit in a silo—we want to see it applied somewhere out there in the world, in the areas of health and medicine.

Q: What does your research focus on?

A: On the math side of things, I develop models that can identify underlying patterns from a large set of exposures. On the application side, I work on things related to nutrition epidemiology and population health disparities. I try to understand how the patterns we see in large diverse populations are impacted by different demographics, such as race and ethnicity, gender, geography, and income. Oftentimes if you have a demographic majority in a study population and you run a model, that majority will drive whatever trend you see. So I try to make flexible models that are able to uplift those smaller subgroups and bring them more to the foreground.

Q: What is a recent study where you applied these methods? What was the main takeaway?

A: My primary focus has been in nutrition epidemiology. I recently worked on a study with [nutrition expert] Walter Willett where we were looking broadly at dietary patterns in the U.S. among low-income adult females, and at how those diets might differ based on different demographic factors—in this case, someone’s racial or ethnic identity. We wanted to get a better idea of what that landscape is and how we can better improve their diets, because diet is a precursor for a lot of different chronic diseases, including cardiovascular disease and cancer.

In general, we know that those of lower income tend to have a less healthy diet. But when we parsed things out by different races and ethnicities, we started to see some nuance with that. For example, we saw that people in Hispanic and Mexican ethnic groups consumed a lot of legumes, which are known to be healthy. And we saw that non-Hispanic Asians tend to have a slightly healthier diet quality compared to other racial and ethnic groups.

It’s very easy to make national nutrition policies for everybody, but our study suggested that there’s a lot of nuance that goes into what we’re eating. I’m really hoping our research can inform more targeted, intentional policy to help decrease health disparities by focusing on those populations that have a greater need.

Q: What are you working on next?

A: Right now I’m focused on two projects. The first is a continuation of the nutrition study on low-income women. We are working to develop a method that will identify dietary pattern changes over time and how they associate with cardiovascular disease risk in understudied U.S. adult female populations. We recently received data from the Black Women’s Health Study, which has followed the diets of over 58,000 U.S. Black women since 1995. With the last diet assessment in 2021, we are really excited to see how diets may have changed during COVID-19 and model their subsequent impact on cardiovascular health.

Another project we have been working on in my lab is using our methods to identify how different aspects of the social environment can interrelate and contribute to health disparities. In a lot of analyses that consider socioeconomic status, what ends up happening is that researchers will throw a bunch of factors into one model, say race and ethnicity, income, and education. The problem with that is statistically, you’re saying that each of these factors is independent of one another.

But really, there is an overlap that happens. It’s not just race that might be associated with a health outcome—it could be race as it manifests with poverty, as it manifests with whether you rent versus own a home. What does your daily transportation look like? Is the head of your household a male or a female? What is the language primarily spoken at home?

I want to get a little bit deeper, thinking about all of these other different, intersecting factors that might come into play. The goal is to get a better idea of how we’re defining social determinants, because they don’t all happen in isolation.

Jay Lau

Photo: Suzanne Camarata