As patient care organizations nationwide move forward to try to leverage data analytics tools in healthcare, the challenges of how to translate theoretical gains into actual gains in care delivery and outcomes improvement continue to pose significant obstacles. That was the consensus of a panel of healthcare leaders gathered to discuss the topic on Wednesday during the Health IT Summit in New York, sponsored by the Institute for Health Technology Transformation (iHT2—a sister organization to Healthcare Informatics, under the corporate umbrella of the Vendome Group LLC).
The panel discussion, entitled “Data Analytics for State-of-the-Art Care,” took place on Wednesday, Sep. 30, the second full day of the Health IT Summit, at Convene, a conference center in lower Manhattan. Shafiq Rab, M.D., vice president and CIO at Hackensack (N.J.) University Medical Center, served as moderator, and his fellow panelists were Al Villarin, M.D., CMIO and director, Division of Quality Analytics, at Staten Island University Hospital, Northshore-LIJ Healthsystem (and a practicing emergency physician); Jay Srini, chief strategist, SCS Ventures, and adjunct faculty assistant professor at the University of Pittsburgh; Bill Fox, vice president, Healthcare and Life Sciences, at MarkLogic (San Carlos, Calif.); and Joseph Hobbs, regional CIO, NetApp Healthcare (Sunnyvale, Calif.), and a former hospital CIO.
When Dr. Rab asked how the use of data analytics has evolved to the place where it’s at right now in healthcare, Srini said, “The first part of it is that initially, we needed to collect information. And when you look at data, we looked axes such as volume, variety, and velocity; but more important than all of those is the value of the data. And how do we analyze it? We’ve automated it now,” and that was a first, important phase. “So how do we aggregate it?” she asked. “And how do we provide it to caregivers and other end-users, and even patients? Liquidity is important.”
Hobbs noted that “We’ve spent a lot of time in the last few years normalizing data” in the healthcare industry. “Another big challenge we have is physician documentation in that some of the data we’re trying to leverage is not discrete data. And so we’re still working to pull out discrete data. So it’s taking all these things,” and making them analyzable, that is one of the biggest challenges right now.
“How do we drive analytics to the bedside?” Rab asked.
In turn, Villarin answered his question with a question. “How do we sell cars today? It’s the same question, but with a different response,” Villarin said. “I love a car that can give me great power, comfort, and usability.” Meanwhile, in the healthcare delivery context, he said, “You have to design a graphical user interface like anything else: cars, TVs, xBox, iPad. It has to be ergonomically modeled for end-users--not just physicians, but anybody. We have to build for this as a ubiquitous interaction, beginning from medical school. We have to have a unification of the data stream going to those clinical people. And that’s where the EMR comes in,” he said. “Right now, we have disparate data. How do we wrap analytics around NLP [natural language processing], voice recognition, ADT [admitting/discharge/transfer], telemetry systems? It’s all important.”
Villarin went on to say, “The hospital is a brain, like our brain, but it needs stimulus. You have to make things automatic. If you have to think about accessing data from systems, you’re lost. So it has to be automatic and sent to end users. So GUI, automation of data, and embedding the intelligence for the clinician, to know what’s going on with the environment of data around them.”
Very importantly, Fox said, “We need to present physicians and nurses with the same experience they’re getting everywhere else. There was this old saying attributed to Microsoft, nobody’s going to want a computer in their house. Nobody wanted “that” computer in their house, it had to be changed. The same is with analytics. If we’re looking within healthcare, that’s not where we’re going to find the solution to how these things will be adopted. Banks are getting people to go online all the time, because everybody’s worried about their money,” he noted. “How can you make that portal experience for the consumer, so they actually go to your website, rather than Google, for health information? And how do you get that doctor to go to your portal and get the same kind of information each time?”
With regard to that, Hobbs said, “I was talking to a CIO yesterday, and he was saying, we’ve got 7 million reports, but people in his hospital are saying that they have to go look for it. It’s about putting into the clinical workflow.
Srini pointed out that “The human brain can look at about eight different concepts at any particular time. And now, with the explosion of new information, and the number of drugs being released every day, it’s impossible to keep up.” There is so much data that could be helpful to clinical care, she noted, for example, significant amounts of genomic data that oncologists would like to have available to them for managing their care delivery. “There are many options, but a lot of the information isn’t in their workflow. So until we can connect these various pieces into their workflow,” she said, it will remain difficult to optimize the use of analytical information, on the part of physicians and other clinicians.
