How nurses are charting the future of AI at America's largest hospital network

Each day, across HCA Healthcare’s 190 hospitals and approximately 2,400 ambulatory sites of care, nurses repeat the same task around 60,000 times: passing patient-specific information from one nurse to another during shift change to help ensure care continuity and safety.

Even with the widespread digitization of electronic health records, many care workflows remain paper-based and time consuming.

Nurse handoffs, which are often completed using a clipboard at the patient’s bedside, are a crucial step to ensure uninterrupted care between provider teams, as one nurse passes along their carefully collected information to the next. Patient safety requires timely and relevant handoff communication, making these records much more than simple paperwork. Reliability, confidence, and accuracy are paramount in these handoffs, which underscores the importance of accurate and complete information.

“Depending on the types of patients and which nurses are giving the report, you could end up with a pretty big stack of papers,” explains Samantha Hall, a registered nurse at TriStar Hendersonville Medical Center, an HCA Healthcare hospital in Tennessee.

A few years ago, HCA Healthcare launched its Digital Transformation & Innovation (DT&I) department to look for ways technology could improve processes and workflows across its facilities. When the team surveyed HCA Healthcare nurses, the nurse handoff process was identified as a top area that could be improved.

At the time, generative AI was just beginning its massive popularity. Given the amount of information involved in nurse handoffs, this seemed like a natural place to deploy technology that could quickly ingest, analyze, and query information in a natural, conversational way. But care teams had to have confidence in the responses. As keepers of this critical communication, it was paramount for nurses to own and shape the nurse handoff product, scrutinizing the process to ensure responses to data requests met their high expectations.

“Nurse handoff communication is critical to any hospital and is fundamental to the role of registered nurses for safely exchanging communication and care about a patient,” said Dr. Whitney Staub-Juergens, vice president of Transformation Operations, DT&I. “It’s how they fundamentally get organized for the next 12 hours.”

Getting handoffs right with AI offered HCA Healthcare a huge opportunity to support better patient care. It typically takes a nurse at HCA Healthcare facilities approximately 40 minutes to complete the handoff for their shift. This adds up to an extraordinary 10 million hours in aggregate per year for all HCA Healthcare nurses.

“The smoother the handoff, the more time we have to spend with patients,” Hall said.

Partnering with care teams to build a solution

The technology industry prides itself on disrupting the status quo to catalyze change. But unless new technology makes clinicians’ jobs easier, with a measurable positive impact on patient outcomes, caregivers can be wary of adopting it. The last thing nurses and doctors want to do is spend more time in front of a computer screen, responding to alerts.

For the Digital Transformation & Innovation initiatives to work, HCA Healthcare leadership knew they had to start with the experiences of clinicians. To transform patient care, notes Dr. Michael Schlosser, senior vice president and chief transformation officer in charge of DT&I, the needs of caregivers must be prioritized, too, with their involvement in the creation of the necessary tools and workflows.

“One of DT&I’s guiding principles for digital transformation is to address the administrative burden on clinicians,” Dr. Schlosser said.

To achieve this quest with the least disruption to daily operations, HCA Healthcare built two Innovation Hubs in its hospitals staffed by engineers who work directly with caregivers. “You need a hospital that understands digital transformation,” Dr. Staub-Juergens said. “By creating a hub where care teams can test and learn, we keep the voice of the care team first and foremost.”

Collaborative testing and learning in a centralized care environment traces back to ancient times. This tried-and-true model has not only allowed doctors to treat multiple patients, but also to perfect their healing techniques on patients with similar ailments. With access to 21st century technology advances, DT&I’s vision is to transform the centralized healthcare environment into a self-learning, continuously improving partner for patient care.

“Imagine a room that’s sensing and scanning what's going on with the patient all the time, and providing insights to the care team,” Dr. Schlosser said. “Instead of the nurse having to spend 15 minutes every hour documenting the patient’s condition and then handing notes off to another nurse, the hospital room walls, and everything within, are watching the patients for us and collecting all that data in an intelligent way. Nurses and doctors always have the most up-to-date insights and can spend their time with patients instead of writing notes.”

How can healthcare make intelligent spaces that partner with care teams — spaces that care — a reality? With its ability to process structured and unstructured elements of the EHR, generative AI appears to be the perfect foundational technology. Facilitating the nurse handoff process is an initial step toward training AI foundation models and language models that can do far more to ease caregiver burden and free up time for patients.

“We wanted to train a large language model system to organize information like a nurse and figure out key pieces of information that it needs to extract from the chart, then organize them in an intelligent way that the oncoming shift could easily consume,” Dr. Schlosser said.

