Max Topaz Uses Data Science to Help Nurses Care for Patients

As a boy growing up in Siberia, Max Topaz watched as his 84-year-old grandmother was turned away from a medical clinic. She had a broken hip, but the doctors deemed her too old to treat. A few months later, without medical care, she died. It was a cruel irony as she tended to dying soldiers as a nurse during World War II. Yet, when it came time for her to receive life-saving medical care, she was denied.

“I decided right then I had to do something to improve health care systems, especially for those who most need it,” Topaz says.

Now a professor of nursing at Columbia University and an affiliated member of the Data Science Institute, he is a pioneer in using data science to help nurses streamline their jobs and better care for millions of patients.

The four million nurses in the U.S. routinely take detailed notes on their patients, which results in a massive goldmine of medical records—data that’s often overlooked. Topaz uses machine learning, natural language processing, and text mining to wade through nursing data and electronic health resources. From that, he creates algorithms and AI tools that help nurses make informed clinical decisions.

“As a researcher and a teacher, I’m guided by the Data Science Institute motto to use ‘data for good,’” Topaz says. “Data science has revolutionized many professions, but I use it to create tools that help nurses improve health care for patients.”

A Nimble Thinker

One tool Topaz created, Nimble Miner, is an open-source natural language processing software clinicians may use to mine millions of patient records, including the myriad clinical notes nurses take on patients. The software flags salient characteristics such as indications of drug and alcohol abuse, mentions of problems such as patients falling at home, and worsening symptoms. These characteristics are then used by nurses to identify patients who are at-risk and need timely interventions.

Topaz also uses millions of clinical notes to better understand which nursing interventions will help patients recover faster. For example, as a research associate with the Visiting Nurse Service of New York, the largest not-for-profit home care organization in the U.S., he uses nursing data to understand which treatments are best suited to treat chronic wounds or identify early symptoms of Alzheimer’s disease. A leader in using natural language processing for nursing, Topaz believes that by mining millions of nursing records “we are making critical nursing data useful, which in turn improves our ability to provide personalized care to every patient.”

Most recently, Topaz developed a patient-prioritization tool called PREVENT to help homecare nurses determine which newly discharged hospital patients need immediate care. Medicare requires that patients receive a home visit within 48 hours of their discharge. For high-risk patients, however, waiting 48 hours could lead to deteriorating conditions. As it is, visiting nurses have little way of knowing which discharged patients are at high risk and should be visited first. Electronic health records (EHRs) do not include high-risk data, so nurses are often left to their best judgment. And that leads to a problem.

Records show that 30 percent of newly discharged patients are readmitted to the hospital within a week, taking up limited space and driving up medical costs. About half of those readmitted cases, Topaz found, could have been prevented had they been seen by nurses right after they returned home. To mitigate the problem, he used data mining and regression modeling to calculate which patients should be prioritized for home visits.

Topaz used five data points on each patient—sociodemographics, medications, depression, learning ability, and living arrangements—to create an algorithm that extracts meaningful patterns from the nursing data. He tested PREVENT in a pilot study on patients seen by the Visiting Nurse Service and found that patients who were identified as high-risk and received a nursing visit a half-day sooner than those in a control group were much less likely to be readmitted to the hospital. The results were published in the journal Research in Nursing & Health.

Topaz is currently testing PREVENT in clinical trials at NewYork-Presbyterian and Allen Hospital. He is also studying EHRs at the two hospitals to see if they may be updated and automated to include data that would help visiting nurses make better decisions. He uses natural language processing to evaluate the EHRs as well as the voluminous notes that nurses take on patients.

“Researchers have used NLP and artificial intelligence to assist doctors, but I’m a nurse, so I use AI to help nurses,” says Topaz. “At four million strong, nurses are the largest segment of health care workers. Anything I can do to improve their work improves health care for all of us.”

From Russia to Israel to America

Soon after his grandmother died of complications from her hip injury, Topaz left Siberia for Israel, where he served as a medic in the army. Like his grandmother, he cared for soldiers in life and death situations. He also helped develop the Israeli army’s EHR system, which taught him the essential role that technology can play in improving health care.

After completing his military service, Topaz enrolled at the University of Haifa to study nursing informatics, a field that allowed him to combine his love of technology with his passion for helping people. He earned a bachelor’s and master’s degree there, completed a doctoral Fulbright Fellowship at the University of Pennsylvania, pursued a postdoctoral fellowship at Harvard Medical School and Brigham Women’s Hospital, and taught in the School of Nursing at the University of Haifa for two years. In 2018, he was appointed the Elizabeth Standish Gill Associate Professor of Nursing at Columbia University Irving Medical Center and became a Data Science Institute faculty affiliate.

Topaz says he was drawn to Columbia because of the nursing school’s international prominence in the field of data-driven nursing research; Columbia nursing professors Suzanne Bakken, Gregory Alexander, Kenrick Cato, and Sarah Collins Rosetti are all prominent in the field of nursing informatics. He is part of a team of Columbia professors recently awarded seed funds from the Data Science Institute to create an introduction to data science course for nursing students, which would include machine learning and other data science techniques to improve health care.

“I strongly believe that we are approaching a new brave world where data science can help us produce better health care for all patients,” he says. “And I’m excited to be part of it.”

— Robert Florida, Data Science Institute

About garen