Inside UW Health's generative AI pilot for nurse efficiency
UW Health nurses are piloting a generative AI tool that drafts responses to patient messages to improve clinical efficiency during ongoing staffing shortages.
As the largest group of healthcare providers, nurses are the backbone of the healthcare system. However, they have largely been left out of EHR optimization. A generative AI pilot project at UW Health is looking to change that.
"So much of the emphasis around technology has been to support our providers," said Rudy Jackson, chief nurse executive at UW Health. "That was done intentionally and appropriately. Having our colleagues working hours into the evening trying to complete documentation or finish up reports was an issue."
However, Jackson said that with ongoing nursing shortages expected to worsen, it is more important than ever to include nurses in health IT development.
According to data from the National Council of State Boards of Nursing, the national nursing workforce lost 100,000 registered nurses during the COVID-19 pandemic and is estimated to lose an additional 900,000 workers by 2027.
"We're having to think differently about our care models, our care delivery systems, how we utilize RNs and how we leverage technology to help offload some of those tasks so that it allows the nurses to spend more quality time with our patients," said Jackson. "We can't bury our heads in the sand and expect this to go by."
Streamlining workflows with generative AI
In April 2023, Microsoft and Epic launched a program at UW Health to trial a generative AI tool that drafts responses to patient messages among a small group of physicians. In September, the first group of nurses began trialing the tool.
Since then, more than 75 UW Health nurses have used the feature, collectively drafting over 3,000 messages across upwards of 30 departments.
The tool drafts responses to messages from the Epic patient portal, MyChart, using a large language model. A nurse reviews and edits each draft to ensure its accuracy and appropriateness before the system sends the message to the patient.
"Whether it's telephone calls or MyChart messages, a lot of that interaction is coming through the nurse, so it's just really cool that this is a tool that we can utilize to help with everyday work that we're doing," said Amanda Weber, registered nurse clinical supervisor at UW Health.
She noted that the AI-generated drafts have been a helpful starting point when replying to patients, allowing her to respond to patient portal messages more quickly.
"It generates specific recommendations, and then I can decide if that's appropriate or not," she said. "Because I work in a specialty clinic, the recommendations are sometimes really generalized, which is helpful, but then I can go in and add patient-specific information on what we find on chart review and what their diagnosis is and their medical history."
The solution is currently only available to nurses who have worked in the clinic for several years to ensure that AI-generated drafts do not sway newer nurses to reply to a patient message in a certain way, Weber said.
"We wanted the nurses that were using it to have a really good foundation in the clinic first," Weber noted.
Incorporating nurse feedback
The generative AI tool allows nurses to give feedback on drafted responses within the EHR, which the system then sends back to the EHR development team.
The better that we as a nation can support our nurses that are out there taking care of patients, the better they're going to provide care for our patients. It's just really as simple as that.
Rudy JacksonChief nurse executive at UW Health
"We have the option to provide direct feedback on what worked and what didn't work with particular responses that were generated," she explained. "Any improvements or recommendations that we have, we can input that right then and there."
"I think as we're getting more involved around utilizing the technology, it's just so incredibly important that nursing be part of the conversation during the development process to determine whether or not it's going to really benefit our workforce," said Jackson.
"There's nothing that could replace the voice of the caregiver because, at the end of the day, they really understand where all of those bottlenecks are," he added.
He also underscored the role of leadership in ensuring nurses' voices are included in the conversation as health IT continues to advance.
"As nursing leaders nationally, we have to be at the forefront having these discussions and talking about the implications so that when we roll out these systems, we're putting our best foot forward," Jackson said. "We really need to be involved on the front end with the actual design well before even the implementation of some of these tools."
UW Health is exploring other ways generative AI could streamline workflows for nurses, including drafting care plans and end-of-shift notes.
"The better that we as a nation can support our nurses that are out there taking care of patients, the better they're going to provide care for our patients," said Jackson. "It's just really as simple as that."
Hannah Nelson has been covering news related to health information technology and health data interoperability since 2020.
