When Purity Mugambi saw red balloons in the UMass Amherst library, she didn’t expect it to change the course of her PhD.
At the time, she was “minding [her] own business,” as she puts it. But the woman holding that event was leading a session for women on how to advocate for themselves in clinical settings—how to talk to doctors in emergency rooms, what to ask for when they receive medication, and what to do when they don’t feel heard.
“I was like, why would you need to do that?” Purity remembers. “She told me the history of how women with heart diseases have been misdiagnosed, mistreated. I was stunned.”
As a computer scientist already interested in algorithmic bias, especially in language and facial recognition systems, Purity had been looking at the ways machine learning models fail to serve all users equally. That conversation at the library shifted everything: from bias in computation to bias in healthcare. From abstract algorithms to urgent, real-world inequalities.
She brought the idea to her advisor, Dr. Madalina “Ina” Fiterau. That conversation set off a series of connections that would eventually grow into a fully interdisciplinary research team: Dr. Rae Walker, a nursing professor focused on health equity; Dr. Stephanie Carreiro, a physician-scientist at UMass Medical School with expertise in emergency medicine; and Dr. Joohyun Chung, a nursing professor with a background in biostatistics.
Those early collaborations weren’t always easy. “The paper we initially submitted was completely off,” Purity admits. “Coming from computer science, we wanted a really complex model—but in healthcare, people care more about interpretability.”
Her new collaborators pushed her to think less like a computer scientist and more like a clinician. “Dr. Walker explained that we shouldn’t be looking at differences ‘by race’ itself, but at how racism leads to different treatment. That completely changed how I thought about the problem.”
The critiques weren’t rejections—they were invitations to grow, and they demonstrated the power of interdisciplinary collaborations and conversations in sharpening research questions and approaches. “My collaborators weren’t just telling me how to think in different fields,” Purity says. “They were teaching me how to be a more well-rounded researcher. That’s the best gift I’ve gotten from this project.”
Together, the team shaped a project that is now in its sixth year—one that is reshaping how clinicians, researchers, and engineers think about bias in healthcare. And thanks in part to support from an IDS Team Grant, Purity has taken the work from promising concept to interdisciplinary, nationally recognized research.
A Hidden Pattern in the ICU
The team’s focus is heart attacks. It’s one of the few conditions where treatment should be relatively standardized across cases: the symptoms are acute, and the interventions are well established. But the initial dataset that the team examined was clear: even in this controlled space, disparities are emerging.
“Prior research had showed that the way women present with a heart attack is different from men,” Purity explains. “So, during triage in emergency rooms, women were being misdiagnosed and sent home—even when they were experiencing a heart attack. This could be one reason why mortality rates are often much higher for women than men.”
Their project uses large, anonymized datasets to analyze how patients in the ICU are diagnosed and treated during cardiac events, and whether those treatments vary based on factors like sex, race, age, insurance status, or region. Because the data is anonymized, the team relies on proxies for certain factors, like insurance type as a stand-in for socioeconomic status.
But the real innovation lies in the scale of their datasets—and how they use them.
“We don’t just analyze one dataset. We’ve developed tools that can compare findings across multiple datasets,” Purity says. “That way, we can identify patterns that aren’t specific to a single hospital or population—but signal something broader across the system.”
Over time, they added more collaborators—clinicians like Dr. Michael Sherman and his fellow Dr. Patric Gibbons, who brought real-time ICU expertise to the project. They expanded their datasets to include institutions from across the U.S. and even the Netherlands. They refined their models, limited confounding variables, and learned to account for regional and institutional variation in treatment.
And once they added datasets from across the country and the world, a fuller picture emerged.
“When we expanded it, we saw major disparities in race and region for things like treatment and pain,” Purity says. “Folks in the U.S. Midwest and Northeast were treated better than in the U.S. South. In some datasets there weren’t big racial disparities, but there were sex disparities. In Amsterdam, which is more homogenous and has universal healthcare, the socioeconomic disparities were harder to identify, but sex-based differences still showed up.”
Those findings are now being compiled into a journal article—a major publication from the study, though not the team’s first.
From Seed to Scale
The IDS team seed grant arrived at a crucial moment. At the time, Purity was working as a teaching assistant, splitting her time between teaching responsibilities and research. The grant allowed her to devote herself fully to the project.
But while not having to split her time was a game-changer, it wasn’t the most transformative part. Dr. Fiterau insisted that Purity use part of the funding to travel and present her findings at conferences, such as the Society for Critical Care Medicine and the European Society for Intensive Care Medicine. Two of the field’s most influential venues, these experiences proved invaluable.
“As a computer science person talking about disparities in healthcare, my work was hugely well-received,” she says. “That made me feel like it was meaningful for the good of the world. It was affirming, as a young and upcoming researcher, that other people thought my work was valid and sound.”
That validation helped her weather early doubts. “There were a lot of people that doubted me. I don’t like putting myself in the public eye,” she adds. “But promoting this has been great, even if a bit uncomfortable.”
A Researcher with Range—and Roots
Born and raised in Kenya, Purity earned her bachelor’s degree in computer science from the University of Nairobi. After graduating, she joined IBM Research – Africa—an uncommon opportunity, made possible when IBM opened a lab in Nairobi just as she was finishing her degree. There, she worked on projects tied to Ebola response, water resilience, and primary care access.
What inspired her most were the projects that combined computing with broad social impact.
“That’s always been what has inspired me,” she says.
When she came to the U.S. for her PhD, she brought that passion with her. But it hasn’t always been easy.
“In Kenya, I never felt underrepresented. I didn’t feel weird being a woman in computer science,” she says. “I didn’t start to feel weird until I came to the U.S.”
It Takes a Village
Even so, Purity has been lifted by a community of mentors—many of them women, many of them from disciplines outside her own. She is quick to emphasize that none of this happened alone. “It takes a village,” she says more than once.
Her project reflects the best of that: interdisciplinary collaboration, patient-centered design, impact-anchored research. Purity is about to finish her PhD. But with such a remarkable research experience under her belt, her promising academic and professional life is just getting started.
And for us at IDS, Purity’s journey shows exactly why the team grant program exists—to help students stick with STEM and plant the seeds so that they learn the skills and develop the community and team to conduct cutting-edge research that tackles big social problems that are game changers.