Howard Garrison Advocacy Fellow
Alyssa Columbus
Alyssa Columbus is a graduate student at Johns Hopkins University.
Describe your interest in participating in the program:
Columbus: Most of my applied work sits downstream of federal money. The National COVID Cohort Collaborative, where I’ve been a consortium member since 2022 and a coauthor on several papers, exists because public agencies paid to assemble it, and the Global Burden of Disease estimates I help produce with the Institute for Health Metrics and Evaluation have been cited in 177 policy documents from bodies including the World Health Organization, the Council of the European Union, and the World Bank. I’ve watched those numbers shape what governments fund. What I haven’t done is help make the case in Washington for the infrastructure and statistical capacity that let work like that exist at all, and that’s the gap I’d like this fellowship to help me close.
I’d advocate for two things. First, predictable support for shared biomedical data resources, since they take years to assemble and lose value quickly once maintenance funding lapses. The second is the analytic capacity that makes them usable, which is where my own research lives: two capable analysts given the same cohort and the same question can produce defensible answers that differ, and knowing when an analysis is finished, and how much its conclusion depends on the path taken to reach it, is the center of my dissertation in trustworthy data science. That capacity costs a fraction of what generating the data costs, and I think the argument for it hasn’t been made often enough to the people who write the budgets.
I’ve had generous preparation for parts of this. The Research-to-Policy Collaboration at Pennsylvania State University selected me as a Data-Driven Policy Voices Fellow, the American Statistical Association sponsored my place at the American Association for the Advancement of Science (AAAS) Catalyzing Advocacy in Science and Engineering (CASE) workshop, and the Federation of American Scientists invited me to write a policy memo on open science and data privacy. As a Center of Excellence in Regulatory Science and Innovation (CERSI) Scholar with the U.S. Food and Drug Administration, I’ve worked on how to report analytic variability so regulators can use it. However, every one of those was a single episode, and I’ve never carried anything through a full cycle from a Hill meeting to a district follow-up to a published op-ed. I’d come in with two concrete aims: to place at least one op-ed on research funding in a Maryland or Southern California outlet, and to bring what I learn back to the Association for Molecular Pathology and to the statistical societies I serve, where methodologists are underrepresented in advocacy relative to how much of the research enterprise we touch. Ten months alongside people who’ve already made that move, including the discussion sessions with Howard Garrison, is how I'd like to learn it, and I'd expect to learn as much from the other fellows, who’ll be closer to the bench than I am. I hope to keep doing this work long after the cohort year ends.
How do you plan to use the knowledge and experience gained through your participation in the Howard Garrison Advocacy Program?
Columbus: The most useful thing I could do with this training is teach it. Advocacy is rarely part of a biostatistics curriculum. I’m on the faculty market this cycle, so I’d like to build it into the group I lead from the beginning: a session in my methods course on how appropriations and authorization work, and standing encouragement for students to write for audiences outside the field, which is the sort of thing most of us pick up a decade in, if at all.
Second, I’d bring it to the societies I belong to. The Association for Molecular Pathology works on questions of evidence and test performance where a methodologist has something specific to offer, and my work as a Center of Excellence in Regulatory Science and Innovation Scholar with the U.S. Food and Drug Administration focuses on how analytic variability should be reported so decision-makers can use it. Seeing how a federation reconciles priorities across member societies would help me contribute there as an advocate and a methodologist, and to do the same in the statistical societies I serve.
Third, I’d keep writing and start placing. I’ve written for Forbes, Significance, and my school’s magazine, and coaching on op-eds and letters would help me reach readers in Maryland and in Southern California, where I grew up. I'd most like to keep the congressional relationships, since they take longer to build than any piece of writing.
Using no more than 250 words, describe your research as you would to a non-scientist.
Columbus: Two years ago, headlines reported that roughly three-quarters of American adults were already overweight or obese, and that the share could pass 80 percent by 2050. I was one of many coauthors on that study. What readers didn’t see is that the estimate rested on hundreds of small decisions: which of 134 data sources to trust, how to fill gaps for states with sparse data, which model to project forward. Each was defensible. Together they moved the number.
My research is about the gap between how certain a finding looks and how certain it is. Hand the same data and question to ten capable scientists, and you’ll often get ten answers that don’t agree, and none of that disagreement reaches the single figure a journal prints. I build software that makes it visible by running an analysis every reasonable way at once and reporting how much the answer moves.
I also work on a question analysts rarely ask out loud: how anyone knows when an analysis is done. That call usually comes down to habit, a deadline, or a supervisor’s judgment, and I’m developing more principled ways to make it, drawing on the mathematics of knowing when to stop looking and commit.
This matters where the numbers do something. Health agencies use burden estimates to decide what to fund and whom to screen, and a figure that arrives with no account of how firm the ground beneath it is asks for more trust than it has earned.
Briefly describe any past or present participation in additional career exploration activities, experiences, and/or programs.
Columbus: I’ve spent a decade testing where a statistician’s work fits outside a university, and the exploring has shaped what I want next as much as any course did. Before my doctorate, I was a data scientist at Pacific Life and then an information security analyst and data governance specialist at SchoolsFirst Federal Credit Union, where I learned that privacy and governance rules decide what can be studied long before anyone reaches the analysis. I later consulted in strategy and analytics at Deloitte, took part in McKinsey and Company’s Insight Program in 2025, and since 2024 have worked as a freelance expert on artificial intelligence safety and governance for several frontier laboratories, which showed me how technical evidence gets weighed when policy is moving faster than the evidence is.
The policy side has drawn me most. My Fulbright year included a seminar on the European Union and NATO, organized by the Fulbright Commission in Brussels, where I sat with officials who make decisions on evidence they have no time to verify themselves. I’ve reviewed grants for the Chan Zuckerberg Initiative, the National Institutes of Health, and two National Aeronautics and Space Administration open science panels, written examination questions for the National Board of Public Health Examiners, and completed the National Science Policy Network’s Communicating Science for Policy certificate.
I try to pass this on by mentoring for the United Kingdom Research and Innovation Policy Fellowship and speaking on career panels at universities and for the Girl Scouts.
Alyssa Columbus is a member of Association for Molecular Pathology.