The first fully automatic clinical documentation application powered by OpenAI's GPT-4 is scheduled to be widely released later this year, .
This latest artificial intelligence (AI)-powered application, known as Dragon Ambient eXperience (DAX) Express, will be able to create an automatic clinical note based on recordings of patient visits, said Diana Nole, MBA, executive vice president and general manager of Nuance Healthcare, which was purchased by Microsoft in 2022.
"It's the latest solution that addresses one of the biggest industry issues and challenges, which is physician burnout," Nole told Ƶ. "We not only reduce the amount of time that the physician has to spend on clinical documentation -- we've had estimates that it reduces per visit 7 to 10 minutes -- so they can actually see more patients."
"It also gives a much higher level of satisfaction between the physician and the patient, so that there's more compliance in terms of their treatment pathways and what each of them needs to do," she added.
Nole explained that DAX Express will combine ambient and conversational sound recordings with the generative AI capabilities of GPT-4 to capture and analyze patient encounters with their provider and produce a draft of a standardized clinical note for the provider's review. The recording will be captured using the provider's mobile device.
The whole process is supposed to take a few seconds after each patient visit, and it will allow the provider to focus on the patient while still having final sign-off on the documentation from the visit, she added.
"The physician can go into the exam room, speak with the patient face-to-face, focus on them, [and] build that trust," Nole said. "Then automatically when they leave that patient room, they have this clinical note waiting for them."
However, not everyone in the healthcare industry is completely convinced that adding an AI-powered application will be the salve to the issues of provider burnout. A by the Pew Research Center revealed that fewer than half of Americans believe that AI can improve patient outcomes, and 60% said they would feel uncomfortable if their provider relied on it for medical care.
The commentary around this announcement was also met with skepticism on .
captured the uncertainty around providers using Nuance's DAX application, the earlier model of this AI-powered automatic documentation application, which notably included an additional reviewer working behind the scenes to prepare the draft clinical note for a provider's review. A senior executive at Nuance acknowledged in the article that trusting AI can be challenging in part because of its proclivity to "hallucinate," a term coined to describe the confident falsehoods frequently produced by AI chatbots, like ChatGPT.
Still, J. Scott Smitherman, MD, MBA, chief medical information officer at Providence Health, a national not-for-profit health system, said the DAX technology has already made major improvements in his day-to-day clinical practice.
"My experience with [it] is that my last patient is at 4:40 in the afternoon, and I walk out right at 5:00 with my notes done, which is not the experience I was having before," Smitherman told Ƶ.
However, he said that the current DAX application still takes up to 4 hours to generate each clinical note because it requires a human reviewer, so the potential of the new GPT-4-powered application to instantly draft each note could be the thing that changes how healthcare providers view this technology.
"I am very excited about that concept because this issue of turnaround time is one of the last pieces that has kept this from being a solution that everybody wants," he said. "This is something that truly has a chance to take a piece of doctor's work and make it significantly faster."
Nole said that the benefits of this new application are that it is built using the privacy and security of the Microsoft Cloud service paired with the power of GPT-4. The application allows for users to completely control their data, which is fully encrypted and HIPAA-compliant, she noted.
The application will be made available to a select group of Nuance's clients during the summer for a private preview, then the company will fully release the application in the fall, Nole said. Nuance currently has more than 550,000 providers using its current Dragon applications.
Nole emphasized that these large language models are trained on millions of encounters using Nuance's conversational and ambient AI, which allows the model to improve with each new interaction it records. She added that because physicians can edit the drafted documentation, the model can also use that feedback to improve.
"I really do feel that this is the future. This is the way that medicine should be practiced," Nole said. "It should be the physician having the time and ability to focus on the care of the patient."