Too often, healthcare is compared with other industries when it comes to leveraging data to improve outcomes and enhance engagement. But what’s often missing from the discussion is a very important observation: “In banking and retail, data is a required ingredient of the process. In healthcare, data is still a byproduct of the process,” according to Omid Shabestari, MD.
It’s a critical piece in the puzzle — at least, for now. In a recent healthsystemCIO.com webinar, Shabestari, who is Director of Health Analytics at Carilion Clinic, discussed the enormous potential of advanced analytics solutions, what it takes to establish a solid ecosystem, and what he expects to see in the coming years.
Dr. Shabestari kicked off the webinar — Designing the Framework for a Cognitive Computing Ecosystem — by providing some clarity around cognitive computing, a term often misused interchangeably with artificial intelligence. Defined by IBM as systems that “learn at scale, reason with purpose, and interact with humans naturally, cognitive computing uses elements of AI, machine learning, natural language processing, network networks, and reinforcement learning to simulate the human thought process.” In simpler terms, it enables computers to gain experience as they’re exposed to the results of their recommendations, and ideally, make better decisions.
“There’s a tremendous amount of data coming to us,” noted Dr. Shabestari. “We need to be able to leverage it to provide good predictions and act upon them to intervene earlier.” Logic dictates that the earlier any problem can be detected, the better the outcomes — and the lower the cost. It makes sense, particularly when applied to risk-based models put into place for reimbursement. “Now there are justifiable, monetary values as well. That’s why we see a lot of organizations going toward cognitive computing solutions.”
The big question, of course, is how this can be done in an efficient and affordable way. Of course, as with many initiatives, it starts with implementing the right ecosystem and ensuring governance is in place.
Below are some of the key points from the webinar:
- Start with triage. Although computers won’t “solve” the challenging intake process, they can suggest a solution. However, predictive models do yield false positives, he cautioned. “Make sure you tune your models to a degree that produces a manageable number of cases for intervention.” When looking at cognitive computing solutions, Dr. Shabestari urged leaders to ensure that the operational and clinical sides are involved.
- Quality is king. One of the most important components in successfully leveraging cognitive computing is data quality. “Without good information, you put yourself at risk,” he added. Therefore, CIOs and other leaders must identify data scientists with operational skills in addition to mathematical, statistical, and modeling knowledge; support them, and surround them with capable people.
- Close the loop. As systems become more comprehensive and more information is available, it becomes increasingly crucial to create a closed loop around it. For example, if patient data is missing, users can get in touch with administration to ensure the blanks are filled in during the next appointment. “You need to make relevant information available to the people who can take action to improve the quality of data,” Dr. Shabestari stated.
- Integrate at the point of care. Any predictions that are made must be tied to the point of care, even if they appear to be separate from the actual care process. “You want it at the point where people make decisions. And that requires quite a bit of integration and expertise built around your data science team,” he said.
- Layering up. The typical cognitive computing environment includes three layers: incubation, production, and maintenance. Incubation involves having test cases that enable leaders to determine whether predictions will be successful, and if they are, demonstrate evidence to back a proposal.
- To the cloud. Although there’s no right or wrong answer when it comes to choosing an optimal cloud model (hybrid versus on premise), Dr. Shabestari is a proponent of the former. “Many solutions, algorithms, and program languages are open-source,” he noted. “You can build them on premise, then move them to cloud for intensive prediction capabilities.” This enables organizations to save on costs that would have incurred from retaining data that’s no longer needed.
- Maintenance matters. The process of maintaining cognitive computing models never stops. Leaders must constantly insert controls around factors that reduce risk and subsidize them, he noted. When that happens, “You get more resources, which means more wiggle room for identifying more cases,” he noted.
- Governance above all — literally. Sitting on top of every ingredient is a solid governance structure, Dr. Shabestari emphasized. But not just data governance. “We need to think beyond that and start talking about information governance. This isn’t just raw data; it’s what you do with it, and how you execute the information produced from that data.”
Finally, Dr. Shabestari urged CIOs and other leaders to think of cognitive computing as the sum of many parts that, when executed collectively, can result in improved outcomes. “Creating the right ecosystem involves everyone in the village, and it involves a lot of effort, but it pays off,” he said. “This might be our best chance for providing better care for many patients.”
To view the archive of this webinar — Designing the Framework for a Cognitive Computing Ecosystem — click here.