Dr. Riskin is the CEO of Health Fidelity, a leading provider of natural language processing solutions. He is also a Consulting Assistant Professor of Surgery at Stanford University and practices one day a week out of the Stanford affiliate hospitals. I recently had the opportunity to get his thoughts on how data-driven innovations are transforming the health care industry.
Castro: In what ways do you see data changing health care today?
Riskin: Data is used daily to define a new generation of healthcare. Not only do patients do research on the internet, request medical support by e-mail (in some systems), and share their own medical stories online, but the actual care delivered now includes apps and remote technologies that offer supplemental care.
The most fundamental change related to healthcare is the redefinition of practice, often known as data-driven healthcare. Data-driven healthcare is a big data approach to healthcare, leveraging information learned from treating millions of patients to personalize care for the few. This turns a half century of medical practice using evidence based medicine on its head. Instead of defining care for millions based on a randomized trial performed on hundreds, an individual can have their care based on what worked in their situation seen in the broader population.
Castro: How do you think data-driven health care is evolving?
Riskin: The earliest manifestation of data-driven healthcare was care determination based on manual patient registries targeted to specific conditions. For example, a trauma registry was kept manually at Stanford for years, allowing publication of unique insight on a few specific variables.
More recent efforts have focused on using automated technologies to leverage unstructured clinical data, the 80% of healthcare content that resides in narrative notes in the electronic medical record and has previously been unused for analytics. Over the last two years, we have seen as many publications using automated data technologies to define care from unstructured data as in the previous 20 years put together.
The changes we should expect and demand as data influences care include: 1) Use of large data sets to enhance standard of care where possible and define standard of care where it was previously undefined, 2) Population based health, supporting preventative care and tracking of diseases to improve overall health in the community, and 3) Defined and published quality determination to assure areas of weakness are addressed and the entire system strives to improve care.
Castro: What kind of impact will data-driven health care have on the average patient?
Riskin: The average patient will see little change over the next five years beyond the fact that when they visit a doctor, their information is now being entered into computers. Over time, they will see a phasing in of better access to care, including e-mail access to nurses and doctors, improved mobile support for health, and better and clearer information on quality of hospitals and individual practitioners. In the 10-year time horizon, I believe patients will actually see improvement in quality of their care based on data-driven approaches, heavily influenced by frequent local, regional, and national quality improvement programs.
Castro: When should we start to see some of the more revolutionary uses of data in health care?
Riskin: The most revolutionary use of data in healthcare will require three factors: Availability of processing power to address big data, availability of large aggregated data sets, and availability of technology to fully leverage these data sets. The first challenge, processing power, has been addressed over the last decade. The next challenge, aggregated data sets, is a political challenge both at the national and local levels. While deidentification and aggregation of data is technically possible (if challenging), the political will to share data from hospital to hospital and within regions is only slowly building. We will need to see how health information exchanges and responses to national requirements progress before we can determine when we can achieve secure control of aggregated robust clinical data sets. The final challenge of developing technology to fully leverage these data sets requires powerful systems to make unstructured data usable (natural language processing and ontologies) and data mining technologies to derive meaning from the usable information. These technologies are just now mature enough for initial use, are being integrated into high end practices, and should evolve over the next several years to offer rapid expansion of data driven healthcare capabilities.
Imagine how Google revolutionized the internet. First, a basic information set was available as of the late 1990s. Then, Google provided tools to make the information usable and meaningful which lead to a virtuous cycle that brought increasing tools and information into common usage. Assuming that security, privacy, and statistical bias issues are addressed, this provides a highly desirable vision for healthcare.
Castro: What are the major challenges to more use of data analytics by doctors?
Riskin: At the moment, increasing use of data analytics is far from the thoughts of the average doctor. In my practice, almost all the doctors I know love technology and embrace any technology that makes their lives easier and the care they offer better. Unfortunately, the last five years have been rough on doctors. The average doctor has been forced to adopt complicated electronic medical records systems where a data entry clerk can use as many as 50 clickthroughs and dropdowns to document any given encounter. Additionally, doctors are required to follow new workflows that feel artificial within usual medical care. The combination of these factors make the last few years the most painful era in healthcare technology. The argument is that effort creates infrastructure for future data-driven healthcare. I agree, but it doesn’t make the process painless.
Over time, the system needs to be easier and more fun. Think of the early text messaging cell phones, where a person had to tap a single key three times to type a single letter. It was barely usable. Then, mini keyboards and tap screens made texting and e-mailing easier and texting became commonplace. When cell phones started offering natural input, a user could choose voice, tapping, or typing, whichever was preferable. This makes the process so enjoyable that even I rarely type a text message any more.
Healthcare data technologies have to reach this level to make doctors not begrudgingly use them, but actually embrace them. I want to use natural input to document encounters and receive data analytics on patients, health systems, and populations. The information I receive needs to be meaningful. This can’t be an internet search from the mid 1990s. It has to be relevant, meaningful information that improves the level of care I provide. This has to be the goal of all next generation healthcare IT vendors.
“5 Q’s on Data Innovation” is part of an ongoing series of interviews for Data Innovation Day by ITIF Senior Analyst Daniel Castro. If you have a suggestion for someone who should be featured, send an email to Daniel Castro at firstname.lastname@example.org.