At LI hospitals, the artificial intelligence revolution has already begun

The words “artificial intelligence” evoke a futuristic world, but at certain Long Island hospitals, the future is here and now.

At some hospitals, nurses track the severity of patients’ symptoms with help from artificial intelligence, a broad term that encompasses computer programs that can be fed huge volumes of data and trained to analyze new data.

Others use A.I. to predict which patients are at risk of becoming ill again because they don’t follow instructions after they’re discharged, or those who are healthy enough to be allowed to sleep through the night instead of being awakened to have their vital signs checked. Still others use the technology to speed the analysis of sleep studies that help diagnose conditions such as sleep apnea.

The ventures vary widely in their origins, scope and funding. One is a new company called Truveta, formed in an unusual alliance between New Hyde Park-based Northwell Health and 19 other health systems across the country. The company, which recently announced $200 million in new private funding, pulls information from millions of the networks’ patient records — anonymized to protect confidentiality — and provides real-time analysis to health care providers.

A.I. projects in use/under development at LI health care institutions include:

Northwell Health has joined forces with 19 other health systems to start a company called Truveta, which recently announced $200 million in new private funding from its member networks and its CEO, Terry Myerson. Using information from millions of the networks’ anonymized patient records, the company provides real-time analysis to health-care providers.

NYU Langone Hospital-Long Island in Mineola has launched an A.I.-powered program that tracks COVID-19 patients’ vital signs, lab results and other information, recording 17 data points every 30 minutes to detect signs of potential deterioration.

Mount Sinai South Nassau in Oceanside uses A.I. to detect patients’ risks of falling or becoming severely ill, and to predict how much nursing care they will need. 

Stony Brook University’s Department of Biomedical Informatics has received more than $5 million in federal grants to research the potential use of A.I. in diagnosing and treating cancer.

Catholic Health uses A.I. to analyze patients’ brain waves, breath patterns, cardiac signals, leg movements and other data points recorded during sleep studies, speeding up the completion of reports that are reviewed by board-certified physicians. 

Sources: Northwell Health, NYU Langone Health, Mount Sinai South Nassau, Stony Brook University, Catholic Health

Northwell sees “revolutionary potential” in A.I., Dr. Martin Doerfler, Northwell’s senior vice president of clinical strategy and development, said in an interview, “and we wanted to be part of it.”

On a different scale, another new program got its start on a local nurse’s laptop during the coronavirus surge last year. After months of research and development, it evolved into an A.I. tool that flags COVID-19 patients at NYU Langone Hospital-Long Island who are at high risk of becoming severely ill in the next 12 hours.

Decision-making stays with humans

The A.I. program “doesn’t take over your decision-making and it never should,” said Jeanmarie Moorehead, senior director of operations at the Mineola hospital. “But it is definitely value-added, tremendous value-added to the clinician.”

What the A.I. efforts have in common is an ambitious effort to use specialized computer programs to comb through columns of data too vast to be understood by a human being, detect patterns and use that information to guide health care providers in diagnosing and treating patients.

The use of A.I. in health care is on the rise, with global funding in the sector reaching $8.5 billion from January through September — nearly double the amount in all of 2019, according to CB Insights, a company that tracks A.I. investments. The United States was the biggest spender, with investments in A.I. in health care totaling $5.45 billion from January through September, the company reported.

Dr. Martin Doerfler of Northwell Health.

‘We need to know the answers that are hidden inside the fragmented data …’

– Dr. Martin Doerfler, Northwell’s senior vice president of clinical strategy and development

Health care technology, including A.I., “is clearly seeing an increased level of investment,” especially over the last year and a half, said Peter Micca, a partner and national health tech leader with Deloitte & Touche LLP in Manhattan. “COVID has only accelerated the awareness around the importance of technology in health care.”

One hurdle is that, in contrast with industries such as finance and social media, health care data “is completely fragmented,” Doerfler said. “We need to know the answers that are hidden inside the fragmented data, and you don’t get the answers until you get the data sets large enough that you can find the answers quickly.”

