Matlin Gilman, PhD ’26, studies the real-world health effects of policy decisions and builds artificial intelligence tools to advance health care
After a decade working in health policy research in universities, nonprofits, and the federal government, Matlin Gilman arrived at Harvard T.H. Chan School of Public Health with a steadfast belief that, as he put it, “rigorous policy analysis and modern data science can be powerful tools for improving people’s health and promoting health equity.”
As a PhD candidate in population health sciences at the Harvard Kenneth C. Griffin School of Arts and Sciences, Gilman has carved a unique academic path in pursuit of turning that idea into a reality. Within Harvard Chan School’s Department of Social and Behavioral Sciences, he has homed in on how to analyze the health effects of policy decisions using advanced statistical methods. And as a cross-registered student at Harvard’s John A. Paulson School of Engineering and Applied Sciences (Harvard SEAS), he has focused on studying artificial intelligence (AI) and machine learning techniques to help empower hospitals and clinicians to improve patient care.
Measuring the effects of abortion bans
When the U.S. Supreme Court overturned federal abortion protections in June 2022, reproductive health care shifted dramatically. Thirteen states banned abortion that year and four more have enacted bans since then. For his dissertation, Gilman set out to measure how these state abortion bans impacted reproductive, maternal, and infant health outcomes in the two years following the Supreme Court’s ruling.
“Access to reproductive care changed dramatically in some states and not at all in others,” Gilman said. “That kind of variation makes it possible to study cause and effect—but only if you can construct a credible picture of what would have happened without the bans.”
To construct this picture, Gilman used Bayesian modeling, a statistical approach that dynamically integrates new data to improve predictions. He built a model that predicted state-level reproductive health outcomes if abortion access had been upheld, taking into account each state’s pre-ban trajectory as well as broader national trends—such as changes in the economy and the COVID-19 pandemic—that the model captured indirectly. “The gap between the model’s predictions and the reproductive health outcomes that actually resulted was the estimated effect of the bans,” Gilman explained.
Gilman found that in states that introduced abortion bans, birth rates increased more than would have been expected without bans, particularly among Hispanic and Black women and women whose highest educational attainment was a high school degree. In contrast, Gilman said, “we didn’t see a significant effect among women with a college degree, which suggests that the workarounds available to some populations—such as expanded telehealth or traveling out of state for abortion care—may be out of reach for others.”
Gilman’s research also found increases in neonatal mortality—deaths within the first 28 days of life—in ban states, driven largely by deaths from severe birth defects. These states also had higher than expected rates of mortality among Black infants. Gilman did not find a statistically detectable effect on maternal mortality, though he noted that maternal deaths are rare enough that small changes are difficult to detect. “The full impact of abortion bans on maternal health may take years to emerge,” he said.
Building AI tools for health care
During his time at Harvard Chan School, Gilman’s focus expanded beyond measuring the effects of policies. He became interested in exploring how complex health care information could become more accessible and useful for the people who need to act on it—whether that’s a hospital trying to understand its own performance or a clinician searching for the best available evidence.
Gilman decided to complement his public health studies with the study of data science. He cross-registered at Harvard SEAS, focusing on machine learning, the architecture behind large language models, and engineering.
For one of his projects at Harvard SEAS, he analyzed Medicare’s Value-Based Purchasing program, which adjusts hospital reimbursements—by as much as millions of dollars—based on clinical outcomes, safety, patient experience, and efficiency. He built a machine learning model to predict whether a hospital would receive higher or lower reimbursements according to their performance, and identified which factors mattered most for that prediction.
“The model shows individual hospitals what to focus on to improve their performance under the program. Patient experience and efficiency drove the predictions more than the other domains,” he said. “That’s especially valuable information for safety-net hospitals, which often face financial instability.”
In another project, Gilman conceived of an AI-powered web application that clinicians could use to effectively and efficiently search through published medical literature to stay on top of current research findings, a gargantuan task as thousands of new studies are published daily. He built the tool with his Harvard SEAS classmates. It uses large language models to understand the meaning of a user’s question, then retrieves relevant studies from an indexed database of medical research and generates a synthesized answer with direct citations. Each cited study is labeled with its conflict-of-interest status, how widely it has been cited, and how recently it was published. Users can filter results on any of these attributes and switch between clinical and research modes so that responses are tailored to their needs.
“The literature is vast, and no one can read it all,” Gilman said. “We built a tool that synthesizes the evidence, links it to its sources, and lets the user decide which studies to prioritize—so the answers are both useful and verifiable.”
Looking ahead
As his PhD nears completion, Gilman is exploring roles in academia, health systems, and technology. His ultimate aim is to use data to help improve health outcomes and make well-being achievable in all communities.
“We have more health data than ever, but data alone doesn’t improve outcomes,” he said. “Someone has to do the careful work of figuring out what it means and making it useful. That’s the part I’m most interested in.”