Data analytics driving medical breakthroughs

Using big data to save lives

A hospital is usually a pretty busy place, but the neonatal intensive care unit at Toronto's Hospital for Sick Children has been buzzing with even more activity than is customary. Thanks to a new technology partnership, the hospital is working to use analytics to predict more accurately than ever before which premature babies are at most risk for disease and infection.

The hospital is in a study to monitor temperature, heart rate, blood saturation and blood pressure levels on preemies, collecting streaming data from electronic devices that monitor the premature babies.

Sick Kids, as the hospital is known, is in good company. Healthcare providers -- from insurance firms to hospitals to service suppliers -- are lining up to adopt advanced technologies to help them take better care of their patients, in many cases becoming more proactive and more personalized than ever before, with the hopes of saving money, too.

Susan Feldman, vice president at research firm IDC, says technology is enabling a process "that will change our lives. One of the things going on in healthcare is understanding we need more of our decisions based on evidence so we can more appropriately process and identify information, and bring it to decision makers in an actionable way."

Other industries to follow

Healthcare is just one of many verticals that are using or will make use of tools that can rapidly analyze information. Emergency preparedness, terrorism detection and fraud detection are all likely to follow quickly, says IDC vice president Susan Feldman.

What these fields have in common, Feldman says, is that they must deal with a lot of incoming data from multiple sources in multiple formats. Feldman believes the market will be divided into platforms -- like Watson -- that provide a set of highly integrated tools; applications built on these platforms will be specific to a particular problem.

"In the future, we will see ecosystems of applications emerging around a relatively small number of platforms," Feldman says. In this scenario, smaller vendors will build specialized applications, adding in the right information sources, and designing the workflow and tools so that users can complete their tasks more efficiently and productively.

Some of the upcoming vendors in this space that Feldman is watching include:

  • Autonomy, a provider of "meaning-based technology," to make sense of unstructured data to draw business value. Feldman says it does probabilistic search and inferencing, as well as content management and is moving into the e-discovery area.
  • Vertica, an analytics platform provider bought by HP last year, is another. Its features include real-time loading and querying, analytics, data compression and columnar storing and execution capabilities.
  • Oracle, which offers "less in the way of linguistic capabilities" has strong search functionality and content management, Feldman says. Oracle recently bought Endeca, a platform provider to store, manage, search and analyze both structured and unstructured information.
  • EMC, which sports an assemblage of technologies -- including, of course, storage.

Feldman says other vendors that have pieces of the technology to pull together unstructured information into an "answer machine" include Microsoft Research, as well as SAP and SAS.

- Esther Shein

It's not just healthcare that is on the verge of benefiting from predictive analytics. Other industries that struggle with vast amounts of ongoing data where this type of technology makes sense include finance, emergency response, entertainment and the legal field, to name a few. (See sidebar.)

Society is "on the cusp of being able to do more than ever before, and support decision-making in ways that have not been possible before," says Dr. Harlan Krumholz, professor of medicine at Yale University and a physician who is also involved in big-data research projects with large hospital consortia.

That said, however, specific results will be up to watchful physicians to ensure, he warns. "Those claims about being able to improve medicine ... have to be looked at carefully. The question is: What do people do with the information" from these systems?

"If wrong decisions are made based on [incorrect] assumptions about information, they won't obviously provide any value and could [instead] be harmful," Krumholz says.

Fewer sick babies

Back in Toronto, the hospital is processing its data in real time using IBM's InfoSphere Streams , software that can correlate and analyze thousands of real-time data sources. The University of Ontario Institute of Technology (UOIT) is using the software to collect streaming data from electronic devices that monitor the premature babies.

The technology is giving UOIT the ability to make sense of the data and analyze it in ways that include, they hope, discovering the onset of sepsis and various other conditions before these problems occur, says Dr. Carolyn McGregor, the Canada Research Chair in Health Informatics at UOIT.

This test has been running in parallel with current clinical practice so doctors and scientists can compare the two approaches, which they are currently in the process of doing. One day's worth of data is copied and sent back to UOIT for the offline analytics component.

The platform, known as Artemis, or "data baby," has been input with a set of clinical rules that serve as a layer of analytics to help it make predictions for the first time, says McGregor, who is also a professor and associate dean at UOIT. Today, medical devices at the bedside give broad information, she explains. Devices provide readings at a very high frequency, but "a human has to be able to analyze" the results, which are "constantly changing," McGregor says.

Final results have not yet been released -- they're expected sometime in late April for peer review and should be public by year-end. But initial results have proven Artemis's "robustness as an approach," McGregor says. The study, of over 400 patients in three sites, has collected "the equivalent of two decades of patient years" worth of data, she explains.

While medical personnel have some traditional indicators for the onset of infection -- such as body temperature -- Artemis will provide "a much richer environment," McGregor says, to analyze a range of different signals for a variety of various conditions that babies can develop.

Toronto's Hospital for Sick Children is testing a data analytics system to more accurately predict which premature babies, like this one, are at highest risk for infection. Photo credit: Carlos Garcia Rawlins / Reuters

"This system was designed to predict the onset of sepsis 24 hours before it became clinically apparent," says IBM 's Chief Medical Scientist Marty Kohn. "It is using structured data to look for patterns that allowed [the hospital] to predict the onset of serious disease based on clinical observation. In a case like this, if you can intervene an hour earlier you can often improve outcomes dramatically."

"This has the potential to reduce a baby's average length of stay in the hospital, as well as to save lives," McGregor says.

UOIT and hospital staffers also plan to use the data to do more clinical research "to look beyond what people already know from watching a heart rate for example; what else we can find from looking at physiology," McGregor explains. By running new algorithms, she says Artemis would be able to tell clinical staff with high probability that a baby's behavior change might be correlated to infection.

