Research

Doctoral Students Aim to Make EHRs More Responsive to Nurses

July 2018Andrew Schwartz

As electronic health records (EHRs) have become ubiquitous, as the technology has become more sophisticated and as clinicians have become used to – resigned to, in some cases – the presence of EHRs, researchers and clinicians have turned their attention to figuring out how to fully achieve the technology’s promise.

Clinician burden is among the most prominent stumbling blocks, but most of the research has focused on the physician burden, despite nurses’ daily need to assess information from EHRs to make frontline clinical decisions. Two doctoral students from the UCSF School of Nursing – Kirsten Wisner and Daniel Linnen – aim to change the balance.

Feeling the Disruption

Kirsten Wisner Wisner’s interest in the topic began in 2011, shortly after she received her master’s degree as a perinatal clinical nurse specialist (CNS) from the School. While she was working at Salinas Valley Memorial Healthcare System as a labor and delivery nurse – where she still works and is the facility’s Magnet program director – the hospital went live with its EHR.

“I found it to be extremely disruptive in terms of staying on top of things with patients,” she says. She notes that labor and delivery is typically a nurse-managed model of care, where nurses have a lot of autonomy, responsibility and accountability as they navigate the tension inherent in monitoring normal, uncomplicated births while also having to decide if a situation is complicated enough to demand a consult or engage the team to escalate a clinical concern. Such decisions often depend on nurses’ ability to understand the significance of an array of clinical findings, and Wisner was concerned that the EHRs were making it harder.

In particular, she says, flow sheets in patients’ paper charts used to give nurses a quick sense of what mattered. Now, to understand the hierarchy of concerns, clinicians have to cognitively create that picture themselves from what can be an overwhelming array of EHR data. “Often this is in situations that are very dynamic and fast-paced, including emergent situations, where we need to see a story that helps us project and anticipate multiple potential clinical trajectories.”

She brought her concerns to the School’s Audrey Lyndon, Wisner’s advisor from her master’s program, who is a national expert on communication in labor and delivery nursing and now chair of the School’s Department of Family Health Care Nursing. After finding few studies on how EHRs affected the quality of nursing care and no studies in labor and delivery, Wisner decided to return to the School to begin her doctoral program in 2014, so she could research those effects.

“By then, the literature was growing about EHRs’ unintended consequences, but the studies were mostly around physicians and computerized order entry,” says Wisner, who is about to begin her dissertation. The studies that touched on nursing were mostly time series studies around workflow and handoffs; they did not address the cognitive work of nurses.

Pilot Study Leads to Focus on Clinical Grasp

Consequently, in year two of her PhD degree program, Wisner conducted a qualitative pilot study to describe and convey what nurses were experiencing. Her interviews with labor and delivery nurses explored their perceptions about how the EHR affected their work, and the findings centered on nurses’ ability to see and track how their patient was doing and to understand and communicate that – particularly when transferring care or in the case of a sudden emergency.

Wisner says the nurses she spoke with were concerned about losing track of how the patient was doing, about the failure of electronic fetal monitoring systems and EHRs to interface effectively and how the amount of additional work EHRs create steals time from direct patient care. Nurses in the study also talked about the importance of their interactions with patients and their support persons, and how the EHR interfered with their perceptual, emotional and intuitive work. Wisner found herself searching for a framework that could help her understand and explain why this kind of interference might be clinically significant. Ultimately, she landed on the idea that EHRs, in their current form, detract from what Professor Emerita Patricia Benner famously described as “clinical grasp.”

“I’m proud to be at UCSF surrounded by such legacy,” says Wisner. “I land on clinical grasp, and Patricia Benner is there. My qualifying exam chair, Kit Chesla, was on Benner’s research team, and she worked closely with me during the literature review and helped me frame the importance of what I was finding. Likewise, Audrey Lyndon is my dissertation chair and advisor; I consider her the most influential mentor in my life. And I’m working with grounded theory, which was founded at UCSF as well.”

Nursing’s influence and professional knowledge may not be captured effectively in the EHR. One of Wisner’s qualifying exam papers was a literature review, which uncovered 18 studies that described incidental findings that bore out her concern about the effect on clinical grasp. “Clinicians described loss of overview of the patient, increased cognitive work while navigating the EHR, loss of individual and team situation awareness, issues with EHR-generated summary tools and concerns over changes to face-to-face communication.”

