NewYork-Presbyterian Developing Innovative AI Tool

‐

By Christopher Cheney

NewYork-Presbyterian is involved in a unique effort to develop an artificial intelligence tool for the early detection of cardiovascular disease.

HealthLeaders is conducting its AI in Clinical Care Mastermind program through December. The program brings together nearly a dozen healthcare executives to discuss their AI strategies and offerings.

One of the advantages of NewYork-Presbyterian is that it is affiliated with two medical schools, Columbia University Vagelos College of Physicians and Surgeons as well as Weill Cornell Medicine, says Ashley Beecy, MD, medical director of AI operations at NewYork-Presbyterian.

“There are research teams across the enterprise developing AI models and working on translational research to bring the models to point of care,” Beecy says. “They also conduct clinical trials to understand both the factors for the safe use of AI and the efficacy of the models when integrated into the healthcare system.”

A team comprised of clinicians and researchers from Columbia University Irving Medical Center is developing an AI tool for the early detection of cardiovascular disease.

“There is a lab at Columbia, CRADLE (Cardiovascular and Radiologic Deep Learning Environment), led by Dr. Pierre Elias, doing incredible work looking at different types of cardiovascular data for early detection of disease,” Beecy says. “For example, an AI tool can use electrocardiograms (EKGs) to identify potential markers for structural heart disease that a cardiologist reading an EKG can’t identify with the naked eye. Depending on the score the AI model generates, it will prompt a cardiologist to order an echocardiogram to confirm diagnosis.”

Additionally, NewYork-Presbyterian’s Enterprise Heart Failure Program, led by Dr. Nir Uriel, with physicians from its affiliated medical schools, is collaborating with Cornell Tech and the Cornell Ann S. Bowers College of Computing and Information Science to transform cardiovascular health and heart disease prediction and prevention using AI and machine learning.

Assistive AI tools

In addition to the innovative cardiovascular disease detection AI tool, NewYork-Presbyterian is rolling out several assistive AI tools that are impacting clinical care, according to Beecy. A few examples include:

  • Ambient scribe: This AI tool has been focused on the outpatient setting. The tool transcribes a conversation between a clinician and a patient, then generates a draft clinical note to document the interaction. “One of the interesting findings so far is that ambient scribe is saving a few minutes for clinicians, but when you survey them, clinicians feel strongly that ambient scribe reduces their documentation time,” Beecy says. “This reflects changing documentation from a writing task to an editing task, which reduces the cognitive load.”
  • Risk prediction: NewYork-Presbyterian is using an AI tool that alerts nurses about potential fall risk for inpatients at the health system’s Lower Manhattan Hospital. “We are getting feedback from nurses on whether they find the AI tool’s alerts useful and whether the alerts are preventing falls,” Beecy says. “We want to work closely with nursing teams as stakeholders to ensure that the AI tool’s alarms are not interrupting their workflow. We want to make sure it is augmenting their work in a positive way.”
  • Radiology: The health system is using an AI tool to triage radiology images. One example is the use of AI for identification of stroke, including possible intracranial hemorrhage. The goal is to notify care teams for early intervention.
  • In-basket messaging: This is one of the first AI tools adopted at NewYork-Presbyterian, and it helps clinicians respond to patient questions in their electronic in-boxes. “It is very similar to ambient scribe,” Beecy says. “People feel as though it is saving them time since the task moves from writing to editing. The metrics for time savings are variable, and it will be interesting to see whether the in-basket messaging tool reduces clinician burnout over the long term.”

For each assistive AI tool, there is still a human in the loop, Beecy assured.

For clinical care teams, right now AI is something unique and independent of the way clinicians currently practice medicine, Beecy explains.

“In the long term, we are going to find that AI becomes ubiquitous in the way we practice medicine,” Beecy says.

In the short term, the focus of AI tools in clinical care at the health system is on administrative tasks and diagnostic accuracy, according to Beecy.

“This allows us to focus on the human input and the patient interactions for care processes, without radically changing how we work,” Beecy says.

Over time, more data will be digitized and transformed into useful data sets that can be harnessed by AI tools, Beecy explains.

“We will generate insights from this data such as wearables and digital pathology,” Beecy says. “AI systems will become more of a part of our decision-making process, and we will see more algorithmically guided care.”

Christopher Cheney is the CMO editor at HealthLeaders.