More diversity in data

Incomplete data sets often lack diversity of race, gender, socioeconomic status and other characteristics, and overrepresent middle-aged white men with health insurance, Doerfler said. By contrast, said Terry Myerson, Truveta’s CEO, the data set drawn from its 20 networks represents 16% of all clinical care provided in the United States and reflects “the diversity of our country.”

The goal of Truveta, Myerson said, is to “empower our clinicians to be experts” and “help families make the most informed decisions about their care.”

Some industry analysts warn of potential pitfalls in the adoption of A.I. At the annual conference of Stony Brook University’s Center of Excellence in Wireless and Information Technology this month, Daniel Holewienko, executive director, big data and business intelligence at Henry Schein in Melville, said failing to embrace A.I. would put health care companies “at a competitive disadvantage.”

Still, he said, those adopting the new technology can face high costs and difficulties integrating it into their current systems, among other challenges. Protecting privacy, preventing bias and making sure clinicians do not place excessive faith in the machines are among the other concerns, health care providers say.

Dr. Joel Saltz, founding chair of the Department of Biomedical Informatics at Stony Brook University, said the industry has proceeded cautiously in adopting A.I. The advanced technology has become more widely used in the last five years or so, he said.

“These things are incremental, especially in health care, because you’ve got to make sure they’re safe and effective,” said Saltz, who is working with colleagues on a project led by the federal Food and Drug Administration, focusing on the use of A.I. in digital pathology. Such tools, he said, are used for “decision support,” to aid doctors and nurses rather than replace their work.

Tool for treating cancer

Stony Brook’s biomedical informatics department is working on three projects funded by more than $5 million in federal grants to research the potential use of A.I. in diagnosing and treating cancer. An A.I. program can examine hundreds of slides and analyze millions of cells, complementing doctors’ ability to visually classify tumors, Saltz said. “Think about the difference between a paper map and Google Earth,” Saltz said. “It really opens up a whole new way of doing things.”

It’s possible that some of the research could be put into clinical practice within 10 years, he said.

In some cases, the COVID-19 crisis has sparked innovation by doctors, researchers and nurses as they raced to understand the new virus and find ways to save patients’ lives. Nurses have been key players in using and, in at least one case, helping to develop the new technology.

At NYU Langone Hospital-Long Island in Mineola, for instance, computers are running a new A.I.-powered program that keeps an eye on COVID-19 patients’ vital signs, lab results and other information, using patients’ electronic medical records to monitor 17 data points every 30 minutes and detect signs of impending danger.

A paper version of the program was born of necessity during the first COVID surge in early 2020. At the time, nurse clinician Cathrine Abbate was seeking a rapid, consistent way to communicate with her fellow nurses and doctors about the severely ill patients suffering from a new and brutal virus.

Saving COVID patients

On video conference calls before and after their shifts, Abbate and other nurses brainstormed about the warning signs that tended to precede a rapid decline in patients’ condition, such as needing large amounts of oxygen or not being able to eat or move. With that information, she used Microsoft Word to create a blank grid that she printed out at her home in Huntington Station. The grid included seven columns, tracking information about the patients’ condition. In the hospital, using copies of the grid made it easier for nurses to quickly rank the severity of each symptom and give an overall rating from 1 to 10, with 10 being the worst, she said.

“We needed to be able to fluidly communicate with each other about how the patients were doing,” Abbate recalled. “It was just a way to create a language for ourselves.”

Nurse manager Sarojini Seemungal helped implement the new system on the 30-bed unit, and alerted her own managers. Moorehead brought it to the attention of researchers at NYU Langone in Manhattan who specialize in analyzing data.

The researchers spent months meeting weekly with nurses and developing an A.I. program that provides information to a rapid response team of critical care nurses at the Long Island hospital who give special attention to the highest-risk patients, said Dr. Yindalon Aphinyanaphongs, director of operational data science and machine learning at NYU Langone Health.