With regard to his organization’s success so far, Villarin noted that “We went with a vendor that was able to aggregate the evidence-based medicine, collect millions of articles, and place the evidence into our engine, that we use for analytics. Then they built a front end that would allow itself to be queried.” Northshore-LIJ has already had some preliminary success with regard to their ability to get their information system to artificially create information to send to physicians, “in the background,” in order to alert them to impending signs of pneumonia and severe myocardial infarction (heart attacks). Indeed, he noted, “We’re predicting a 50-percent decrease in severe MI, when a patient is admitted with symptoms. We’ll have the evidence six hours in advance that the patient may have a severe cardiac event. And it’s potentially available in pharmacy, ICU, discharge, etc.”
Villarin went on to say, “You have the evidence, you have the data, you put the two together in an engine, and give the physicians the ability to query the system. That’s where we’re hoping to have success.
Hobbs noted the challenge of data normalization, saying, “We had some analytics, but around quality control and infection control, etc. But the last year, we did nothing but focus on normalization, among three EMRs. We spent a entire year just normalizing data, to make sure the data sets were usable and mappable. And you have to do that, because the moment someone starts not trusting the data, the time and effort are wasted.”
Srini pointed out that “The Triple Aim is really about experience, cost, and value. And you can really triangulate it to provide the best value. And rather than letting the perfect be the enemy of the good,” she said, “if you can get things to be 80-percent accurate, you can get amazing results.” For example, she noted, at one patient care organization, “Using data we had available to us, including claims data, lab data, and so on, we were able to predict who our pre-diabetics were, at a rate of 98-percent accuracy.” So, she said, “While it’s a large and complex problem” to robustly apply analytics to patient care process improvement, “there are quick wins.”
A fellow presenter at the Health IT Summit, Luis Taveras, senior vice president and CIO at Barnabas Health in New Jersey, noted the success that he and his colleagues have had with alerts around emerging sepsis for inpatients. “We’ve implemented the sepsis algorithm and sepsis alert in the last six months,” he said, “and we’ve had many cases where the team was not aware that a patient was becoming septic. But the vitals and all the lab results are a part of the data being sent to the team. And it’s not just the predictive but the prescriptive, because the system will tell the care team what to do.” As a result of that analytics-facilitated clinical improvement work, he noted, “We’ve had a 30-percent mortality rate from sepsis in all our facilities, and now we’re saving a lot of lives. And now the question is, what do we do next?”
The key, really, said Villarin, referring to the potential capabilities of the “data lake” and “data cloud” concepts, “I don’t care about 99 percent of the data, the proposition that doesn’t touch me or my patient, speaking as a clinician. What’s important” to clinicians, he said, is the return on investment that an organization can get from inputting certain types of data. “If you can provide me with hotspots—it’s the ability of the analytics to give information to different end-users in a way they care about—so the pharmacist, the epidemiologist, the ED doctor, the nutritionist, can all use it. You have to be flexible about how data is provided and used. That’s smart data. Big data is bad data unless it can be used intelligently and provided in ways that people can really use it effectively.”
With so many different types of data being collected from so many different sources, Rab asked his fellow panelists, “How do we tie all this data together? How do I deal with myself in this new world of data analytics and state of the art care, where everybody knows about me?”
“The new millennial generation, and I can speak for my kids—they put what they eat, what they wear, where they are, on social media,” Srini pondered. “And I think that whole fear factor has evolved,
and become less of an issue, at least for younger generations in the U.S. And she noted the flip side of the ubiquity of data, in terms of how care management programs might someday be able to help patients, when she said that, “For example, a provider organization might be able to send a patient who is asthmatic a reminder, during a severe smog alert, to get their inhaler now, at the pharmacy.”
Fox said, “I’m a fairly serious athlete, and there’s a fairly big push among people to use devices, including wearables,” to help optimize sports performance, “and I don’t buy into that. So some of this will involve personal decisions. I think Jay’s right,” he added “There’s a lot of data that will simply be out there and be known about us as individuals. And so we’ll have to make personal decisions” about how to participate as healthcare consumers, in the ubiquitous-data phenomenon.
“Where we’re going with the data,” Villarin said, “is the need to build up the ability to work with data ever faster. Aggregating data will help us utilize modern technology and algorithms to help a hundred times more people than they can today, and that includes healthy people. Eventually,” he predicted, “we will have the applications of patients on the outside, putting information into a database, and the database analyzing that data from watches and running shoes and telemetry,” and other sources, to support enhanced health. “It’s the Internet of everything—everything’s going to be available.”