When Data Science Meets the Real World

By Ursula Rogers

If you’re like most people, you probably don’t spend a large chunk of your waking hours thinking about medical data. I do, but then, it’s my job. As Senior Informaticist for Duke Forge, I live and breathe medical data: how to find it, understand it, extract it, clean it, combine it, and most importantly, make it useful for answering questions. But whether you’re obsessed with it or indifferent to it, medical data—and just as importantly, the ways we think about and handle that data—can have profound effects on our lives.

I received a first-hand lesson in this when my daughter was born with serious health challenges. A problem with her airway required a long-term stay in the hospital’s pediatric intensive care unit, as well as multiple surgeries and intensive home care. During this time, it might have looked from the outside like my daughter was getting great treatment. We were at a world-class hospital and we had caring and competent support from individuals on the medical team.

But there was a catch.

In fact, there were a lot of them. There were significant errors in my daughter’s medical record. Information wasn’t being shared effectively across providers on her team. Bad data “stuck” to the medical record even after it was corrected. As my daughter passed from one medical crisis to another, these data issues (and the confusion they caused) snowballed to the point where I had to leave my job because managing my daughter’s care required my full-time attention. Eventually, my husband and I took the extraordinary step of seeking care at a different institution, one several states away from our home in North Carolina.

In a way, our experience was a microcosm of shortcomings that exist across the entire healthcare system. Almost everywhere you look, you see similar problems. Patient care is atomized and often involves multiple specialists who may lack tools for effective communication. Medical data are collected in electronic health records that primarily serve the needs of payers instead of those of patients and providers. And as we learn to apply cutting-edge approaches such as machine learning, we’re increasingly aware of the gap between a patient’s actual condition and what the data seem to be telling us about that condition. These challenges are especially frustrating because at the same time, we’re sitting on an unbelievable wealth of information that could empower patients and physicians, advance research, and improve care—if only we can figure out how to fit all the pieces together.

Fortunately, I’m in a position to do something about it. At the Forge, our unofficial motto is “free the data.” We know that tackling big health challenges means that we have to innovate and build new tools for working with data. We have to match pressing medical questions with the right information and ensure that data can flow freely to and among diverse teams that can apply their experience and insight to solving problems. And ultimately, we have to figure out how to pull together a fragmented story that’s scattered across multiple screens of an electronic health record and turn it into a coherent and true narrative.

No patient or family should ever have to worry about whether bad data might be steering medical care in the wrong direction, or whether good data is failing to reach a care provider who can act on it.

We can do better. We have to do better.

This blog post was originally published April 4, 2018 on the Duke Forge Website.