Join an event that enterprise leaders have been trusted for nearly 20 years. VB Transform brings together people who build real enterprise AI strategies. learn more
How about chatting in a way that can be done with Health Records and ChatGpt?
This question, initially posed by medical students, caused the development of Chate. Stanford Health Care. Currently in production, the tool accelerates chart reviews for emergency room admissions, streamlines patient transfer overviews, and integrates information from complex medical history.
Early pilot results showed clinical users significantly increased their search for information. In particular, emergency doctors have reduced chart review times by 40% during critical handoffs, Stanford’s SVP and Chief Information Officer Michael A. Pfeffer said in today’s Fireside Chat. VB conversion.
It is based on decades of work health facilities to help reduce doctor burnout while improving patient care and collect and automate critical data.
“It’s a very exciting time in healthcare, as we’ve digitalised and put it in our electronic health records for the past 20 years, but it’s not really transforming,” he said in a chat with VB editor Matt Marshall. “With new, large-scale language modeling technology, we are actually beginning to do that digital transformation.”
Return to true face-to-face interactions, how Chatehr helps reduce “Pajama Time”
Doctors spend up to 60% on administrative tasks rather than directly caring for patients. They often have important “Pajama time“Sacrifice Individual and family time to complete administrative tasks other than normal working hours.
One of the big goals of Pfeffer is to streamline workflows and reduce those extra time so clinicians and administrative staff can focus on more important tasks.
For example, much information is provided through an online patient portal. AI now has the ability to read messages from patients and draft responses that humans can review and approve of their submissions.
“It’s kind of a starting point,” he explained. “It doesn’t necessarily save time, but it’s interesting, but actually reduces cognitive burnout.” He also said that messages tend to be more patient, as they can tell the model to use a specific language.
Moving on to agents, Pfeffer said they are a “fairly new” concept of healthcare, but offer promising opportunities.
For example, patients with cancer diagnosis usually have a team of experts who review records and decide on the next treatment procedure. However, preparation is a difficult task. Clinicians and staff should pass information across patient records not only on EHRs but also on imaging pathology, in some cases genomic data, and clinical trials that may suit the patient well. All of this must be together for teams to create timelines and recommendations, Pfeffer explained.
“The most important thing you can do for a patient is to make sure the patient is receiving proper care and requires an interdisciplinary approach,” Pfeffer said.
The goal is to build agents in Chatehr, which can generate overviews and timelines and create recommendations for clinician reviews. Pfeffer emphasizes that it will not be replaced and prepares “incredible overview recommendations in a multimodal way.”
This allows medical teams to currently provide “actual patient care.” This is important amidst the shortage of doctors and nursing.
“These technologies will change the time doctors and nurses do administrative tasks,” he said. And when combined with the surrounding AI scribes who take over their note-taking duties, medical staff concentrates more time on patients.
“That face-to-face interaction is valuable,” Pfeffer says. “We’re going to see AI move more towards clinician-patient interactions.”
“Amazing” technology combined with interdisciplinary teams
Before Chatehr, the Pfeffer team deployed Securegpt to everything in Stanford Medicine. The secure portal has 15 different models that anyone can tinker with. “What’s really powerful about this technology is that it really opens up to so many people to experiment,” says Pfeffer.
Stanford takes a variety of approaches to AI development, building its own models and uses a mix of secure, private off-shelf (such as Microsoft Azure) and open source models when needed. Pfeffer explained that his team is “not entirely specific” to either one, but suits the one that suits the best for a particular use case.
“There are so many amazing kinds of technologies now, so if we can connect them in the right way, we can get a solution like what we built,” he said.
Another credit for Stanford is the multidisciplinary team. Pfeffer has gathered Chief Data Scientists, two informaticsists, Chief Medical Information Officer, Chief Nursing Information Officer, CTO and CISO in contrast to AI Chief Officer or AI Group.
“We combine informatics, data science, traditional IT and wrap it around architecture. What you get is this magical group that can do these very complex projects,” he said.
Ultimately, Stanford views AI as a tool that everyone should know how to use, Pfeffer emphasized. Different teams need to understand how to use AI. This means that when you meet up with business owners and come up with ways to solve problems, “AI is part of their way of thinking.”
