Health Tech

How Can Health Systems Cut Through the AI Noise? 3 Leaders Answer

During a webinar, three health system leaders discussed strategies they use to ensure successful AI deployment. They agreed that one of the best ways health systems can prime themselves for fruitful AI deployments is beginning with a specific problem to solve — as opposed to having AI startups tell them which problems need to be addressed.

Hospital administrators are constantly inundated with pitches from AI companies claiming that they can solve healthcare’s myriad inefficiencies. On Thursday, three health system leaders came together for a webinar to discuss strategies they use to cut through all that noise.

During the webinar, which was hosted by Reuters Events, the executives agreed that one of the best ways health systems can prepare themselves for fruitful AI deployments is beginning with a specific problem to solve — as opposed to having AI startups tell them which problems need to be addressed.

Denise Basow, chief digital officer at Ochsner Health, pointed out that health systems have to think very intentionally about the vendor acquisition processes in order to prime themselves for AI success.

“Don’t come to us with a solution. Come to us with a problem, and we’ll help find a solution. Because if [vendors] come to us with a solution, we usually end up kind of dashing their hopes and dreams as we go through with the prioritization process and governance and all those things,” she explained.

Hospitals need to not only be clear on the specific pain points they are seeking to solve, but also their goals and the metrics they will use to measure success, another panelist noted.

Reed Smith, chief consumer officer at Ardent Health Services, encouraged health systems to formulate their own roadmaps for success. This is important because it ensures that the health system has more control over what its AI deployment looks like and is not easily influenced by the business goals of a vendor, he said.

“A lot of vendors out there have their own product roadmaps, and they have their own development roadmaps. That’s not bad in and of itself, but I would caution against marrying yourself to that roadmap of a vendor or a solution provider because then you’re beholden to that course of action,” Smith explained.

Just because a health system creates an internal, fixed roadmap for success doesn’t mean that it can never change its plan. Smith emphasized that it’s okay to have to make tweaks or reformulate the plan — he also pointed out that having a strong objective in place beforehand can make those adjustments go more smoothly.

In Smith’s view, hospitals shouldn’t be afraid of failure when it comes to AI deployment, given that many of the healthcare field’s AI tools are still in an iterative phase. 

Many hospitals still get discouraged by AI pilots that didn’t go as planned, though, noted Ainsley MacLean, chief medical information officer and chief AI officer at Permanente Medicine

She brought up the idea of “death by pilot,” which refers to the demoralizing phenomenon hospital leaders feel when tech pilots repeatedly end up with less-than-desirable results.

“I’ve been thinking about that a lot because I think that pilot work has a tremendously important role. At the same time, if you want to really leverage the cost savings and the quality enhancements, not to mention the speed of implementation, you sometimes have to buy. And so we do a mixed approach,” MacLean said.

Basow said that she has “definitely talked to” some hospital leaders who have decided that they’re not going to conduct any more pilots, and instead only buy tools that have proven their ability to work at similar organizations. But this introduces a bit of a “chicken or the egg” paradox, she pointed out — how can you adopt a tool with proven outcomes without the pilot programs that prove outcomes?

At Ochsner, leaders don’t agree to pursue a pilot unless they have the organizational buy-in required to eventually scale the program, Basow said.

“That doesn’t mean that you scale no matter what. It means that you’ve identified upfront the outcomes that will be meaningful. Let’s identify those upfront, and if we meet those outcomes, we will be able to fund it and scale it,” she explained.

Completing a successful pilot and then not knowing how to pay for it is “the worst thing you can do,” Basow added.

In some cases, partnering with a vendor might not be the best choice, though. 

If vendors aren’t willing to work within the health system’s parameters for success, then it could very well make sense for the organization to build an AI tool on its own, Smith declared. This could also potentially lead to the health system commercializing the product down the line, he added.

“Otherwise, you could end up down a path where you’re trying to bend a solution to meet your needs, and then you’re disappointed when it doesn’t work,” Smith remarked.

Photo: metamorworks, Getty Images