It was also disturbing that while nurses still read physicians’ narrative notes, physicians often lose track of nurses’ notes – and so nursing’s influence and professional knowledge may not be captured effectively in the EHR. Findings from her literature review and pilot study informed the design of Wisner’s dissertation study, which is in the Institutional Review Board (IRB) site approval process and which will use grounded theory to look intensely at the effect of EHRs on the cognitive work of nurses.

None of this is to imply that Wisner believes a return to paper records would be an improvement. Rather, she says, if she finds that EHRs compromise nurses’ clinical grasp and can define how that occurs, those findings could inform the next generation of EHR design and evaluation.

Zeroing In On Precision Risk Profiles for Nursing Care

That, perhaps, is where Linnen’s doctoral work comes in. His interest in synthesizing information effectively to improve bedside nursing has already garnered a considerable amount of attention. A critical care nurse and board-certified nursing informatics specialist, Linnen works in the Kaiser Permanente Division of Research, having received his master’s degree from the School’s Nursing and Health Systems Leadership program (a predecessor to the current MS-HAIL program) in 2013. At Kaiser, Linnen has led and collaborated on numerous clinical analytics projects, many at national scale, including electronic sepsis recognition, delirium risk prediction, safe alarm management, clinical event notifications and electronic dashboard design and development. He was honored with the “Emerging Leader” award from Alpha Eta, the Sigma Theta Tau International chapter at UCSF, in 2014, and was selected to be a 2016-18 Jonas Nurse Leader Scholar. In 2018, he received the Carol A. Lindeman Award for a New Researcher from the Western Institute of Nursing (WIN) for his study “Risk Adjustment for Hospital Characteristics Reduces Unexplained Hospital Variation in Pressure Injury Risk,” which appeared in Nursing Research in July/August 2018.

Specifically, Linnen aims to combine biostatistical and computational methods – including predictive analytics – with massive data sets to predict clinical outcomes for high-risk patients. His interest in the topic began in 2008, when he was working as a critical care nurse and Kaiser rolled out its EHR implementation.

Linnen, who had arrived in the United States from Germany in 2002, was excited because he had long felt that existing documentation on patients was inadequate. He quickly became his facility’s clinical expert in EHR use and appreciated the way the EHRs helped improve workflow and could ultimately streamline nursing work. But he was not blind to the many problems the first generation of EHRs was causing.

“I saw the limits, particularly around predictive algorithms aimed at providing clinical decision support,” he says. “If you look at the quality expectations of nurses – falls, pressure ulcers, catheter and central line infections, delirium, good family dynamics, preventing rebound – we have very crude tools to assess risk.” He says that the tools that do exist in EHRs are largely adapted from paper-based tools, which where designed for fast manual calculations and so are not particularly precise.

“The result is that a large number of patients in a critical care unit have a fall risk wristband,” says Linnen. “And as a standard of care, if everyone receives every intervention all the time, it’s very inefficient, so the notion of risk adjustment for nursing-related interventions really appealed to me.”

Eventually he found that while massive data sets have enabled better patient risk profiles, if you don’t adjust for hospital characteristics, the risk profiles remain inadequate predictors at the health system level. That is a problem for direct patient care and – in a system with many different hospitals – it can also cause quality improvement efforts to aim at the wrong target if the wrong hospitals and units are receiving poor performance scores.

Leverage the Latest Technology While Learning Lessons of the Past

Daniel Linnen Linnen believes that by deploying some of the latest advances in data science, such as artificial intelligence and machine learning, risk profiles can become even more precise and timely, as opposed to today, where retrospective operational reports offer information to change practice, but don’t necessarily prevent misperceptions in real time. Health systems could have dashboards that present a more timely and accurate picture of risk and patient safety concerns related to nursing care. That is one aim of the next stage of his work.

For nurses, however, Linnen is also cognizant of the danger of adding to alert fatigue. “So long-term, I could imagine having a platform, co-located with the EHR platform, with different risk parameters for different risk factors for nurse-sensitive concerns. Engines could communicate the risks clearly and efficiently to nurse-managers, who can then triage the information,” he says. “We wouldn’t create another monitor, but instead, we’d have something that puts the nurse back in charge.”

He cautions that before jumping aboard the next technology trend, he has become increasingly aware that two things must always come first. One is a well-curated database with clean data, good standards, clear definitions and shared language. The other is researchers who know how to ask the appropriate questions of the mushrooming data sets now available. Both elements can be lost in the rush to upgrade, a problem that has continually plagued health care technology implementations.

“The simplistic assumption that if we make more information available, we will improve care hasn’t turned out to be true,” says Wisner. “We have to synthesize the information in a relevant and timely way, so we can make better clinical decisions.”