Artificial intelligence fueling the need for a digital workforce in healthcare

First movers such as Mount Sinai, Fresenius and Houston Methodist are running AI pilots and enabling employees to work alongside machines.
By John Sharp
12:15 PM
Share
AI digital workforce

NEW YORK — Healthcare organizations implementing artificial intelligence must move to “decision centricity.”

That’s according to Michael Berger, VP of Population Health Informatics and Data Science, Mount Sinai Health System.

WHY IT MATTERS

Mount Sinai defines decision centricity as it’s “mission is to improve population health and the effectiveness, efficiency, and equity of health care by developing, analyzing, and disseminating evidence to guide clinical decision making and policy.” 

That theme was repeated in various flavors throughout the AI4 Healthcare Conference here last week.

Berger said, for instance, that it is important to empower users with tools to make decisions. He also quoted his CEO Ken Davis on the topic of population health “If our beds are filled, it means we've failed.”

AI and machine learning are one of the ways hospitals such as Mount Sinai focus on keeping patients out of those beds. Berger said Mount Sinai runs analytics against some 1.5 billion data points from the EMR, claims and other sources to create patient profiles through propensity matching.

THE BIGGER TREND

Mount Sinai is not the only hospital moving on AI and machine learning. Fresenius Medical Care, Houston Methodist also discussed work each has already undertaken.

Fresenius Medical Care, a national dialysis provider takes a pragmatic approach to advanced analytics. According to Len Usvyat, Vice President of Integrated Care Analytics, a steering committee of both executives and technical staff prioritize projects including AI models. The portfolio includes cognitive insights, operational optimization and prioritization algorithms. Fresenius, in fact, has big data with 1,600 variables per patient, according Usvyat.

Michelle Stansbury, VP of Corporate Business and Revenue Cycle at Houston Methodist, said the system has created a digital workforce receptive to the use of artificial intelligence and intelligent process automation. Early adopters included supply chain, insurance verification, scheduling and physician credentialing.

Stansbury emphasized the importance of stakeholder management in rolling out AI and clearly identifying roles and responsibilities including those for bots and people alike.

WHAT COMES NEXT

Stansbury’s approach fits with the concept of obtaining FDA approval for artificial intelligence algorithms. She even suggested credentialing bots in the sense of testing them to ensure they stay in their role. Beyond pilots, Houston Methodist moves to production, growth and optimization.

During a panel on Population Health: A Data-Driven Model for Patient Care, Northwell Health’s Chief Innovation Architect Vish Anatraman said the system uses probabilistic programming to focus on what’s important within 15,000 data points per patient.

Once this is accomplished, the next step is to deploy the AI models into workflow by acting as a sidecar to human providers. They emphasized incremental change to help physicians get used to the idea of working at a higher skill level, leaving tasks of identifying patients at risk to the analytics.

AI CHANGING MINDSETS

As AI and machine learning continue being implemented in provider and payer organizations, the technology and people driving adoption are also changing the mindset of stakeholders from skepticism to understanding the usefulness of AI in simplifying tasks.

As we have also seen in the ongoing HIMSS Media Focus on Artificial Intelligence during November, keeping an eye on how AI is effectively implemented and including stakeholders to create a digital workforce that operates along side machines is now essential for health and IT leaders.

John Sharp is the Senior Manager of the Personal Connected Health Alliance at HIMSS.

Focus on Artificial Intelligence

In November, we take a deep dive into AI and machine learning.