When you go into Netflix to choose a movie or Amazon to buy a book, they’re ready with proactive suggestions for your next purchase, based on your past history. Isaac Kohane, MD, PhD, would like to see something similar happening in medicine, where today, patients often find themselves repeating their medical history “again and again to every provider,” as Kohane recently told Harvard Medicine.
“Medicine as a whole is a knowledge-processing business that increasingly is taking large amounts of data and then, in theory, bringing that information to the point of care so that doctor and patient have a maximally informed visit,” says Kohane, chair of informatics at Boston Children’s Hospital and co-director of the Center for Biomedical Informatics at Harvard Medical School.
That’s the theory. If it becomes reality, big data could be the next blockbuster in healthcare—one with an enormous impact on how the system approaches medicine. Through it, Kohane believes, doctors will be able to:
- pull in knowledge from the entire population with a disease—including which drugs work for different patient subgroups.
- make more standardized diagnoses: Rather than record one’s impression of a patient’s heart function after listening through a stethoscope, “you can attach a computer to a microphone, and get consistent, reliable diagnoses of which valve is affected,” Kohane told Harvard Medicine.
- get help interpreting patients’ medical history, family history and genomic background in real time, gaining better insight into their health as well as the preferred therapies.
Data on environmental exposures, lifestyle habits, diet and epigenetics (the control mechanisms that decide which genes are turned on or off) could be added into the mix. And all the while, doctors could be feeding information back into the system.
The clinical notes doctors write represent huge reserves of untapped data. Kohane was recently an author on a fascinating paper in JAMIA about using a natural language processing algorithm to capture narrative data from electronic medical records. In simulations, algorithm did a good job at clinically classifying patients with rheumatoid arthritis and coronary artery disease, “two disparate diseases whose diagnosis typically relies upon a complex combination of signs and symptoms, diagnostic tests, and clinician reasoning.”
But for big data live up to its promise, many barriers around sharing health data must be surmounted, even within a single provider system. “If you ask any large health-care system, ‘How many patients do you have with this characteristic? How many patients of this kind did your doctors see? What was their average length of stay?’ they will not know,” Kohane told Technology Review last year.
In Nature Biotechnology last month, Kohane and Kenneth Mandl, MD, MPH, called for a “federalist system” in which research networks combine centralized data-management functions with local authority over datasets. Each institution would be a node that could join a variety of networks, similar to the SHRINE network operated by Harvard’s clinical and translational science center.
“For healthcare to become a data-driven enterprise that learns from itself and drives towards the practice of precision medicine, health systems, hospitals and even provider practices must become instrumented for discovery research and cost-effective federation of data,” they write.