Dr. Kaeli Yuen, Veterans Affairs artificial intelligence product lead within the VA Office of Information and Technology’s Office of the Chief Technology Officer, provided an update on the agency’s review of several AI pilots at Federal Insights from an American Council of Technology-Industry Advisory Council. Exchange meeting in Washington, DC, this week.
WHY IT MATTERS
Yuen said that while many veterans are skeptical about recording the doctor’s visits, according to a report by MeriTalka public-private partnership, doctors are enthusiastic about.
“It’s pretty incredible,” added Dr. Susan Kirsh, VA deputy assistant secretary of health for discovery, education and affiliate networks, added.
Yuen said the VA is working to balance the “privacy and security that veterans expect, while providing them with the convenience of experience.”
The forum brings government employees together to improve government outcomes. On May 29, ACT-IAC hosted Yuen and Kirsch for an FIE event focused on the VA’s journey to improve care for veterans using AI.
In 2021, more than two dozen VA offices developed the agency’s AI strategy, becoming one of the first federal agencies with an AI roadmap, leading to testing of AI for use during clinical encounters and for benefits administration.
Yuen said many offices within the VA are requesting genAI tools to draft correspondence and documents and summarize information, such as the veteran user experience survey data, the story said.
“People from all over VA want this,” she said.
However, experimenting with generative AI interfaces to help address the VA’s overall administrative burden is challenged by how to measure success with pre-existing workflows, Yuen said.
“I think we’re applying the tools to a process that was built without those tools in mind and maybe there’s a different way we should be doing things,” she said.
THE BIG TREND
The burden of clinical documentation is at the top of physicians’ minds, said Whende Carroll, clinical informatics consultant at HIMSS, parent company of Healthcare IT newsat the HIMSS23 Global Conference and Exhibition.
The fatigue it causes among doctors, nurses and other health care workers can worsen side effects, increase errors and result in “complications” that cause poor patient outcomes, she said.
While Carroll said AI, and natural language processing in particular, can reduce waste and help with medical documentation productivity that causes physician fatigue, genAI has been found to result in lengthy and overly complex reports and increased risk on errors in government reporting.
Srini Iyer, chief technical officer at Leidos Health & Civil Sector, said ahead of his HIMSS24 presentation on designing genAI to maximize benefits for healthcare organizations that while testing Google’s Med-PaLM 2 – a chatbot piloted by the Mayo Clinic and questioned others who federal lawmakers have used — his team encountered challenges.
They focused on the three top needs of healthcare executives for testing use cases, Iyer said.
By using the Leidos vector store, the researchers achieved better results when testing Med-PaLM 2 for accuracy, he noted. Using the specialized storage was ideal because it allowed the large language model to find relationships between unstructured data points – allowing it to remember those relationships over time, he explained.
The amount of information and complexity can increase the risk of errors, such as in mandatory reporting by regulators, “potentially impacting patient care and reimbursement,” Iyer said.
ON THE RECORD
“I think I can speak for the local communities, we want to spend all our time taking care of the patients,” Kirsch said.
“And the documentation has become quite high over the years. So this is something that is transformative.”
Andrea Fox is editor-in-chief of Healthcare IT News.
Email: afox@himss.org
Healthcare IT News is a HIMSS Media publication.