The terms ‘artificial intelligence’ and ‘pragmatic’ don’t typically go together. But if healthcare is going to realize the true potential of AI and machine learning, perhaps it’s time that the two become better acquainted, said Marc Mangus, Principal Specialist Solution Architect at Red Hat, during a recent discussion, which also featured John Halamka, MD (President, Mayo Clinic Platform) and Zafar Chaudry, MD (Chief Digital & Information Officer, Seattle Children’s).
Pragmatic AI, Mangus noted, is focused on “driving toil out of the system” by automating tasks that are repetitive or don’t necessarily require human intervention. “That’s where you find the low-hanging fruit.”
Halamka agreed, noting that boards and executive teams don’t want to hear about innovation or digital transformation when they’re dealing with negative margins, broken supply chains, and burnout. Instead, it should be about “trying to meet people in the trenches with business solutions that solve problems in the economic headwinds of 2023.”
One of those problems is a shortage of people and skills, which has led organizations like Mayo Clinic and Seattle Children’s to leverage AI to automate administrative tasks like helpdesk tickets and billing. “Everybody wants technology departments to do more for less,” said Chaudry. And so, his team is utilizing tools to fill open positions, rather than eliminating roles by implementing tools.
It’s a strategy that Mangus believes will position them well. “You’re going to get a lot of bang for your buck by shoring up automation around repetitive tasks,” which can then enable providers to spend more time on patient care. “People are doing some great work.”
“Good AI and ML”
But, like any major initiative, it comes with myriad challenges that need to be overcome, according to the panelists. Below, they shared advice on how to harvest that low-hanging fruit.
- All about data. First and foremost, “good AI and machine learning starts with highly curated data,” said Halamka, who described a 10-year agreement between Mayo Clinic and Mercy to create a “distributed data network” that enables users to develop and validate ML models that can detect diseases earlier and improve overall health. The key, he added, is to “make sure you develop great data, have transparency around where it does and doesn’t work, bring something of value to the user, and make sure it fits into the workflow.”
- Build to scale. “Creating scale is really important if we’re going to train these models,” said Halamka, who has adopted a cloud-neutral approach. “Every cloud provider has its own strengths. Embrace the cloud. Create large scale, deidentified, secured data sets, and create processes that make collaboration easy. It’s only by creating this ecosystem, this community, that we can innovate to meet the needs of our patients.”
- Begin at the end. Another key step, according to Mangus, is to start with outcomes in mind. Failing to do so can send initiatives off the rails. “Who am I actually trying to serve? What value am I delivering to those that I do serve? And how is it really making their lives better? It’s that simple. It’s not anything to do with technology.”
- Solid foundation. Infrastructure, on the other hand, does have to do with technology, and should be a top priority. However, the “plumbing required to stand up production-grade AL and ML solutions often gets overlooked,” said Mangus. “Everyone looks at the shiny thing, which is the algorithm. But what about getting access to data? What about integrating with disparate data sources? What about securing it? What about managing the performance and monitoring it?” All of this can place a significant burden on IT, he noted. “That’s where Red Hat comes in, by pulling together the pieces to make innovative projects run, whether it’s on-prem or in the cloud.” Having a platform, he noted, allows leaders to “focus on the business-critical part of the algorithm, which is the differentiating or innovative thing you’re trying to do.”
- Know the naysayers. One component that can’t be overlooked is resistance to change — which is more common than many believe, said Chaudry. “When we automated our tickets around setting up servers, there was a level of resistance from internal teams as to, ‘What does that mean for my job? Am I losing my job?’ Whereas I would’ve thought that technologists would tend to embrace new innovation, it has become a hard sell.”
The best way to manage that fear, Chaudry noted, is through transparency. “People want to know, what will my life look like and what’s in it for me? The truth is that some roles will no longer be needed. You shouldn’t lie about that. You have to be clear.” For administrative staff, it might mean being retrained, upskilled, or moved to other areas of the organization. By showing people their potential — and giving insight into what the “new world” will look like, it can eliminate the fear factor, which then promotes collaboration.
Even providers can benefit from some clarity, Chaudry added. “If you don’t show the nurse, the allied health professional, or the physician what their life will look like once you give them a device, they’re not necessarily going to embrace that. And so that’s the approach we try to take.”
Creating a Community
What’s also critical, according to Halamka, is building buy-in by sharing the vision with clinicians. For example, clinicians will be able to see twice as many patients without sacrificing safety or quality, which will improve their performance. It’s also reassuring them they’ll have constant support. “All of a sudden, a lot of the barriers melt away, because you’ve aligned incentives and created a community.”
Mangus concurred, noting that it’s critical to remain focused on the value stream by constantly asking what services are being provided and how they will change peoples’ lives. It’s also building that into the DNA of the project and involving all of the stakeholders to be part of the solution.
Finally, as with all initiatives, the education piece is paramount. “It really comes down to embracing a learning culture,” and rewarding people for upskilling, which Mangus believes is a critical component of any successful project. “A rising tide raises all boats. We can learn together and be part of a community and make each other better.”
To view the archive of this webinar — Leveraging AI’s Low-Hanging Fruit to Help Your Organization Survive the Downturn (Sponsored by Red Hat & Carahsoft) — please click here.