Nashville, Tenn.-based HCA Healthcare has started using generative artificial intelligence technology from Google to document emergency room visits and speed up nurse handoffs.
The 182-hospital system has rolled out the clinical documentation at four hospitals, where it's being used by 75 emergency physicians, while the patient handoff tool is in testing at UCF Lake Nona Hospital in Orlando, Fla., HCA and Google said Aug. 29.
"We thought we had a much longer road in front of us to tackle some of the really big problems in healthcare, just because of how complicated they are," said Michael Schlosser, MD, senior vice president of care transformation and innovation at HCA Healthcare, at a press briefing. "And then generative AI came along and we realized the future was probably closer than we really thought it was."
The partnership between the tech giant and one of the nation's largest health systems shows the speed at which healthcare is adopting generative AI, not even a year since the debut of ChatGPT. The two organizations began collaborating in 2021 to support HCA's digital transformation.
"We spent a lot of time talking to our clinicians and asking them directly about what their challenges are, and 'What would you like 'care transformation' to focus on and to solve?''' Dr. Schlosser said. "And the doctors unanimously said documentation."
In the four HCA ERs, physicians have hands-free devices outfitted with the Augmedix medical transcription app that record their interactions with patients. The data is then fed through Google's PaLM large language model, instantly creating a medical note that the providers review before it's uploaded to the EHR. HCA plans to scale the program to other departments and hospitals.
"There's a holy grail that the doctors always talk about, which is that I could just go provide care and the documentation would take care of itself," Dr. Schlosser said. "Well, we think that that future is coming into vision with the advent of these large language models."
At UCF Lake Nona Hospital, a generative AI platform digests data about a patient from the previous 12 hours — such as vital sign changes, medication adjustments and lab work — to create a summary and care plan for the incoming nurse. The handoff tool is currently in the "proof-of-concept stage," Dr. Schlosser said, but could eventually transform into a "gen AI assistant" that helps clinicians in a variety of ways.
The health system is also testing Med-PaLM, Google's large language model designed specifically for healthcare. HCA's AI platform is not storing or learning from its patients' data, and humans are always involved in the process, Dr. Schlosser said. ER patients give consent before AI listens to their interactions with clinicians.
Dr. Schlosser said this is only the beginning for generative AI at HCA. He envisions the technology one day "nudging" clinicians in the EHR on what questions they might ask or a possible diagnosis they might have missed.
"Ultimately, I do think decision support is an interesting use case, but I want clinicians to fully embrace AI as a partner that's making their life easier before we start getting into some of those more controversial areas," he said. "Let's build some confidence by having the AI do things that they love that it's doing for them, so that we can break the ice and then build up."