Dr. Christopher Phillips on the Histories of Statistics & Data in Medicine
On March 17, our podcast hosted Dr. Christopher Phillips, a Professor and Historian of science, medicine, and statistics Carnegie Mellon University—and also a member of our Seminar! Beginning in the Fall of 2019, Dr. Phillips joined in on our public events and Friday lunchtime sessions. On our podcast interview, he shared how joining the Seminar’s interdisciplinary conversations about data and (reference intended!) information ecosystems has revealed the need for and rewards of approaching the same topics from distinct disciplinary and methodological viewpoints. And during our chat, I was alerted over and over to how valuable a historic approach to understanding science is. So often, we view STEM fields and workplaces as intrinsically separate from, and thus competing against, the humanities. This perceived divide has real-world consequences, among them the myths of STEM disciplines as ahistorical or apolitical, and the ultimately dangerous devaluing and underfunding of humanities programs.
But Dr. Phillips’ work stands as a testament to the very real insights to be gained from a historical approach to math, science, statistics, and medicine. His current research focuses on the long histories of precision medicine and statistical approaches within. In the wake of the ongoing COVID-19 pandemic, the concept of precision medicine has come under renewed scrutiny. Precision medicine proposes that medical practices ranging from decisions, diagnoses, treatments, and products can be tailored to precise subgroups of patients—taking into account their genetics, environment, and lifestyle, rather than a “one size fits all” approach. For many of us, COVID is the among the single greatest public health crises we’ve witnessed in our lifetimes—and so it makes perfect sense why precision medicine is an attractive medical model: patients hoping to avoid infection or complications want their doctors to outline a precise approach, which takes into consideration their unique and individual risk factors. So how does precision medicine become precise? In very recent years, “precision” has often come to mean “data-driven” or “statistical”; when your doctor gives you a diagnosis or prognosis, they are not only considering you and your body, but also the group of patients whose biomarkers—any of a broad range of things happening in your body, like resting heart rate or X-Ray findings—match yours. And in some ways, this statistical approach is truly brand-new.
Dr. Phillips shared on the podcast that “for most of the history of medicine, statistics has actually been fairly irrelevant to clinical medicine, and that’s because when you go to the doctor you don’t want average advice, right, you want advice for you.” Indeed—and while many patients might not actually know that their prognosis has anything to do with massive data sets of strangers’ genetic profiles, there are real pharmacogenetic reasons why a patient, risk-adverse or not, would absolutely support this approach. And although Dr. Phillips shared that this next best thing in medicine isn’t necessarily all that new, there is something truly novel at play here: how much decision faith we’ve collectively placed on the 21st century’s brand of big data.
As Phillips points out, statistical analysis and medical doctors have worked hand-in-hand for centuries. Additionally, long before “big data” was even a concept, doctors in the mid-20th century were using statistical methods to aggregate, generalize, and then apply information about increasingly large populations of patients. So the fact that these precision methods have verifiable histories should provide a necessary dose of skepticism about recent claims that big data can surely cure the incurable, or that we can’t beat COVID without it; while doctors are rightfully trained to strive toward error-free excellence, history tells us that medicine is wrong, imperfect, and even violent, just as often as it gets things right. Plenty of medical and scientific advances that we now know to be poisonous, lethal, or otherwise harmful were used widely and even awarded prestigious honors , as recently as just a few decades ago.
But far from a warning to mistrust all doctors and scientists, Dr. Phillips instead wants us to understand that there is no single scientific advancement that can cure all of our medical woes—including big data’s rapid entry into precision medicine. Data seems, so often, to simply tell us the truth. But our Seminar guests tell a whole different story; from medicine to social media, and from cartography to topic modeling, data virtually never reveals much certainty. And in a 2019 article on data in baseball, Dr. Phillips put it plainly: data that is stable and reliable is an accomplishment of labor, infrastructure, and volunteerism—not a natural state of the data itself.
In order to bridge the perceived gaps between the humanities and STEM, it is essential that we re-evaluate math, science, and statistics—and big data—as deeply historically and socio-politically situated. The field of History tells us exactly this—science, numbers, and statistical analysis are utterly informed by the world they’re in, whenever and wherever that is. On the podcast, Phillips shared that the Seminar has been helpful for maintaining this sort of helpfully skeptical eye; only amid a community of diverse scholars can we accrue information about a single topic from a wider array of viewpoints. Science and math are historical, sure, but in that history we must also consider economics, politics, popular culture, fine arts, and indeed, how that history bears on our contemporary.
So if you’re afraid that you’re “not a math person,” “not a science person,” or simply “bad with numbers”—a set of myths I certainly told myself over the course of my K-12 education—remember that a lot more goes into numbers than just the numbers themselves. On our podcast, Phillips remarked that in the context of high school and college Science and Math classrooms, teachers can position “historic development as the way to actually get people interested in the material.” In other words, if the precise mathematical concepts of, say, falling objects are not appealing to a student who is so-called “bad” at math, a teacher might first introduce them the story of Galileo’s Leaning Tower of Pisa experiment. Amid the blazing speed of 21st century big data, we must remember our (very) long shared histories, which confirm that there are so many ways of knowing.