Nurses training an AI model to assist nurses

HCA Healthcare’s Digital Transformation & Innovation department started working on the Nurse Handoff tool with Google Cloud in early 2024. The first step was to train the AI models to generate meaningful content.

“We trained the model in partnership with frontline nurses, data science and product teams, and the Google team,” noted K.C. DeShetler, a registered nurse who is also the product owner for Nurse Handoff on the DT&I team. “We fed the model the information nurses want to know, prompted the model in a multitude of ways, used retrieval augmented generation to identify citations for the generated content, provided templates for organizing information the way we want, and so forth.”

The team developed a tool, named Nurse Handoff, that shows nurses the electronic health record on one side and the AI-generated output on the other. Using a hospital-provided mobile device, nurses are able to easily review and add information throughout their shift. Google’s MedLM models, running within the Nurse Handoff app, can ingest, analyze, and develop a cohesive and concise view of pertinent patient information for the oncoming nurse for handoff.

Traditionally, nurses rely upon their recollection of events, conversations, and data points, leaving room for error and inconsistency in the information given during handoff. Using Nurse Handoff during their shift, nurses can easily find the information they and their peers have collected through the automated shift report and add their own notes, so the system keeps building. This can include relevant patient data from notes, orders, tests, and more. All this happens within a highly secure cloud environment, to keep patient information confidential.

“We went through that process three or four different times,” Hall recalled. “And each time, it became a little bit more accurate, a little less filled with fluff that we don't need. The more we worked with it and provided our feedback, the more useful it became for the handoff setting.”

Giving nurses the voice in development

This human-informed AI training is key to the tool’s success and effectiveness. Not only must the AI model learn to provide the salient information, but interfacing with the solution must also feel natural to nurses, so that it will become an indispensable addition to their toolbelt. Using drag and drop widgets, nurses can arrange the user interface, or UI, in a way that aligns with how they get organized.

“This is getting creative, thinking differently from the UIs that the traditional hospital industry provides,” Dr. Staub-Juergens said. “It’s really important for adoption that we give the nurse a sense of, ‘this is my assistant.’”

The iterative design process has also provided valuable insights into how DT&I needs to approach change management, communications, and education as they plan the organizational rollout. The team started with acute care medical surgical units, because narrowing the use cases helped ensure the foundational model was delivering reliable results. They could then expand to more complex care areas in the hospital, adapting and monitoring the data sets accordingly.

“Participating in the evolution from idea to impact has been incredible,” DeShetler said. “Our frontline nurses have guided iterative development at every step. From data content and structure to care process delivery, nurses haven’t just had a voice at the table, they’ve had the voice.”

DeShetler adds that early adoption of the Nurse Handoff tool is promising, particularly for new nurses. “Our new graduate nurses have been particularly eager to use the new tool, telling us they feel more supported, more confident, and more prepared to perform what can typically be an intimidating process for a novice nurse.”

HCA Healthcare’s goal is to pilot the nurse handoff solution in five hospitals, gather additional feedback, then prepare the version they’ll roll out for use by all 99,000 nurses working across their system. As pilot testing continues for the Nurse Handoff tool, feedback from nurses at one HCA Healthcare facility have rated it as 86% factual and 90% helpful.

Achieving the vision, one process at a time

Given that HCA Healthcare’s DT&I department is aiming for maximum impact with minimum disruption, capturing input from care teams is crucial to success. “When I talk to direct care nurses and nurse leaders about technology and the emergence of AI in nursing,” Dr. Staub-Juergens said, “I often challenge them to be bold, be brave, take the keys to the car, get in the driver's seat, and use their voices to drive the design of the solutions moving forward.”

Dr. Staub-Juergens wants to see nurses at the table, partnering with other professionals to design meaningful solutions. “I’m exceptionally proud of our TriStar Hendersonville Medical Center nurses,” she said. “They're not technologists by trade. Most of them had to learn about AI, but they don’t have to worry about being technical experts. They can focus on what would make their job easier and better in the future."

Dr. Schlosser believes that working directly with physicians and nurses will yield many effective AI-driven solutions. But to succeed, it’s crucial to start with a determined, inclusive, and collaborative approach, plus the right expertise and technologies. The breakthroughs will follow from there.

“I think AI is a completely transformative technology, but it's going to take a bit of hard work to transform our processes to make AI useful,” Dr. Schlosser said. “I think it's 100% doable — there's no massive engineering breakthrough that has to occur in order for us to make a vision like a room that cares come to fruition. If we focus on it, we prioritize it, and we measure it, I have confidence that we will solve it.”