Search Site

Masters of Information

These innovative nurse researchers are using data science and informatics to help patients - and healthcare systems - thrive.

Share

By Kenneth Miller

This article originally appeared in the Fall/Winter 2018 issue of Columbia Nursing Magazine.

 

Recent advances in computing power have enabled researchers in many fields to harness the potential of Big Data as never before. For nurse-scientists, that means a quantum leap in the ability to see how complex factors interact to shape patterns of health and disease in vast populations—and how everything from government policies to caregivers’ Twitter habits affect patients’ well-being. These six researchers are developing new digital tools to analyze those dynamics and to formulate solutions to some of healthcare’s knottiest problems.

 

The Trailblazer

 

In the 1980s, Suzanne Bakken, PhD, RN, helped pioneer the field of nursing informatics, and she’s still breaking new ground. Since arriving at Columbia Nursing in 2000, Bakken—the Alumni Professor of Nursing and professor of biomedical informatics—has built an interdisciplinary training program that’s considered among the finest of its kind. Director of the Precision in Symptom Self-Management (PriSSM) Center and co-chair of the Health Analytics Center of Columbia’s Data Science Institute, she has published more than 300 papers and served as principal investigator for $30 million in grants. She was recently named editor-in-chief of the Journal of the American Medical Informatics Association (JAMIA)—the first nurse to hold that title.

 

Besides bolstering the skills of nurse-scientists, Bakken’s central goal is to improve care for underserved populations—in part, by helping them care for themselves. The PriSSM Center, for example, aims to advance the science of symptom self-management for Latinos through research that takes genetic, behavioral, and environmental factors into account. And the people being studied get a say in how that research is carried out. “For years,” Bakken explained, “I’ve insisted that when we collect data from a community, we make sure to give it back in a way that’s understandable and meaningful and actionable.”

 

An example of this approach is an ongoing study aimed at developing a family health information-management system for caregivers of people with dementia. Team members, including assistant professor Adriana Arcia, PhD, RN, worked with community members to design the study and to develop information visualizations that would help caregivers understand their own health risks, including factors such as diet, exercise, and genetics. One aspect of the project, implemented by associate research scientist Sunmoo Yoon ’04 ’11, PhD, RN, involved mining Twitter to identify network structures among thousands of caregivers and to analyze the emotional tenor of the tweets themselves—all with the aim of creating an app- or Web-based intervention to improve social support.

 

“The beauty of data-science methods is that you have the opportunity to collect information from novel sources, some of them outside of the traditional healthcare system,” Bakken said. “But what’s equally crucial is presenting it in a manner that’s as useful as possible.”

 

The Visual Thinker

 

Jacqueline Merrill ’98 ’06, PhD, MPH, RN, professor of nursing in biomedical informatics, is a leading researcher on how data science can help improve the performance of health organizations. But her route to the field was wildly circuitous. After several years as a diploma-educated critical care nurse, she spent more than a decade as a clothing designer and photo stylist. Then, missing her old profession, Merrill took a job as a school nurse—and found a new mission. “We followed the public health model, which is about preventing illness and promoting health in a large population,” she recalled. “Working at New York City schools, covering 3,000 kids, I realized how much public health nurses, as well as other public health workers, needed support in managing information effectively.”

 

To deepen her own expertise, Merrill pursued a bachelor’s degree in nursing and a master’s in public health, followed by a PhD in public health informatics from Columbia Nursing in 2006. She specializes in data-science techniques such as network analysis and agent-based modeling, which target the complex interactions that can determine an organization’s effectiveness. Her work has had a range of real-world impacts, from helping health departments reduce the incidence of sexually transmitted illnesses to enhancing disaster-response preparedness. She’s currently co-principal investigator on an NIHfunded study analyzing years of Medicare and Medicaid data for over 50,000 patients with dementia who’ve been treated by the Visiting Nurse Service of New York (VNSNY). “My team is mapping a trajectory of care for these people,” she explained. “Where do they live? What services do they use? In what sequence? How do those factors affect outcomes?”

