Despite the promising potential of AI, managing costs associated with AI implementations is still a challenge for hospitals and health systems.
"I think cost is something that we're all struggling with right now," Allen Hsiao, MD, chief health information officer of Yale New Haven Health told Becker's. "Truthfully, these are amazing tools that have the potential to really help with burnout and clinician wellness. But they're not already budgeted costs and can be significantly large when you want to build them up."
One of the primary objectives at Yale New Haven (Conn.) Health when it comes to AI is making sure the organization is considering the compute costs and the need for secure connections with major cloud vendors. Dr. Hsiao said Yale is focused on striking a balance between the benefits of AI and the associated financial considerations.
"We have to proceed carefully and try to understand what the costs are with AI, especially when we scale it up to a larger number of users, or when we want to adopt something enterprisewide," he said.
One way Yale is proceeding with caution is by creating its own AI governance structure for the technology.
"Our chief digital health officer and our chief information officer are working to create an AI governance structure," Dr. Hsiao said. "But we've also been working on different AI initiatives through our current infrastructure. For instance, we have a data health governance group that meets for the system and the school. And I think they'll probably evolve to be a related committee focused just on AI."
When it comes to leveraging AI, Yale is currently working on several pilots with the technology.
"We're working with Epic to pilot AI-driven responses to patient messages," Dr. Hsiao said. "The sheer volume of messages ambulatory clinicians receive from patients is a big challenge. So we're hoping AI will take a bit of the edge off of them."
The other AI use Yale is exploring is using generative AI for documentation.
"We're working with Abridge to really test this out," Dr. Hsiao said. "We're testing this for multiple medical and surgical subspecialties in addition to primary care, where it's been very well proven by several organizations that we spoke to. We're also interested in exploring it in the inpatient and emergency department spaces as well."