Exclusive: Hippocratic AI launches Nurse Advisory Council for LLM safety

Hippocratic AI, a generative AI company developing safety-focused large language models for healthcare, is launching a Nurse Advisory Council that will work in conjunction with its Physician Advisory Council to ensure the safe development of its LLMs aimed at increasing care equity, the supply and availability of healthcare workers, and healthcare access and outcomes.

The California-based company’s LLM concentrates on non-diagnostic, patient-facing applications, such as certification, ensuring proper bedside manner for patient well-being and reinforcement learning from human feedback (RLHF) through a panel of healthcare professionals.

“Clinical input and clinical involvement has been part of the company since day one,” Dr. Meenesh Bhimani, cofounder and chief medical officer at Hippocratic AI, told MobiHealthNews. “The Nurse Advisory Council is really the next step in this.”

“They will be deeply involved in shaping the product itself, fine-tuning the product, but also testing it for safety, identifying any areas that they feel we obviously need to improve or work on, and really giving their unique insight into not just the patient experience, but the care pathways and the use cases. They’ll serve as a resource for us on clinical education and workflows, but really focus on the safety and quality perspective of it.”

The Nurse Advisory Council consists of the following members:

• Erica Bentley, vice president of population health management at Optum

• Serena Bumpus, CEO of the Texas Nurses Association 

• Shawna Butler, nurse economist and managing director at NextMed Health

• Cassie Choi, cofounder and chief health equity and people officer at Pair Team and member of the White House Clinician Innovators Roundtable on Health Equity

• Kenneth Dion, founding principal at TurnPath, LLC

• David Marshall, senior vice president and chief nursing executive at Cedars-Sinai

• Amy McCarthy, director of nursing, women, infants and oncology at Texas Health Resources; president-elect of Texas Nurses Association

• Molly McCarthy, global GTM excellence leader, enterprise informatics at Philips

• Colleen Morley, Case Management Society of America national president

• Patricia Railsback Masson, a nurse scientist at Massachusetts General Hospital and associate scientist at Broad Institute of MIT and Harvard

“We’re going to be asking a lot of questions,” Shawna Butler, an RN on Hippocratic AI’s Nurse Advisory Council, told MobiHealthNews. “We understand what problems we’re trying to solve, so how are we using these tools to actually solve them? And so I think a lot of times what we will be doing as nurses is bringing nursing expertise, wisdom, practical insights and to be testing a lot of that.”

Butler emphasizes that the Nurse Advisory Council will be experts on safety and ensure physical and emotional harm is not done to anyone. 

“The most important thing that we’re doing is, first of all, looking at the datasets because a big piece of what we’re looking at is, where are those datasets from? Who is in them? Who is not in them? And even when you have people in them, are their lives accurately and fairly represented? Is there a degree of bias or maybe some prejudice?” Butler said.  

Still, she emphasizes the task at hand, creating the datasets, is hard work. 

“Even if we get it right for 90% of the people, it’s not okay to get it wrong for 10% of the people,” she said. 

To ensure the datasets are analyzed from multiple perspectives, the Nurse Advisory Council will work with Hippocratic AI’s Physician Advisory Council, announced in October last year. 

“We’re hoping that both Councils will work synergistically. Nurses are the heart and soul of our healthcare ecosystem. They’re very deeply connected to patients, and so their insight and the perspective that they’re going to bring is going to be crucial to our real goal of building a safety-focused model,” Bhimani said.  

“And really, our goal is to augment the challenges that we’re finding in the nursing workforce today and not disrupt it, so we need their insight and input to kind of bring that reality to life.”

Butler says the Councils will co-create to provide each other’s various perspectives in caring for the patient.  

“How does [the tool] actually unfold to make sure that all of the touch points, all of the interactions work seamlessly for all of the different team members, workflows, and responsibilities?” Butler said. 

“We’re gonna do a lot of talking, we’re gonna have a lot of, I don’t know if it would be a debate, but I think it is getting inside of each other’s worlds to really understand how care unfolds for people in the different care pathways?” 

Ultimately, the goal of all parties, Bhimani and Butler agree, is to ensure generative AI is leveraged safely to redefine the standard of care in healthcare delivery. 

“Our goal from the beginning has been to develop a safety-focused large language model for healthcare, and one that we can then use in a patient-facing application,” Bhimani said. “We think we can dramatically increase and improve healthcare access, equity and outcomes by using generative AI.”