 

To make the results of such multilayered studies easier to grasp, Merrill is developing new ways of visualizing
data—and teaching the latest methods to graduate and postdoctoral students at Columbia Nursing. “This is where my background in design is critical,” she said. “Visualization is a way to capture many planes of information in one image. It’s an ideal tool for explaining complexity.”

 

The Safety Maven

 

After graduating from the University of Pennsylvania with a BS in nursing, Sarah Collins ’09, PhD, RN, worked as an ICU nurse at several major metropolitan hospitals. “As I moved from one to the other, I observed how strongly different information systems impacted clinical workloads and the ability to deliver care,” she recalled. “It was really quite striking.” That insight spurred Collins to enter the informatics program at Columbia Nursing, where she earned her doctorate in 2009. And it helped draw her back to the school this past February, when she became an assistant professor of biomedical informatics and nursing.

 

Collins spent the previous seven years as an instructor at Harvard Medical School and senior clinical and nurse informatician at Brigham and Women’s Hospital in Boston, where she focused on leveraging informatics to enhance patient safety. Her most ambitious project, still ongoing, was a multi-site study called Communicating Narrative Concerns Entered by RNs (CONCERN), funded by the National Institute of Nursing Research (NINR), which aimed to
mine electronic health-record data—particularly nursing documentation—to prevent patient mortality in real time. As her co-principal investigator, she tapped former Columbia Nursing classmate Kenrick Cato ’08 ’14, PhD, RN, who has continued in that role.

 

CONCERN grew out of Collins’s observation that nurses often expressed misgivings about a patient’s status, based on subtle signs of deterioration, long before worrisome signals appeared on vitalsigns monitors. Her goal was to identify such early-warning signs in EHRs (electronic health records), and use them to create what she calls predictive healthcare process models of clinical concern (HPM-CCs). In the first phase of the study, published in 2013, she and her team used advanced statistical modeling techniques to analyze EHRs for 15,000 acute-care patients and 145 cardiacarrest patients over a 15-month period. They found that records for patients who died had significantly more optional comments from nurses, as well as more frequent vital-signs checks, over the previous 48 hours—the first time such a correlation had been shown.

 

The next phase of the study is to determine the precise threshold at which intervention is warranted, using natural-language processing and machine-learning techniques to generate HPM-CCs; after that, the team will develop an app to alert clinicians, and test it in the field. But Collins’s research has already confirmed her initial hypothesis. “The evidence is pretty clear,” she said. “If a nurse is worried about a patient, that’s a reason to act—and without delay.”

 

The All-Arounder

 

The computer-besotted son of an emergency room nurse, Kenrick Cato ’08 ’14, PhD, RN, always thought he would go into either healthcare or IT—but after graduating from Swarthmore, he joined the National Guard instead, and wound up serving as an infantry captain in Iraq. When his tour was over, a friend suggested that a career in nursing informatics might be the perfect way to unite his disparate interests. “She was right,” said Cato, who earned his doctorate from Columbia Nursing in 2014, and is now an assistant professor at the school.

 

Cato’s research reflects his wide-ranging intellect and his combat-honed capacity for multitasking. Besides
co-leading the CONCERN study, he’s responsible for a broad variety of investigations. For example, Cato and his colleagues are developing an algorithm that can identify transgender patients from their electronic medical records—a first step toward learning more about the health needs of this poorly understood population. His team is also mining
EHRs over a five-year period to trace patterns of treatment for thousands of people identified as suicidal after admission to emergency departments. “One of the most crucial decisions for ED clinicians is whether and when to refer suicidal patients to higher levels of care,” Cato explained. “Yet there are currently no gold-standard guidelines for how to manage suicide risk in the United States, and traditional statistical approaches may not be well suited to
analyze the large numbers of variables that contribute to optimal treatment of these patients.” Eventually, he hopes to identify which variables predict key outcomes (such as patients’ discharge disposition and overall length of stay), and to use that data to develop decision-making tools.

