Few health systems have generative AI strategy

Despite the hype around generative AI tools like ChatGPT and its competitors, a mere 6% of health systems have a solid strategy in place to leverage these models to improve their financial sustainability, according to a new survey released by Bain & Company. 

Generative AI relies on large language models (LLMs) to ingest and synthesize huge volumes of data. They use a natural language interface, such as a simple text box, to take requests from a user. The results can include completely original images, narrative text, or other types of content, although these products may suffer from biases inherent in the source materials.  

In the healthcare context, generative AI has shown promise for making sense of enormous volumes of complex clinical data, assisting with documentation, discovering new drugs, and automating a variety of time-coming tasks. 

The survey revealed that health system leaders are excited about the potential to apply these tools to everyday operations in the care environment and the back office, with 75% of executives stating they believe generative AI is ready to reshape the industry. But how, exactly – and when – remains unclear. 

Where can generative AI make an impact?

In the near term, leaders view the opportunities as mostly administrative and financial in nature. Respondents to the poll believe generative AI can have the most significant impact on charge capture and reconciliation, structuring and analyzing patient data, and optimizing workflows.   

Fewer believed that this type of AI is ready to take on more purely clinical tasks, such as decision support and risk stratification. This tracks with other industry polls that indicate there is still a major trust gap between clinicians and AI.  

Interestingly, respondents were least confident in generative AI’s ability to mine clinical documents for research purposes and aid with clinical trials, despite the fact that these areas have been primary targets for exploration and align with the strengths of large language models in general. 

Even in the longer timeframe of 2-5 years, health system leaders don’t see generative AI making a big dent in research and drug discovery. But they do feel that it will be ready to enter the patient-facing arena with clinical decision support, treatment recommendations, and predictive activities. 

Executives are more likely to believe AI will make an impact in the realm of patient experiences and health system navigation, as well, over the next five years. Generative AI may be able to assist in such tasks as interacting with payers and providing help with care coordination, as well as playing a role in call centers for both administrative purposes and clinical questions.   

 Barriers to implementing generative AI in the healthcare ecosystem 

Health systems may be slow to take a strategic look at generative AI because they feel the barriers are too high to cross right now, the survey indicated. Financial constraints are particularly worrisome for hospitals and health systems, which have been struggling through a grim post-COVID economic environment.   

Approximately half of participants cited resource constraints as the top challenge to implementing generative AI at scale, followed by the lack of available technical expertise to make it happen. A significant number also pointed to lack of clarity in the regulatory environment and organizational resistance as reasons to keep generative AI off the priority list for the moment. 

What doesn’t faze them? Ethical concerns, clinical risks, or unclear benefits of the technology.  These items were at the bottom of the list of barriers by quite a notable margin, hinting that AI might be more of a focus for health systems if the financial factors weren’t so much of an issue. 

Ready or not, here it comes 

Even though most health systems say they don’t have a strategy for generative AI, how much do they really need one? With health IT vendors like Epic simply baking generative AI into their products to enhance and expand known capabilities, health systems are going to be able to use generative AI sooner rather than later, whether or not they have developed their own roadmap for doing so.   

That could take some of the burden off health systems that don’t feel prepared. Working alongside vendors to replace or augment traditional analytics methods with generative AI techniques is likely to be a lot easier than it was to develop a health IT environment from scratch, so much of the hard work has already been accomplished.   

Just like we no longer talk about having an “analytics strategy,” and instead focus more on the specific tasks that require analytics as a background competency (chronic care management; value-based contracting; workflow automation), generative AI will soon become so ubiquitous in healthcare products and services that it won’t require a special call-out. It will just be the way things work. 

That doesn’t mean health systems don’t have to think carefully and strategically about how to design their next-generation technology stack, particularly in areas where bias could be an issue for patient safety or care quality. 

However, leaders should consider thinking about generative AI in context: it’s yet another valuable tool that will allow organizations to further existing capabilities and fuel the new products used to accomplish important financial, operational, or clinical goals, not necessarily a stand-alone strategy that can (or needs to be) implemented in fully controlled circumstances according to a color-coded timeline. 

AI has always been a wild, messy discipline that moves at the speed of light, and health systems certainly need to be prepared to choose the right generative AI tools that meet their needs. But the barriers to entry might be lower than executives think, especially in organizations that have already invested in a strong foundation of data-driven infrastructure and processes.  

Jennifer Bresnick is a journalist and freelance content creator with a decade of experience in the health IT industry.  Her work has focused on leveraging innovative technology tools to create value, improve health equity, and achieve the promises of the learning health system.  She can be reached at jennifer@inklesscreative.com.