The use of generative AI may not help reduce burnout in the medical field, new research suggests.
Previous research has shown that increasing time spent using electronic health record (EHR) systems and handling administrative responsibilities is burdening physicians.
So while some have heralded artificial intelligence as a potential solution, a recent study by the US healthcare system found that large-scale language models (LLMs) are simplifying clinicians’ daily tasks. It turns out there isn’t.
What is artificial intelligence (AI)?
For example, a 2023 observational study conducted at Brigham and Women’s Hospital in Boston, Massachusetts, investigated the impact of using AI for electronic patient messaging.
The researchers created a large-scale language model to respond to mock questions from cancer patients and compared its output with responses from six board-certified radiation oncologists.
Medical professionals then edited the AI-generated answers into “clinically acceptable” answers and sent them to the patient.
New research suggests that generative AI may not help with physician burnout, as previously thought. (St. Petersburg)
The study, published in The Lancet Digital Health, found that the draft LLM caused “a risk of serious harm in 11 out of 156 survey responses and a risk of death in 1 survey response.” There was found.
“Adverse reactions are largely due to misjudgment or miscommunication of scenario accuracy and recommended actions,” the researchers wrote.
First augmented reality abdominal surgery performed in Chile: ‘Revolution’
The researchers say that the LLM-assisted results (compiled by physicians) represent a “best-of-both-worlds scenario” that reduces the burden on physicians while ensuring patients receive accurate information. I concluded that there is.
“These early findings…demonstrate the need to thoroughly evaluate LLMs in their intended clinical context, reflecting the precise task and level of human supervision,” the study concluded.
The researchers concluded that the LLM-supported results represent a “best-of-both-worlds scenario” that reduces physician burden while ensuring consistency of response and improving patient education. (St. Petersburg)
medical billing code
Another study from New York’s Mount Sinai Health System was evaluated Four different types of large language models for performance and error patterns when querying medical billing codes.
Google Bard goes Gemini: What you need to know about the AI upgrade
The study, published in the journal NEJM AI, found that all LLMs tested performed poorly when querying medical codes, “often producing codes that convey inaccurate or fabricated information.”
The study concluded that “LLM is not appropriate for use in medical coding tasks without additional research.” This research was funded by the AGA Research Foundation and the National Institutes of Health (NIH).
One study found that all tested LLMs performed poorly on medical code queries, and that this issue requires further research. (St. Petersburg)
The researchers noted that while these models can “approximate the meaning of many codes,” they “also exhibit an unacceptable lack of accuracy and a high propensity to tamper with the code.”
“This has important implications for billing, clinical decision-making, quality improvement, research, and health policy,” the researchers wrote.
Message to patients and doctor’s time
A third study published by JAMA Network, from the University of California, San Diego School of Medicine, evaluated AI-drafted replies to patient messages and the time doctors spent editing them.
Study finds widespread inaccuracy when answering medication questions
The assumption was that generative AI drafting would reduce the amount of time physicians spend on these tasks, but the results showed that this was not the case.
“Generated AI replies are associated with a significant increase in read time, no change in reply time, and a significant increase in reply length. [only] “Several benefits have been observed,” the study found.
The researchers suggested that “rigorous empirical testing” is needed to further evaluate the AI’s performance and patient experience.
A study from the University of California, San Diego, found that generative AI causes “increased read times, no change in response times.” [and] The length of patient message replies has been significantly increased. (St. Petersburg)
Doctors’ thoughts on AI
David Atashloo, MD, Chief Medical Officer of Qventus, an AI-powered surgical management solution in Mountain View, California, reacted to the study’s findings in an interview with Fox News Digital. (He was not involved in the study.)
“We are expanding our portfolio of low-risk, highly automatable tasks such as schedulers, medical assistants, case managers, and care navigators, who have traditionally served as essential but often overlooked ‘glue roles’ in healthcare. “We believe there is tremendous potential for AI,” he said.
“It’s important to set realistic expectations.” [AI’s] performance. ‘
“These professionals are critical to orchestrating processes that directly lead to clinical outcomes, but they also spend a significant amount of their time performing administrative tasks such as parsing faxes, summarizing notes, and securing necessary documentation. ”
Click here to sign up for our health newsletter
In automating these tasks, Atashroo suggested that generative AI could help improve the efficiency and effectiveness of clinical care.
“When considering the implementation of generative AI, it is important to set realistic expectations about its performance,” he said.
“Even the people doing these tasks today aren’t perfect, so standards aren’t always perfect.”
“Standards are not always perfect because even the humans currently performing these tasks are not reliable,” the AI expert said. (St. Petersburg)
He suggested that in some scenarios, AI could act as a “safety net” to catch team members’ oversights.
Atashloo pointed out that sometimes challenges go unresolved “simply because we don’t have the time to work on them.”
“Generative AI will help us manage cases more consistently than our current capacity.”
“When considering the implementation of generative AI, it is important to set realistic expectations about its performance.”
Safety and efficacy are “of paramount importance” in AI applications, the doctor said.
“This means not only developing the model with rigorous quality checks, but also incorporating regular evaluations by human experts to verify its performance,” he said.
CLICK HERE TO GET THE FOX NEWS APP
“This double layer of validation ensures that our AI solutions are both reliable and trustworthy before being scaled.”
Atashloo also said that “transparency in the development and implementation of AI technology is essential to building trust between hospital partners and patients.”
For more health articles, visit: www.foxnews.com/health.
