AWS, GE leaders talk hurdles to data-sharing, AI implementation

ORLANDO – The healthcare business is an unusual one: Although trillions of dollars are poured into the industry, billions of people worldwide don’t have reasonable access to care.

Part of the solution to that gap, explained Amazon Web Services Chief Medical Officer and Director of Machine Learning Dr. Taha Kass-Hout, may be found in artificial intelligence, and in technology more broadly.

“Innovations like precision medicine, conversational bots, AI scribes and APIs for data interoperability are great examples of how we can help improve care, close gaps in care, provide more efficiencies and also provide more equitable care,” said Kass-Hout in a fireside chat at the HIMSS22 Machine Learning and AI for Healthcare Forum on Monday.

Additionally, given the move toward the digitization of health data, particularly via the cloud, the question becomes how to use that information for the benefit of patients.  

One hurdle, as other HIMSS22 panelists pointed out earlier in the day, is the sheer amount of unstructured data being created.  

“Every health organization, payer or life sciences organization is trying to structure this information,” Kass-Hout said. “If you do, you can make better connected decisions, you can design better clinical trials, you can operate more efficiently or you can detect better trends in a population.”  

Vignesh Shetty, SVP and GM of Edison AI and Platform at GE Healthcare, told attendees at the fireside chat that bias is another issue for would-be AI implementers to contend with.

“A lot of times, people say, ‘I don’t trust AI,'” he said. “But it isn’t as much about the algorithm; it’s about the data that was used to create the algorithm and that could lead to potential bias.”  

“Breaking the ‘black box’ is not an easy task,” Kass-Hout chimed in, referring to promoting transparency around AI algorithms. “It’s really very hard. Understanding the bias that went into the model is also really hard.”  

Another major challenge, he said, is that data today is locked in “thousands of incompatible formats.”

“For many business reasons, they’re locked behind different silos,” Kass-Hout said. “You want all this information to come together at the point of care, where you sort of have a 360-degree view of every patient. So, you can understand what’s going on with them today, but also really try to forecast and predict what’s going to be next.”

In this way, he said, we can “start moving the care system from ‘sick’ care to, really, healthcare.”  

Organizations with fragmented data in their own companies, Kass-Hout suggested, should start with concrete use cases – including operational efficiencies.

“Start with the data that you need to address that use case,” he said. “By working through an end-to-end use case crisply, you’ll be able to bring a lot of this information together, and this is where you start realizing a lot of the value in machine learning technology.”  

“Start small and then scale,” said Shetty.   

“Don’t get held back by the fact that there are silos within the enterprise; that’s the reality for almost every enterprise today,” he continued.   

Overall, said Shetty, the industry is at “very early stages” of machine learning and AI.

“Healthcare and health tech [are] at an inflection point,” he said. “The specific marriage of human intelligence with some of these tools to drive better clinical and operational outcomes is something that I’m super-stoked about.”


Kat Jercich is senior editor of Healthcare IT News.
Twitter: @kjercich
Email: [email protected]
Healthcare IT News is a HIMSS Media publication.

Source: Read Full Article