How Precision Health Is Transforming Medicine
After decades of investing in achieving the lofty goals of so-called precision medicine, the field of medicine has realized that we may have been targeting a necessary, but not sufficient definition of success. “Precision health” is a far more precise term to describe our evolving efforts to use big data to tailor individual treatments for each patient.
This change is more than semantics. It reflects one of the key insights that have emerged during the last few years of intensive research and planning: Health care is no longer completely defined by the always essential relationship between medical professionals and their patients. Increasingly, patient health is being safeguarded and improved by a broad range of experts who are collaborating on the ever-growing array of questions that involve not just medicine but the social, psychological, and physical health of patients.
This is the idea behind the new Precision Health Initiative at the University of Michigan. It involves 19 schools and colleges across the college, including several historically under-connected to medicine, such as law, social work, and literature. It is being led by leaders from three distinct schools, Public Health, the College of Engineering and, of course, Medicine. This broad collaborative is similar to efforts underway at Stanford University, Indiana University, the University of Washington and the National Institutes of Health’s one million-person precision medicine initiative.
We need the best minds from a variety of disciplines because the challenges of modern health care far exceed the skill sets of doctors. At its core, precision health involves drawing on the health records of tens of thousands, and one day, millions, of patients to find the best course of treatment for each patient.
A data-driven approach for health today involves mind-boggling amounts of data. So much information is in play that it is already challenging the ability even of computers to store and analyze it. A single human genome takes up about 100 gigabytes of storage space; just 10,000 patients would require millions of gigabytes. And the genome is just one data point. The promise of precision health depends on collecting far more information from each patient — including their social situation, economic resources, family and medical history, diet, exercise routine, and exposure to environmental factors, to name just a few.
Indeed, the unique genetic sequence that each of us possesses does not define who we are, it defines who we can be. The interaction of the environment with each of us can determine which of our genes are “turned on” and which are muted. For one person, exposure to cigarette leads to lung cancer. For another, perhaps not. Paradoxically, recent breakthroughs have led us to realize that we do not understand the full health impacts of every aspect of our environment — whether air quality or diet or socioeconomic status — which may, in turn, influence gene expression.
The goal is to replace our a-few-sizes-fit-all approach — if you have heart disease or melanoma, for example, here are the two possible treatments — with a more tailored approach that creates subsets of similar patients to determine the best options based on a myriad of factors. This will allow us to use data-driven, evidence-based treatments to replace two of the key biases that now affect health care choices: the biases of the organization (a health center that specializes in bone marrow transplants is more likely to recommend that as treatment for leukemia than chemotherapy), and the anecdotal experiences of the doctor.
The more data we have, the more precise we can be — but only if we can sift and sort it to find the patterns hiding in that mass of information. That is why we need non-medical experts, including computer scientists and biostatisticians and social scientists, to analyze and interpret it. In fact, we are confronted with so much data — made even more complicated by the fact that people and their environment are not static but ever changing — that precision health will increasingly depend on advances in artificial intelligence as only machines will be able to make the key connections in the mountains of data.
Such machine learning is already paying dividends. For instance, each year an estimated 500,000 patients in American hospitals develop a dangerous intestinal infection called Clostridium difficile, or C. diff. A team of researchers produced an algorithm that assesses a variety of factors — about 10,000 factors per patient, per day, including patients’ specific medication history, their location in the hospital, length of stay, vital signs, lab test results, and more — to determine who is most vulnerable to contracting C. diff. It has already in use and saving lives.
A similar effort is underway to determine who might be at greatest risk for developing an addiction to opioids, which often starts with a legitimate prescription for pain medication.
Despite these promising developments, we must resist the natural urge to oversell the current effectiveness of precision health. It is not yet vastly increasing the range of treatments for illness. Instead of producing miracle cures, it is helping caregivers identify which available treatment might be more effective for particular patients, such as which patient, for example, might better respond to a bone marrow transplant rather than chemotherapy for leukemia. It is allowing physicians and patients to have more open and transparent conversations about the benefits and drawbacks of various treatment options.
While still in its infancy, precision health is a milestone in human history. By harnessing a wide array of human and technological genius to the complex question of well-being, we are coming ever closer to realizing the goal of medicine which has guided every physician since Hippocrates: to deliver the right treatment at the right time to the right person.
Marschall S. Runge, MD, PhD, is Executive Vice President for Medical Affairs and Dean of the Medical School for the University of Michigan.
