Much of the public discourse in health care in recent years has been around population health and the many techniques and technologies that can be employed to glean actionable insights from big data, close gaps in care and more effectively target patients for prevention and care management programs. Organizations that have been able to sift through the hype and develop practical programs are starting to achieve real results with their big data initiatives. One aspect of this new frontier of leveraging the power of large data sets that is less understood is “cognitive computing.”
Often conflated with earlier concepts of artificial intelligence, or AI, and machine learning, cognitive computing is actually a powerful and proven technology that promises to deliver on many of the as yet unmet expectations around health care big data.
Hospital and health system leaders need to build a better understanding of cognitive computing and its applications to be ready to take advantage of this next evolution of big data-driven health care innovation.
What Is Cognitive Computing?
The key characteristics to understand about cognitive computing systems are that they can:
· Understand and interpret natural language
· Extend what humans and machine can do
· Help experts to make better decisions by penetrating the complexity of big data — including structured and unstructured data
We say that these systems “learn” because they are able to take in massive amounts of structured and unstructured data — things like national influenza vaccination rates or every Journal of the American Medical Association article published in the last decade — and use a variety of technologies, including natural language processing, to associate and correlate those many disparate data sources in very complex and sophisticated ways according to their meaning. In this context, the concept of learning also includes the ability to refine the system’s understanding of its data sets, and make continuous improvements to the associations and potential applications of that data based on expert human feedback and ongoing data acquisition.
Another important aspect of cognitive computing is the ability to “understand” natural language. This means that a health care provider with access to a cognitive computing system does not need to know how to structure a database query or construct a Boolean statement; they can just ask a question in a natural manner. That kind of capability is powerful for a few reasons. It not only means that users can launch powerful searches without programming support, but it also allows the system to provide a wider range of response, possibly bringing in information that would not have been retrieved, constrained by a structured query.
Together, the power to learn about massive data sets and to process naturally phrased questions allow cognitive computing systems to make the vastness of big data — including many kinds of unstructured data that would otherwise be impossible to aggregate — available and useful for providers and other health care professionals. Where traditional big data systems have given organizations data-driven insights limited to certain types of data and types of queries, cognitive computing systems yield knowledge-driven insights that complement and transcend structured queries.
Is Cognitive Computing Appropriate for My Organization?
For health care organizations to assess whether this next evolution of big data and insight represents an opportunity to more efficiently improve outcomes and lower costs, there are a few key questions to consider:
Is your organization doing everything it can with the data it has?
The volume of biomedical, clinical, psychosocial, personal and research data available continues to grow at an increasingly overwhelming pace. It is implausible for even the most diligent physicians to keep up with the proliferation of information and, consequently, many providers fail to connect their patients with the best care potentially available to them. If we use cognitive computing systems to give doctors the tools they need to succeed, and empower expertise in every individual caregiver, we can convert information overload into meaningful guidance that allows caregivers to perform at their highest potential.
Where’s the best place to allocate your resources and take waste out of the system?
Cognitive computing also can be applied to the challenge of managing the cost of care, by helping organizations to understand where best to apply limited resources. Getting each complex patient just the right care (and avoid unnecessary care) at just the right time requires a careful balance between the a priori knowledge and the interactions of hundreds of factors — a perfect use for cognitive computing. When you have a system that can provide decision support based on intelligent analysis of all of those elements, and can collect and analyze data on which interventions and pathways are most effective, it becomes far easier to meet the demands of tightening margins in the setting of new value-based payment models.
What’s your strategy for limiting variability of care?
Good doctors know that variability is not always a bad thing. It not only identifies barriers to adopting guidelines, but also can help to identify (and predict) outliers that are performing exceptionally well and incorporate those insights for the benefit of the entire population. When variability does occur, cognitive computing can provide a more nuanced understanding of the factors leading to those decisions and help to make smarter recommendations in the future.
Do you need a cognitive computing strategy?
We believe that in the near future, cognitive computing systems will touch every aspect of care. Think about how your organization manages knowledge and assimilates the variety of data types and sources to deliver useful insights to your doctors, nurses, caseworkers, administrators and more. Knowledge-driven insight can complement your data-driven approach to comparative effectiveness and optimal personalized care plans.
The bigger question at hand is: What form of relationship do you ultimately want with your data and the information that exists in your organization? Cognitive computing represents an opportunity to create a living and learning system that gives clinical, financial and operational leaders ever more accurate and useful insights that enable everyone to perform better. Teaching a computer system to be smarter and better at supporting the work you do is an ongoing learning and transformational journey, a long-term investment in leveraging the insights from your systems and others’ and then disseminating that wealth of knowledge for better care.
Anil Jain, M.D., is senior vice president, chief medical officer and co-founder at Explorys, an IBM Co. Kathy McGroddy Goetz, Ph.D., is vice president, partnerships and solutions, at IBM Watson Health.