If you ever feel the urge to play amateur archeologist and prowl Duke University’s Davison Building, built in the 1930s, you will find the remnants of the original Duke Hospital.
Peer down an anonymous corridor of offices, and the stretcher-height stainless steel lining the walls and stretcher-width doors bear witness to their past lives as operating theaters. Peek through the square window in a door, and it's quite possible that you’ll be looking into the operating room where in 1968 William Sealy was the first person to surgically treat an aberrant cardiac electrical pathway called “Wolff-Parkinson-White syndrome.” Elsewhere, between the former Osler and Long wards on the second floor, was the medical intensive care unit, and on the third floor across from the former Cushing ward was the surgical ICU.
It was in these units during the late sixties and early seventies that Frank Starmer, an electrical engineer and a founding faculty member of Duke’s Computer Science Department, worked with Department of Medicine Chair Eugene A. Stead, Jr. and other faculty to build machines, each the size of a small refrigerator, that captured data from critically ill heart attack patients.
These forgotten appliances, these corridors of history, represent Duke’s long and continuing tradition of using data to solve clinical problems.
Dr. Stead was a visionary who saw clearly the potential that networked computers had for enabling new approaches to patient care. A digital adjunct to the patient’s chart, dubbed the “Prognostigram,” would compile all the data Duke was collecting about patients with heart disease and run calculations to help clinicians project what would happen if they chose among different clinical interventions. Yet, nearly 5 decades later, Dr. Stead would be surprised and saddened that his vision of a digital chart that constantly and pervasively assimilates data and applies computational power to assist clinical care is still an aspiration rather than a reality – ubiquitous electronic health records notwithstanding.
Today, Duke Forge sits squarely amid a warren of hallways and history in the old Davison Building, where our faculty, staff, and partners are steadily hammering away at the raw material of health data in the hopes of realizing Dr. Stead’s dream. As a group of people devoted to “actionable health data science,” we obviously love technology. The fact that we can fix the aorta of a patient with an aneurysm by sliding a synthetic graft through a catheter to then open like an umbrella is astonishing. No less astonishing is the ability to write a handful of lines of programming code that can turn thousands of variables plucked from data contained in insurance and health records into an accurate prediction about whether a patient will be hospitalized in the next 6 months. But what drives us is no different than what animated those former operating theaters, wards, and ICUs down the hallway—it’s just that our tools have evolved.
When I first donned a white coat as a fledgling physician, what motivated me was not the prospect of merely staying out of trouble. Rather, it was and continues to be helping fellow human beings with the best information and best instruments at hand.
However, it can sometimes happen that a focus on tools and how to use them risks distracting us from essentials. For those of us who work intensively with health data, a disproportionate focus on security and minimizing risk—the negative goal of avoiding mistakes—can crowd out the positive goals of helping patients. Now, it’s absolutely true that the fiduciary and ethical obligations that govern how we work with health data are no different than the Hippocratic Oath that binds doctors to “do no harm.” But when I first donned a white coat as a fledgling physician, what motivated me was not the prospect of merely staying out of trouble. Rather, it was and continues to be helping fellow human beings with the best information and best instruments at hand.
In other words: while how we help someone is important, it’s far less important than the basic motivation for caring in the first place.
In the world of applied healthcare, we must remember that technology and data science, as well as the various systems of governance and oversight that overlay them, are only means to an end and not ends unto themselves. We also need to remember that arguments about machine learning methodologies or “container orchestration” platforms (arguments that can approach religious levels of fervor) must always be in the service of obtaining the best outcomes for the people we care for and the frontline clinicians who are doing their best to care for them.
Here’s an example: Susan Jones (not her real name) is a woman in her late sixties with multiple chronic conditions, including diabetes, high blood pressure, congestive heart failure, and obesity. She has difficulty walking, and coming to a doctor’s appointment is a challenge. She also has little help at home. Over the course of the past year, she has been hospitalized five times.
Recently, a cross-disciplinary team from Duke Connected Care, the Duke Institute for Health Innovation, Duke Health Technology Solutions, and Duke Forge built an artificial intelligence platform called “Deep Care Management” to help patients like Ms. Jones. It uses a kind of machine learning called a deep neural network to predict a patient’s chance of hospitalization over the next 6 months for several dozen possible diagnoses. In Ms. Jones’s case, Deep Care Management notified our complex case management team that she has a 97% chance of hospitalization. The team was then able to rapidly mobilize in response to provide her with personal health counseling, a dietitian, physical therapy, and transportation assistance.
A year later, Ms. Jones has had just one visit to the emergency department. She’s also lost 68 pounds, her probability for hospitalization has dropped to 81%, and she now reports that she’s “feeling much better” to her primary care doctor.
Harkening back to the vision that Dr. Stead first articulated in the early 1970s, we now have the potential to build the hundreds or thousands of “prognostigrams” that the health system of the future should be running to proactively support health and to deliver the right care, to the right person, at the right time. Achieving this goal will fundamentally change the information architecture we use to deliver health. But as we work to design and deploy this architecture, we can’t afford to get lost down technological rabbit holes or be paralyzed by indecision while seeking a path forward without risk. If we are to truly serve our community, it’s “Susan Jones” and those like her who must be at the center of all that we do.