 

To Cato, the diversity of these studies illustrates the vibrancy of his emerging field. “Nursing informatics is about much more than improving nursing practice,” he observed. “As this discipline grows, it’s becoming a force multiplier for finding better solutions throughout the healthcare system.”

 

The Language Whiz

 

As a boy in post-Soviet Russia, Max Topaz, PhD, RN, saw his 83-year-old grandmother turned away from a clinic after she broke her hip, because the doctors deemed her too ancient to be worth treating. A few months later, she died of complications from her injury. “That’s when I first decided I wanted to make healthcare systems better,” Topaz recalled. After immigrating to Israel, where he served as an army medic, the teenage tech geek was thrilled to help implement the military’s first electronic health-records system. Moving on to the University of Haifa, he decided to focus on nursing informatics. And by the time he earned his PhD as a Fulbright Fellow at the University of Pennsylvania, in 2014, he was recognized as one of the most promising young researchers in the field.

 

This fall, after completing postdoctoral studies at Harvard Medical School and teaching at the University of Haifa, Topaz was appointed an associate professor at Columbia Nursing. Now 38, and the author of more than 50 papers in peer-reviewed journals, he’s a pioneer in natural-language processing in nursing, which uses machine learning to analyze large volumes of clinical texts or speech. Over the past few years, he has created algorithms that can highlight references to wound care or comorbid conditions in nursing notes, and help clinicians distinguish between urgent and less-crucial drug-allergy alerts. Working with the Visiting Nurse Service of New York (VNSNY), he’s currently developing software that can help home-care nurses decide how to prioritize patients upon admission based on analysis of risk factors, including socio-demographics, comorbid conditions, medications, depression, learning ability, and living arrangements.

 

Besides advancing nursing research, Topaz (who consults for tech startups as well as major healthcare organizations) is eager to help students develop informatics tools of their own. In Haifa, he taught a course on innovation and entrepreneurship, in which students collectively created a mobile app for patient triage in emergency departments—and he plans to offer something similar at Columbia Nursing. “My basic goal,” said Topaz, “is to help move this technology revolution forward.”

 

The Symptom Sleuth

 

Theresa Koleck, PhD, RN, got into nursing for two reasons: “I’ve always liked science—biology, anatomy, physiology, genetics. And I like interacting with people.” But as an undergraduate in the University of Pittsburgh’s nursing program, she realized that the science part excited her most. While working on a breast cancer treatment-related symptom study as part of a research mentorship program, she wondered why some breast cancer patients developed cognitive impairments before treatment while others didn’t. Could genetic differences play a role? Studying a small number of patients for her PhD project at Pittsburgh, Koleck found evidence to support her hypothesis.

 

That only added to her curiosity, however. Cancer symptoms often came in clusters, she knew—fatigue, pain, depression, and sleep disturbances, for example, or nausea and vomiting—and their severity varies from patient to patient. Other chronic diseases had symptom clusters, too, some of which occurred across multiple diagnoses. Could genetic or other factors help explain these groupings, and point toward ways of preventing or treating them? Koleck’s
hunger for answers led her to a postgraduate fellowship in informatics at Columbia Nursing. After completing the program in May, she became an associate research scientist at the school.

 

Koleck is leading an NIH-funded study that will use natural-language processing to mine years’ worth of electronic health records for thousands of patients with cancer, type 2 diabetes, chronic obstructive pulmonary disease, and heart failure. Her team is currently developing ways to identify symptom clusters from these records; next, they’ll explore relationships between the clusters and clinical information, such as patient demographics or laboratory test
results, for patterns suggesting biological, behavioral, or environmental risk factors. “Using EHRs allows us to analyze
huge amounts of data without having to recruit huge cohorts of patients,” Koleck explained. “These new methods enable us to do massive longitudinal studies that would otherwise be impossible.”

 

Her research would also be impossible, she adds, without the mentoring she’s gotten at Columbia Nursing. “Dr. Bakken is so incredibly generous, supportive, and protective of my time,” Koleck said. “She’s given me the skills and confidence I need to succeed.”