Stanford’s ChatEHR allows clinicians to query patient medical records using natural language, without compromising patient data

How would it be to discuss with health files as we could do with Chatgpt?

Initially posed by a medical student, this question sparked the development of Chatehr to Stanford health care. Now in production, the tool accelerates the exams of the graph for emergency admissions, rationalizes patient transfer summaries and synthesizes information from complex medical stories.

In the first pilot’s results, clinical users experienced considerably accelerated information; In particular, emergency doctors have seen 40% reduction in the revision time of graphics during critical transfers, Michael A. Pfeffer, SVP of Stanford and principal officer of information and digital, said today in a cat by the fire at the VB transform.

This helps reduce the professional exhaustion of doctors while improving Patient careAnd is based on decades of work that medical facilities have made to collect and automate critical data.

“This is such an exciting period in health care because we have spent the last 20 years digitizing data on health care and putting it in an electronic health file, but not really by transforming it,” said Pfeffer in a conversation with the editor of VB, Matt Marshall. “With new models of model of great language, we are actually starting to make this digital transformation.”

How chatehr helps reduce “pajamas time”

Doctors spend up to 60% of their time on administrative tasks rather than care for direct patients. They often put important things “pajamas“, Sacrifice Personal and family hours to perform administrative tasks outside of regular work hours.

One of the major Pfeffer objectives is to rationalize workflows and reduce these overtime so that clinicians and administrative staff can focus on greater work.

For example, a lot of information arises through online patient gates. AI now has the ability to read patient messages and write answers that a human can then examine and approve for sending.

“It’s a bit like a starting point,” he said. “Although it does not necessarily save time, which is interesting, it in fact reduces cognitive exhaustion.” In addition, he noted, messages tend to be more suitable for patients, as users can ask the model to use a certain language.

Passing to the agents, Pfeffer said that they were a “fairly new” concept in health care but offer promising opportunities.

For example, patients with cancer diagnosis generally have a team of specialists who examine their files and determine the next treatment stages. However, preparation is a lot of work; Clinicians and staff must go through the whole file of a patient, not only their DSE, but also imaging pathology, sometimes genomic data and information on clinical trials for which patients could be a good correspondence. All these elements must meet so that the team creates a calendar and recommendations, said Pfeffer.

“The most important thing we can do for our patients is to ensure that they have appropriate care, and it adopts a multidisciplinary approach,” said Pfeffer.

The objective is to build agents in Chatehr who can generate a summary and a calendar and make recommendations for the examination of clinicians. Pfeffer stressed that he does not replace, he prepares “just incredible summary recommendations in a multimodal way”.

This allows medical teams to now do “real care for patients”, which is essential in the midst of a shortage of doctors and nursing.

“These technologies will change as long as doctors and nurses spend administrative tasks,” he said. And when combined with Scribes AI AMBIANT Who take control of note tasks, medical staff concentrate more time on patients.

“This face -to -face interaction is simply invaluable,” said Pfeffer. “We are going to see AI move more towards the clinical-patient interaction.”

“Amazing” technologies coupled with a multidisciplinary team

Before Chatehr, Pfeffer’s team deployed Securegpt to all Stanford drugs; The secure portal includes 15 different models with which anyone can tinker. “What is really powerful in this technology is that you can really open it to so many people to experiment,” said Pfeffer.

Stanford adopts a varied approach to AI development, building its own models and uses a mixture of secure and private models (such as Microsoft Azure) and open source models, if applicable. Pfeffer explained that his team is “not completely specific” to one or the other, but rather goes with what will probably work best for a specific use case.

“There are so many incredible types of technologies now that if you can reconstruct them in the right way, you can get solutions like what we have built,” he said.

Another credit in Stanford is his multidisciplinary team; Unlike an AI leader or an AI group or an AI group, Pfeffer has gathered a data scientist, two computer scientists, a medical information director and a director of nursing information, and their CTO and CISO.

“We bring together IT, data science and traditional IT, and wrap in architecture; what you get is this magic group that allows you to do these very complex projects,” he said.

In the end, Stanford considers AI as a tool that everyone should know how to use, Pfeffer said. Different teams must understand how to use AI so that when they meet business owners and offer means to solve problems, “AI is only part of the way they think.”