'Readiness gap' holds back GenAI adoption in healthcare, survey finds

What's behind the GenAI readiness gap

Despite the lack of preparation, 67% of respondents saw GenAI as a solution to prior authorizations problems, 62% believed it could help with electronic health record management, 68% saw Gen AI improving cybersecurity preparedness and 65% said it could boost telehealth or virtual care programs. 

In addition, although 80% of respondents saw "optimizing workflows" as a key company goal, only 63% believed they were ready to use GenAI to help them do it. 

Flawed workflows need to be improved first before layering on a GenAI tool, according to Dr. Matthew Crowson, director of AI and GenAI at Wolters Kluwer Health, an information services firm.

"You can't just air-drop in a tool or a GenAI solution into a suboptimal workflow," Crowson said. "You actually have to reengineer the entire workflow, as opposed to just layering these things on top and hoping you get better outcomes out of them."

Other readiness challenges included staffing shortages and clinician burnout. In fact, 85% of respondents saw recruiting or retaining nursing staff as a top workforce concern with GenAI, and 76% cited "reducing clinician burnout" as an issue.

Academic medical centers have readiness issues because of layoffs and a lack of capital for expenditures, Crowson explained. 

Vacant roles such as in nursing and medical assistant positions create difficulties for organizations to add new tools to their workflows, he suggested.  
 
"Just to do the daily tasks regardless of technology is challenging these days, and everybody is being asked to do more with less," Crowson said.

Also, many healthcare organizations are concerned about security risks in using the cloud. They also are not accustomed to working with a cloud provider, and they are unable to afford the computational hardware to use AI apps locally, according to Crowson.

"There are some barriers there because it requires building that part of your technical strategy for your enterprise to operationalize," Crowson explained. "These things have to run on some hardware somewhere, whether you buy it or you rent it through a cloud vendor."

'Crawl, walk, run': How to integrate GenAI

GenAI can help address staffing issues at healthcare organizations, according to Crowson.

"People are viewing GenAI as a backfill for having to help address some of the capability gaps that the shortages of nurses and other allied providers are producing," Crowson said.

To solve challenges with GenAI readiness, proper training in the technology and the processes for using it will be necessary. Providers lack the technical literacy, data science knowledge and "baseline" AI literacy, according to Crowson.

"Like any tool, you have to know how to use it appropriately and be trained on it as well, and that requires some AI literacy," Crowson said. 

To improve workflow optimization, healthcare organizations must innovate with Gen AI using ambient listening and clinical decision support, according to the survey. Tools such as Suki allow healthcare providers to gain efficiencies in documenting patient visits and connecting it with EHRs.

Crowson suggested a "crawl, walk, run" approach in which healthcare organizations start small with a pilot project. That could mean a three-month pilot with ambient AI dictation, he said.

"Piloting something is a fairly inexpensive way to get in the game and get started and do a lot of self-discovery as part of that," Crowson said.

Several "layers" must be in place to become GenAI ready. That includes training and literacy, being data-ready and having proper connectivity with EHRs.  

"At the patient level, these models have to be hooked up to the EHR to read off a patient's record to be able to provide personalized information as part of that encounter," Crowson said. "Hooking up that plumbing requires some expertise and having your data house in order. That's probably one of the biggest barriers to all this. A lot of organizations aren't particularly data mature. They don't have their data in a format that's ready to be modeled and used by large language models." 

Crowson highlighted the documentation process for surgery approval requests as an area that GenAI can help by drafting appeal letters to send to payers. These letters would otherwise be written by physicians, nurse practitioners or physician assistants, Crowson said. 

"Potentially, we can get to a decision much faster and get the patient on their way, getting the care they need, or where they want," Crowson said. That could improve the "latency" in provider-payer communication, he added.  

Payers can also use GenAI on their end as part of the payment reimbursement process, he said.  

To help with corporate approval for AI projects, healthcare organizations could hire a chief AI officer to take responsibility for educating employees on GenAI, Crowson said.  

Healthcare organizations can get buy-in for AI apps by starting with the "low-hanging fruit," Crowson said. That means automating administrative tasks rather than trying to cure cancer. (However, AI does help with detecting types of cancer such as prostate cancer.) More higher-risk use cases include suggesting diagnoses, which are mostly in the research stage except for radiology and pathology, Crowson said.  

"Given the regulatory environment and given what we know of AI's capabilities, I think we're probably collectively more comfortable deploying these tools on some of these lower-risk everyday burdens, as opposed to the moonshots out of the gate," Crowson said. "I think we'll get there." 

Brian T. Horowitz started covering health IT news in 2010 and the tech beat overall in 1996.