Natural Intelligence: Undergraduate and medical students go beyond the algorithm, use AI to develop tools as part of competition

The prospect of using artificial intelligence in medicine has generated plenty of excitement, but for Neil Sarkar, PhD, MLIS, founding director of the Brown Center for Biomedical Informatics, part of his job is tempering those expectations for health care professionals and students alike.

“We may be excited about AI, but with health data generated in a real-world setting, there is a lot of nuance to it. You have to appreciate that before you try to build a cool new app,” he says.

The inaugural Rhode Island AI for Health (RI-AI4H) Data Challenge + Datathon aimed to give participants—including students, faculty, postdocs, and others—a greater appreciation of those nuances by having them hone their skills in data wrangling and analysis to create tangible, real-world solutions to health challenges across Rhode Island. Advance Rhode Island Clinical and Translational Research co-sponsored the event.

As the name suggests, the event consisted of two parts: a Data Challenge where cross-disciplinary teams used their skills to develop an AI solution, and a Datathon where top-scoring teams got feedback to refine their solutions. The challenge ran from June into July, with several rounds of presentations and feedback, before concluding with a symposium on July 23.

Sarkar, also an associate professor of medical science and of health services, policy, and practice, says participants were tasked with addressing cardiovascular disease and hypertension, and cancer prevention and early detection among the Hispanic and Latino communities across the state. Information was pulled from the SyntheticRI dataset, which contains computationally generated data representing more than 300,000 Rhode Island patients.

Elizabeth Chen, PhD, interim director of BCBI, says SyntheticRI has been incorporated into undergraduate and graduate courses at Brown, and is a major component of the master’s program in health informatics and AI.

“We found it’s a great way to teach students how to work with health data without worrying about the constraints with real data,” says Chen, an associate professor of medical science and of health services, policy, and practice. “It allows them to become familiar with the content and structure of such datasets so they are better prepared for real-world usage.”

Alan Mach '25 MD'29 admits that, prior to competing, he wasn’t knowledgeable when it came to approving health care solutions powered by AI. He and his team developed CardioEquity RI, which identifies cardiovascular patients who may be missing treatments and offers a platform to prepare clinical follow-ups and outreach.

“I’m not too familiar exactly with all the nitty-gritty details of what makes a good AI-based program and what makes it commercializable,” Mach says. “But this challenge taught me about what is expected and what should be accounted for in developing these types of solutions.”

Anna Felten ‘29, an incoming transfer student on Mach’s team, says their research into how to use the dataset produced greater insight into what can make certain AI platforms effective and ineffective. Working on an interdisciplinary team was a major part of their development process, as medical students like Mach provided input from a clinical standpoint.

“We were able to examine our tool from a physician’s perspective and consider how it would be used with patients in the community,” Felten says. “That experience gave me better insight into gaps in patient care.”

There were also questions about protecting the privacy of patients’ data. Felten says their prototype used a third-party application programming interface, so the team had to consider ways to protect the data from outside interference.

“The project raised questions about medical ethics, privacy, and the technical challenges that can feed into AI,” she says.

However, the fact that RI-AI4H wasn’t “simply a technical competition” attracted students like Kelly Yang ‘27. The intersection of data science with clinical practice, public health, and implementation appealed to her and others on the CardioEquity RI team. Using a tool that could examine everything from patients’ blood pressure, medications, and other health information reinforced the balancing act of offering an in-depth tool to clinicians in an ethical fashion.

“The Datathon helped us move from trying to build an accurate tool to trying to design an accountable system around it,” she says.

Another AI solution was CATCH, which stands for Care-gap Alerts for Treating Community Hypertension. Ivan Yu ‘29 says the system uses the SyntheticRI dataset to “identify patients with repeated elevated blood-pressure readings and gaps in diagnosis and treatment,” and triages cases into different priority levels for health care professionals. He says the feedback throughout the challenge led to numerous adjustments for his team’s solution.

“This included differentiating our model to stand out among market competition, ensuring HIPAA compliance per use of protected health information, and refining our target population while incorporating linguistically tailored outreach to make follow-up more accessible,” Yu says.

One of the challenges was narrowing our solution to a specific healthcare need while ensuring that it would support, rather than complicate, clinicians’ existing workflows. 

“We ultimately focused on closing the gap between abnormal blood-pressure readings and appropriate follow-up with an emphasis on supporting federally qualified health centers serving patients who may face linguistic, financial, or social barriers to care. We also had to work through ambiguity presented within the dataset, like distinguishing between a single elevated reading and a persistent pattern or determining what certain information could or couldn't tell us,” Yu says.

Sarkar says the experience gave students a “taste of what health data look like” and he hopes this and future Data Challenges/Datathons will become a breeding ground for future innovations across the state.

“This is a great culmination of things we’re doing across Rhode Island,” Sarkar says.

Yang says the symposium was both energizing and humbling and left an important impression.

“It also reminded us that enthusiasm for AI has to be matched by careful attention to safety, privacy, accessibility, and implementation,” she says.

Rhode Island AI for Health (RI-AI4H) Data Challenge + Datathon Winner

Mala Models on Thayer, Blood Pressure Cascade RI

Christopher Chen ‘29, Alec Zhu ‘29, Annie Wu ‘27, Andres Newman ‘29, Umar Sheikh '29

Runner-up

CATCH, Care-Gap Alerts for Treating Community Hypertension

Wilber Sean V. Anterola '29, Ivan Yu ‘29