'I don't ever trust Epic to be correct': Nurses raise more AI concerns

Nurses continue to voice concerns about artificial intelligence and its integration into EHRs, saying the technology is ineffective and interferes with patient care.

Nurses from health systems around the country spoke to National Nurses United, their largest labor union, for a June 5 story about issues with such programs as automated nurse handoffs, patient classification systems and sepsis alerts.

Multiple nurses cited problems with EHR-based programs from Epic and Oracle Health that use algorithms to determine patient acuity and nurse staffing levels.

"I don't ever trust Epic to be correct," Craig Cedotal, RN, a pediatric oncology nurse at Kaiser Permanente Oakland (Calif.) Medical Center, told the nurses' union. "It's never a reflection of what we need, but more a snapshot of what we've done."

He said the technology does not account for the hours of preparing and double-checking the accuracy of chemotherapy treatments before a pediatric patient even arrives at the hospital.

"In Epic, AI is a tool to help nurses complete documentation and patient messaging quicker, spend more time with patients, and continue to provide high-quality care," a company spokesperson told Becker's. "In 2023, nurses across the Epic community used AI-powered drug interaction checks to prevent more than 60 million potentially adverse events. Nurse input is at the core of our development and implementation processes from the earliest stages, and we encourage customers to have nurses guide their ongoing use of the software."

Hundreds of nurses protested AI at Kaiser Permanente San Francisco Medical Center in April.

“Kaiser Permanente is empowering nurses with state-of-the-art tools and technologies that support our mission of providing high-quality, affordable healthcare to best meet our members' and patients' needs," a spokesperson for the Oakland-based health system emailed Becker's. "We have consistently invested in and embraced technology that enables nurses to work more effectively, resulting in improved patient outcomes and nurse satisfaction, and we will continue to do so."

"As an organization dedicated to inclusiveness and health equity, we ensure the results from AI tools are correct and unbiased," the spokesperson continued. "AI does not replace human assessment."

When Los Angeles-based Keck Medicine of USC revealed plans to implement Oracle Health's Clairvia acuity system, nurses threatened to picket if the health system didn't leave decisions on patient acuity and classification to bedside nurses, according to the article. These nurses said Clairvia was far less transparent and clear on how patients are classified compared to the old system. Keck leadership issued a memo that only a licensed registered nurse could make the final call on nursing hours needed.

"In our ongoing commitment to quality and safety, we continually evaluate new technologies that can simplify administrative tasks, thus reducing staff burnout and allowing for more time with patients," the health system said in a statement to Becker's.

"Keck Medicine, like many other health institutions, leverages evidence-based software to help inform patient acuity using health data from the patient's care team. However, the patient’s needs and acuity are ultimately determined by the assessment of a licensed registered nurse. We also collaborate closely with nursing staff throughout implementation to gather ongoing feedback to ensure we're best meeting caregiver needs."

A nurse with New York City-based NewYork-Presbyterian told the union these systems don't account for the time nurses spend on the type of tasks that aren't easily documented, such as patient education and "compassionate, psychosocial care." The health system declined to comment. Oracle Health didn't respond to a request for comment.

Automated nurse handoffs, which detail information about patients that the next nurse reads before starting a shift, are increasingly popular in healthcare. Jeff Breslin, RN, a float pool nurse at University of Michigan Health-Sparrow in Lansing, Mich., and president of the Michigan Nurses Association, told National Nurses United he once caught a patient's medication missing from an automated handoff that could have been harmful if he hadn't found the drug noted elsewhere in the chart. "It's better to have an actual conversation with someone who has taken care of the patient," he said in the article. "It takes some of the guesswork out of receiving a new patient by having that face-to-face, human-to-human interaction."

"University of Michigan Health-Sparrow currently does not have any generative AI tools and believes organizations must have guiding principles and a responsible approach when using technology in healthcare," Josh Wilda, regional chief digital information officer for University of Michigan Health, told Becker's in an emailed statement.

"These technologies can assist team members by managing information and administrative tasks, allowing more focused patient interactions. However, it's crucial that AI tools serve as assistants in the decision-making process, not stand on their own in providing care and services. As UM Health-Sparrow introduces new, innovative technology to our health system, we will ensure human connections with our patients are at the center of it all."

A nurse with Sacramento, Calif.-based UC Davis Health said sepsis alerts in the Epic EHR are often incorrect, missing patients with the condition and providing false warnings for patients unlikely to have sepsis, according to the article. "Everybody on my unit knows the sepsis warning system is not helpful," said Melissa Beebe, RN, an oncology nurse at UC Davis Medical Center in Sacramento and member of the California Nurses Association/National Nurses Organizing Committee. "I've sent people to ICU and that thing never went off."

The article pointed to a 2023 JAMA Network Open study that found the program missed two-thirds of sepsis cases. An Epic spokesperson told Becker's for an April story: "Last fall, we released an updated version of the sepsis predictive model and we are working with our customers to implement it. The live organizations have seen more timely alerts and fewer false positives. The JAMA study does not reflect the performance of our updated model."

A UC Davis Health spokesperson told Becker's the platform is "not a 100% indicator the condition is happening or going to happen."

"This program provides a probability the condition might happen, with clinical workflows designed to then be followed by humans, using that prediction as a starting point for further clinical assessment and clinician-based decisions," the spokesperson said. "This particular condition can happen quickly, and is frequently fatal, so an early alert that it is possible is in the best interests of patients and any clinician who actually cares about patient outcomes."

Copyright © 2024 Becker's Healthcare. All Rights Reserved. Privacy Policy. Cookie Policy. Linking and Reprinting Policy.


Featured Whitepapers

Featured Webinars