Navigating the path to AI at scale
Because so few healthcare systems use artificial intelligence on a large scale, real-world advice can help healthcare organizations guide the deployment of their AI systems.
Eve Cunningham is currently group vice president and head of virtual care and digital health for Providence, including service lines for virtual care companies, home hospitals, remote patient monitoring programs, virtual nursing and more. She will speak about how the healthcare system is approaching AI work 2023 HIMSS AI in Healthcare ForumDec. 14-15 in San Diego.
Her panel session, “Navigating the Path from Innovation to Scale: Strategies for Success and Sustainability,” also includes Corey Lyons, senior staff solutions engineer for healthcare at VMware, and Tariq Dastagir, assistant vice president of medical informatics and clinical trends, at Humana. .
Their discussion will focus on how healthcare systems can approach AI – from use case assessments and governance – and what they need to consider to develop, test and scale these technologies.
A structure for AI governance is critical
While large language model technology is the latest fascination, Providence has focused on implementing AI capabilities that support and support physicians.
“We can’t forget machine learning, natural language processing, optical character recognition, computer vision, robotic process automation and more,” Cunningham explains.
Providence has created a governance structure for AI that involves physicians and determines their path to AI implementations.
“We lay our groundwork specifically by thinking about the fact that a doctor should always be aware,” she said.
The goal of the AI governance structure – led by Sarah Ozzie, Providence’s Chief Digital and Strategy Officer, and Mark Primo, its Chief Data Officer – is to enable the organization to innovate without bogging down the process with numerous committees get stuck.
She called it a “Top-down, bottom-up approach.”
There are four subgroups leading the AI approach at Providence: consumer-facing, human resources, administration and back-office work. They each evaluate different technology capabilities, different use cases, and more.
There have been several staff requests regarding radiology use cases, she noted.
“There’s a lot of maturity in that area,” so Providence wants to speed up the review process for those deployments.
Cunningham said there have also been many requests to use LLMs to speed up workflows in the clinical environments – “mundane, repetitive tasks” – that are being considered.
With an AI governance framework in place and an emphasis on ROI and key performance indicators, Providence appears to be moving past fatigue for AI-driven ideas, she said.
Validating needs and balancing resources
To evaluate each AI use case, Providence’s AI working groups first ask how the idea can help address challenges related to three strategic priorities outlined in the governance structure.
Those fundamental priorities are staff shortages and burnout, hospital throughput and capacity, and fragmentation of care, Cunningham explains.
The working groups first validate that there is an end-user problem that the technology could potentially solve and ask, “Is it a bull’s eye in addressing these problems?”
They then validate the demand — its impact on prioritization, the resources required, the speed at which it can be adopted and the ease of integration into electronic health record workflows, she said.
“If it’s a very limited use case that really has a very limited audience or limited impact, but it requires a lot of resources, you know that might not be the best thing to start with.”
There may also be additional adoption issues.
“It might work very well for translating certain types of labs or imaging studies and might not work for others, so there would be some adoption issues that we might have to consider,” Cunningham said.
As Providence prepares to look at how to develop a potential AI system — “build versus buy” — if Providence decides to work with a vendor, the working groups evaluate market maturity and information security aspects, Cunningham said.
“Does this make sense, or are we going to act as a development shop for a vendor while they build a solution, which will put some administrative burden on the people involved in implementation?”
Return on investment can be difficult to measure for AI tools that accelerate workflows. For example, they aren’t reducing inbox messages or headcount, she said.
Sometimes great AI system ideas don’t get adopted by Providence.
“We’ve spent money and resources inventing some things, or working with the supplier and then saying, ‘You know what? This isn’t giving us the outcome we expected,'” she said.
These solutions are not scalable and working groups are shifting to prioritize other AI capabilities. But these efforts are not all in vain.
“We were able to learn from them and then use those lessons to come up with a better solution,” Cunningham said.
Andrea Fox is editor-in-chief of Healthcare IT News.
Email: afox@himss.org
Healthcare IT News is a HIMSS Media publication.