6 best practices to help health systems get the most out of AI

Hospitals and health systems are increasingly looking to use artificial intelligence technologies to improve clinical and operational performance. However, actual deployment has been slow-moving and rife with obstacles.

While some of the barriers are technical, others stem from a fundamental misunderstanding of AI and its potential, therefore resulting in ineffective integrations and a lack of support to help the technology reach that potential.

During a Feb. 11 webinar titled "The 2020 Guide to Unleashing AI in Healthcare: Three Expert Perspectives," which was sponsored by KenSci and hosted by Becker's Hospital Review, Parveen Chand, COO for the adult academic medical centers of Indianapolis-based Indiana University Health, Tom Lawry, national director for AI in health and life sciences at Microsoft, and Ankur Teredesai, PhD, co-founder and CTO of KenSci, discussed the concepts and priorities health systems and their industry partners must keep in mind when launching an AI project.

Before launching any AI initiative at all, according to Dr. Teredesai, it is crucial for organizations to understand several core ideas about the technology.

For one, AI should be approached not as "artificial intelligence," but as "assistive intelligence," he said, with organizations adopting a social engineering approach to all AI initiatives. According to Dr. Teredesai, that entails "bringing in diverse stakeholders very early on and establishing baselines on performance and expectations regarding the endpoints of what an AI implementation is expected to achieve."

For another, organizations must follow a principle of "responsible AI," prioritizing ethical, fair and transparent technology. "Today, across the globe, the biggest risk for AI implementations is an increasing amount of models that are black boxes — that are not open to inspection," he said. "They can cause significant harm if we don't understand how or what the AI implementations are doing."

Finally, Dr. Teredesai suggested organizations keep in mind that "operationalizing AI is not the same as building an AI model on a retrospective dataset." That is, rather than merely applying pre-existing models to an organizations' individual datasets, they should be prepared to design and redesign their own AI models based on their specific data needs and the evolution of their organizations.

Beyond these more strategic practices and concepts described by Dr. Teredesai, Mr. Chand provided a hospital-specific perspective, outlining six concrete tactics to improve health systems' chances of success when implementing AI:

  • Align with goals: "By identifying where projects fit into your organizational goals, it will become very clear where to spend your resources and energy."
  • Identify disruption: "Look for areas of the organization that are ripe for disruption … those departments or leaders or functions that are eager to take on risk and lead the way."
  • Prepare for impact: "You've got to understand the potential impact that technologies can have on your work. Think through the applications, then create and celebrate small wins."
  • Develop expertise: "Identify subject matter experts, either through partnerships or organic growth, to help develop some level of expertise."
  • Vet partners: "You must do your own due diligence in selecting a partner."
  • Stay flexible: "Being an organization that has the ability to pivot is something that is going to be instrumental in future success."

Above all, according to Mr. Lawry, the most successful AI implementations will be those that focus not just on the technology itself, but on the people that are integrating and working alongside that technology.

"There are actually two ways — and only two ways — that AI is going to add value to a healthcare organization in a measurable way," he said. "One is that it's going to automate processes, and two is that it's going to augment the wisdom and the work and the service of knowledge workers."

What that means, Mr. Lawry explained, is that AI "is going to change operational and clinical workflow processes." Therefore, human resources and other employee-centric experts must be included in all discussions about AI implementations, which will ensure that humans are not fully replaced by AI, but will instead see their work enhanced and improved by the integration of the new technology.

"Everyone talks about artificial intelligence, but have you ever heard anyone talk about artificial wisdom?" he said. "That includes all those things that are unique to your humans, to your clinical workers, to any knowledge worker — it's more about how we put AI behind their work."

View the full webinar here.

More articles on AI:
Medtronic acquires AI company Digital Surgery
Only 44% of healthcare leaders say employees are prepared for AI adoption: 4 notes
Trump budget includes big gains for AI, quantum computing

© Copyright ASC COMMUNICATIONS 2020. Interested in LINKING to or REPRINTING this content? View our policies by clicking here.


Featured Webinars

Featured Whitepapers