How ChatGPT technology is applied in healthcare
There is a lot of hype around AI in healthcare, and digital healthcare companies are looking to cash in.
But experts aren’t sure how generative AI applications like ChatGPT and GPT-4 will impact clinical diagnosis and decision making. Most say the first wave of adoptions will take place in areas where there are administrative layoffs.
Related: Microsoft and Nuance Unveil ChatGPT Successor for Healthcare
“Obviously there’s a lot of energy and a lot of anxiety,” said Dr. Greg Athor, chief medical informatician at the University of Kansas Health System. “People are just getting off their skis on some of these technologies.”
Instead, the early adoption of generative AI in healthcare is taking place in the less vibrant field of clinical medicine. take notes. Ator is part of a team implementing generative artificial intelligence technologies in the academic healthcare system to help clinicians take notes. The system works with Abridge, an artificial intelligence healthcare company, to summarize clinical conversations from recorded audio during patient visits.
Abridge’s generative artificial intelligence technology is similar to Nuance Communications, a clinical documentation software company owned by Microsoft. Last Monday, Nuance said it was adding Successor to OpenAI ChaptGPT GPT-4 to its latest application to be used in electronic health record systems.
In both cases, users must describe what they see in order for the software to work properly. For example, if a patient has a sore throat, in order for the program to enter information, he needs to verbally convey a specific comment about what the doctor sees.
In addition to entering relevant information into the EHR, both applications delete conversations that are not related to the service plan.
“These are power tools,” said Abridge co-founder and CEO Shiv Rao. “[Generative AI is] a powerful tool in the context of a much larger set of technologies that together make up a solution that can create value in the workflow.”
Medical records are a logical start, Ator said, because clinicians can quickly determine where AI-generated results came from. Clinicians can easily listen to the visit recording again if the AI misses valuable information.
“What you build next to, under, and above these foundation models like GPT-4 is the secret sauce,” Rao said. “There is a certain layer of technology that is now available to all of us, but how we integrate these tools into larger solutions will make the difference between a truly magical experience for physicians and their patients and solutions that seem off the shelf. toys”.
Investor interest remains strong
According to Rock Health, a venture capital firm focused on research and digital health, in 2022 investments in AI in healthcare amounted to $4.4 billion. Although last year the overall figure was down more than 50% compared to 2021, it was in line with 2020 levels.
The same data showed that 2021 was a record year with 224 deals for companies using AI technology. Although 2022 was not as fruitful, it was higher than 2020. While experts say 2021 levels won’t return anytime soon, this space is still of interest.
Skeptics say that while investment in AI remains strong, few of those investments are ready for widespread adoption.
“I think for some time we will still need to keep people informed because AI is far from perfect,” said Eric Brynjolfsson, director of the Digital Economy Lab at the Human-Centered Artificial Intelligence Institute, Stanford University. . . “He can’t do many things.”
Brynjolfsson said trained medical professionals can quickly reject abnormalities on a scan or medical image, while AI can make a wrong diagnosis. While there is the potential to eventually replace some of the roles of clinicians, experts say the human input is still critical.
Generative AI also takes a long time to set up and even in promising areas is not quite ready for prime time. During the summer, Nuance launches the GPT-4 feature.
Athor said the University of Kansas Health System has been rolling out the technology for months. He hopes that this will be completed by the end of the year, but did not want to give a specific time frame. This is largely due to the time it takes to train clinicians and the necessary integration with healthcare provider Epic’s EHR platform.
“Every time you work with a complex system like Epic, which is our core medical system, we have to interact with them. [the implementation] determined by their schedule,” Ator said.
Another potential barrier to adoption may be acceptance by the patient. A Pew Research Center survey conducted in December found that 60% of US adult patients would feel uncomfortable if their healthcare provider relied on AI for their healthcare. Less than a third believe that the quality of their care will improve after the introduction of AI.
Although the study did not specifically ask respondents about the analysis of audio recordings of their visits, the authors of the report found that “concerns about the pace of AI adoption” are widely shared in medicine.
Brynjolfsson said dictation and medical imaging are areas where vendors can improve processes. But he said the future of healthcare will still require medical and human oversight.
Others, however, are more optimistic about future adoption.
“What we’re seeing today is just a sign of what’s to come,” said Dr. Robert Pearl, former CEO of Kaiser Permanente in Oakland, California and current professor at Stanford University. “Everyone focuses on the mistakes of the day or the shortcomings of today. They don’t matter.”
This story first appeared in Digital Health Business & Technology.