Why hospitals are leveraging AI to drive operational intelligence

Despite massive venture investments in healthcare artificial intelligence — about $6.6 billion annually, by some estimates — there are few instances of hospitals using machine learning in real-world applications. Becker's Healthcare publisher Scott Becker was fortunate to catch up with three organizations about their real-world AI use cases at Becker's Hospital Review 9th Annual Meeting in Chicago.

 

This article is sponsored by Qventus.

During a panel discussion April 11, three healthcare leaders shared how their organizations are using artificial intelligence to streamline hospital operations and support frontline clinicians. Panelists included: 

  • Karim Botros, senior vice president and chief strategy and innovation officer at The MetroHealth System (Cleveland)
  • Peter Fleischut, MD, senior vice president and chief transformation officer at NewYork-Presbyterian (New York City)
  • Mudit Garg, founder and CEO of Qventus, the leader in operations management for health systems (Mountain View)
  • Moderated by Scott Becker, JD, publisher of Becker's Healthcare and partner at McGuire Woods

AI and operational improvement

Dr. Fleischut and Mr. Botros represent health systems with distinct operational and strategic needs.

NewYork-Presbyterian comprises of seven hospital campuses, two of which are academic medical centers, as well as a regional hospital network, a physician services group and a community and population health division. The urban hospital system alone includes 2,600-plus beds and 6,500 affiliated physicians, and reported $5.6 billion in total revenue in 2017. MetroHealth is a single-hospital public health system with about 25 care locations in Ohio. The system reported $1 billion in total revenue in 2016.  

Despite the health systems' differences in scale and scope, Dr. Fleischut and Mr. Botros see similar opportunities for AI solutions to improve daily operations that support the patient care journey.  

"Financial sustainability in part depends on getting better at operations," Mr. Fleischut said. "We [at NYP] are thinking differently about how we operate. You won't hear me talk about AI and machine learning in terms of improving clinical diagnoses and clinical decision support — I believe the value [of AI] is in operations, in everything from time keeping to insurance authorization, patient eligibility and moving the pieces to improve length of stay."

MetroHealth also chose to focus AI investments in operational support before tackling clinical decision support.

"Rather than go down the path of clinical decision support, we're focusing on operational decision support and integrating it seamlessly within our workflows," Mr. Botros said. He noted that MetroHealth implemented Qventus in November 2017 to help clinics improve scheduling by predicting the likelihood of patient no-shows days in advance. The solution has had an 82 percent accuracy rate, allowing MetroHealth to improve physician productivity by filling appointment slots with new patients in fewer than 24 hours. 

Making the 'Google Maps' of healthcare AI

When discussing the opportunities for AI in improving hospital operations, it didn't take long for a common theme to emerge. Panelists agreed the key to successful AI use cases was seamlessly integrating AI in clinicians' workflows — to the point that the AI solution itself was invisible.

"Seamless integration [of AI] means staff don't see it happening; it goes on in the background," Mr. Botros said. "It's almost invisible, so things still feel normal and employees' day-to-day tasks stay the same."

To envision "seamless" AI in healthcare, Mr. Fleischut drew an analogy between AI adoption in hospitals and the way consumers use Google Maps. He spoke about their with Qventus to apply AI to operational challenges, as a driver uses Google Maps. "We take care of about 810,000 patient stay days across six hospitals. If you multiply that times the magnitude of decisions that need to be made on any given day, you're into billions of decisions," said Dr. Fleishchut. "Frankly, I don't think it's fair to ask someone to make all of those decisions when the environment is constantly changing so much. We need to provide tools to providers to help manage the magnitude of operational tasks of say, length of stay." NYP is early stages of doing this, deploying across 30 medicine units on both Columbia's and Cornell's campuses.

At Qventus, Mr. Garg is committed to developing AI-powered tools that give frontline providers the same clarity and user experience as Google Maps, whether it's using predictive analytics to recommend which patient to see next or which discharge to prioritize.

"After years of working in hospitals, we've seen the way EHRs were implemented make things harder for frontline teams when it should be making healthcare work better," Mr. Garg said. "That's why we founded Qventus. We believe in the power of data. But that data is useless if it can’t make it easy for our teams to do the right thing. Technology can be used to turn data into action and enable better decisions, as well as a faster, flatter and more flexible organization."

Transformation depends on frontline engagement

Health systems are challenged to implement large-scale, sustainable change across an enterprise. About 70 percent of company transformation efforts fail due to poor employee engagement, according to McKinsey and Co.

"[Organizational change] has nothing to do with technology but [has to do] with the people," Dr. Fleischut said. "It must be addressed at every facet of the organization across the enterprise.  We are now really honing in on actually making sure that the tools we deliver can translate in an actionable, sustainable way to the front lines "

As NewYork-Presbyterian gears up to further deploy AI, Dr. Fleischut said the organization is drawing on an important lesson it learned from implementing telehealth roughly two years ago. The lesson was the importance of engaging every single employee across the enterprise in the technology initiative, no matter how often or little that employee may interact with the technology. "It has to be something that will make it easier for them to do their jobs."

More articles on artificial intelligence: 

Intermedix named one of America’s best midsize employers by Forbes
AMA funnels $27.2M into value-based care startup incubator
Blockchain collaborative Hyperledger adds 2 healthcare members

© Copyright ASC COMMUNICATIONS 2018. Interested in LINKING to or REPRINTING this content? View our policies by clicking here.

 

Top 40 Articles from the Past 6 Months