The program acts as a “tireless monitor,” taking information about thousands of previous patients — including many whose conditions deteriorated — and using it to predict whether current patients are likely to decline, he said.

Hype vs. reality

There’s a lot of “hype” about A.I. and its subset machine learning, a term that refers to computers learning from examples, Aphinyanaphongs said.

“A lot of times when people think of artificial intelligence, they think of, you know, WALL-E,” he said, in a reference to the 2008 animated movie about a lonely robot. But in fact, “the value in some of these models has to do with, not doing something better than humans, but doing things faster than humans can do,” and more consistently, he said.

A tool like the one developed by the nurses and researchers, he said, can take a health care provider who has little experience with COVID, and it “can help elevate their experience and their expertise to the point where they’re functioning at the same sort of assessment level as someone who has seen a lot of COVID patients.”

The program can be downloaded for free by other hospitals that use the Epic electronic medical records system, Aphinyanaphongs said.

Stacey A. Conklin of Long Beach has been

‘[A.I.] takes a lot of the subjectivity away from staffing, and allows us to really put the resources where they’re needed most.’

-Stacey Conklin, Mount Sinai South Nassau’s chief nursing officer and senior vice president of patient care services.

At Mount Sinai South Nassau in Oceanside, computers use A.I. to make sure patients receive precise, personalized care, taking into account the severity of their illnesses and other factors, said Stacey Conklin, chief nursing officer and senior vice president of patient care services. Those at higher risk of falls, for example, get extra help moving around if needed, she said.

A.I. “takes a lot of the subjectivity away from staffing, and allows us to really put the resources where they’re needed most,” Conklin said. “If I as a manager am trying to figure out where to put all of my resources, it’s very helpful for me to be able to look broadly across the unit and see what’s going on with all the patients so that I can ensure that the patients are getting the best care.”

Filling need for speed

At the Catholic Health system’s six sleep labs, A.I. is used to analyze the sleep studies of patients who spend the night hooked up to machines that record brain waves, breath patterns, cardiac signals, leg movements and other data points to diagnose conditions such as sleep apnea, said Brendan Duffy, director of sleep services at the network.

The data can fill hundreds of pages, and analyzing the information is “a very time-consuming, very meticulous” process that used to take one to two hours for each report, Duffy said.

Once the health system started using the A.I. program about three months ago, he said, that time was reduced to about 20 minutes, he said.

The new system means the sleep labs can get patients on the calendar for follow-up appointments more quickly, so patients spend less time driving while drowsy or suffering compromised immune systems due to sleep deprivation, he said.

But despite their remarkable efficiency, he said, the computers can’t have the last word.

A board-certified physician reviews the sleep reports “each and every time, and that’s nonnegotiable,” he said.

At Northwell’s Feinstein Institutes for Medical Research in Manhasset, researchers used A.I. to analyze 24 million patient vital sign measurements. The results helped them predict which patients were low-risk enough to sleep through the night with a nurse looking in on them periodically, instead of being awakened to have their vitals checked, according to an article published last year in the journal Nature Partner Journals Digital Medicine.

The health system also is using A.I. to identify certain high-risk patients, said Dr. Jamie Hirsch, director of Northwell’s data science program.

Getting personal

In presentations about A.I., Hirsch tells his fellow physicians the technology can help identify people such as a fictional patient he has dubbed “Ethel,” a sprightly 87-year-old grandmother who is “fiercely independent,” but who feels overwhelmed in the hospital, lives alone and might need more assistance than she realizes.

In a busy hospital filled with hundreds of patients, a patient like Ethel might not get the hand-holding she needs, he said.

But when an A.I. program is trained to flag patients who are older, live alone and are coping with a bewildering array of medications and discharge instructions, he said, “now you have a patient experience specialist that’s going to come in and say, ‘How are you? Let’s sit down, let’s talk, you know, how can we make your experience better …. How do we get you home, so you can continue living that independent life that you so value?”

He said, “It allows us to focus our energies in the right way, to the right person, at the right time.”

Related Posts