Dealing with data deluge

The ability to make sense of unstructured information and make predictions about diseases that could develop, thereby enabling physicians to make better decisions about treatment, is the goal of a set of tools from IBM that include its highly publicized Watson technology . Voluminous amounts of information come at doctors all the time, especially with thousands of new articles being published monthly in medical journals. Add to that lab results, health care systems and social networking sites for professionals, and it can become overwhelming pretty quickly.

"As you gain more and more information in the world, it becomes impossible for a person to manage a physical library or have an electronic collection unless you can create very intelligent access to it,'' explains IDC's Feldman. "Language is so rich and so variable that there are too many ways of saying the same idea, so you really need technologies that can understand all the various ways we can understand something."

Tools that structure various sources of information by applying analytics are helping the health care industry -- among others -- gain a whole new level of intelligence, observers say. IDC is projecting the overall big data information technology and services market to grow from $3.2 billion in 2010 to $16.9 billion in 2015.

"Why Watson is such a breakthrough is it understands questions and tries to interpret them," while making hypotheses through its text mining and analytics capabilities, says IDC's Feldman. IBM built Watson to do what it calls "Deep Q/A," she says, with QA standing for question and answering.

"It's a whole new approach or technology that tries to understand what people are looking for," she says. The technology understands how one entity affects another; in health care, for example, the computer can analyze how a drug affects a particular type of person who has a certain type of disease.

This approach is much different from how a search engine works. A search engine attempts to answer a question that has been input using a few keywords, says Kohn. "Then you get sites or pages to look at to see if any are actually relevant to the question you have in mind. That takes a lot of time ... it's an ineffective process at best."

In contrast, Watson understands a question that's been posed in natural language, reads literature and, using a series of massively parallel "probabilistic algorithms" to analyze the information it is given, goes out and brings back prioritized suggestions on whatever the topic, Kohn says.

Answering complex questions

WellPoint, which bills itself as the country's largest health benefits company, recognized that the Watson Deep Q/A initiative could be used to automate utilization management. That's the manual, often complicated and time-consuming insurance-approval process physicians are required to go through before treating patients, says Ashok Chennuru, director of technology.

"We know according to guidelines and policies we have for cancer treatment options that [patients] need certain drugs and procedures" that are given or performed in concert for maximum benefit. However, right now the company doesn't have a streamlined approach for doctors' offices to submit multiple pre-op authorization requests together, says Chennuru. "Sometimes not enough documentation is submitted and we have to go back and forth and it can be frustrating."

Those claims about being able to improve medicine . . . have to be looked at carefully. The question is: What do people do with the information from these systems? Dr. Harlan Krumholz, Professor of Medicine, Yale University

The company started using IBM's Content and Predictive Analytics (ICPA) decision support engine last fall to automate the submissions process between the physician and payer, he says. If everything is clearly documented from the doctor's office, it is sent to WellPoint's version of Watson, which determines if the submission meets all criteria and quickly gives a yes or no answer or requests more information, he says.

WellPoint is conducting a pilot in one region on a limited basis to ensure that the tool is working well, Chennuru says. By the end of the year, the company is hoping to deploy it in all 14 states where WellPoint has facilities.

Eventually, once oncology is covered , WellPoint foresees being able to use the tool for other medical issues such as respiratory distress, diabetes and cardiac and kidney diseases. Using Watson, WellPoint says it envisions being able to look at massive amounts of medical literature, population health data and a patient's health record to answer "profoundly complex questions."

The company believes it will eventually be able to develop new applications to allow physicians to load patient medical histories, recent test results, recommended treatment protocols and the latest research findings into Watson. The goal is to be able to discuss the most effective courses of treatment with their patients.

"It's about leveraging the power of the computer rather than oncologists having to keep up with this [new information] they receive on a daily basis,'' says Chennuru.

Preventing hospital readmissions

Seton Healthcare Family, a health care system serving about 1.9 million people in central Texas, was looking for a way to reduce readmissions for patients with congestive heart failure (CHF), which the company says is costly -- to the tune of an estimated $35 billion in the United States. Seton Healthcare says 500,000 new cases are diagnosed every year and more than half of CHF patients need to be readmitted within six months after treatment.

"Hospitals are penalized [by Medicare] for having too many patients who need to be readmitted," says chief strategy officer Travis Froehlich. To predict which patients are most at risk for readmission, Seton Healthcare uses ICPA to identify known risk factors like smoking, so health professionals can focus their efforts on keeping patients at home more efficiently.

To accomplish this, Seton is using the ICPA for Healthcare tool, which lets officials mine unstructured data using natural language processing and search technologies. The company says more than 80% of healthcare data is unstructured and consists of physician notes, registration forms, discharge summaries, echocardiograms and other medical documents.

When combined with structured data, all this can paint a more accurate picture of trends, patterns and deviations, allowing clinicians to make better treatment decisions. She says this data was found in the history and physical part of the medical record in a narrative section.

The information gained from the unstructured data, however, is only as good as the user-created linguistic models using ICPA, Seton says. The accuracy of the linguistic and predictive models themselves depends on how well the model has been optimized by the user.

In terms of ROI, "The cost and angst to get all this information we need in structured data fields is pretty tremendous," says Froehlich. "This has the potential to reduce the need to provide a data field for every possible piece of data and of making everyone type everything into every single field ... which drives people up a wall. Intuitively we believe there is a cost benefit."

Seton Healthcare senior epidemiologist Christine Jesser concurs. "Highly skilled clinicians are spending inordinate amounts of time entering data into structured fields and this can reduce the time and effort it takes." While clinicians were already aware that smoking is an important consideration when looking at someone with CHF, she says the tool came back with some results they didn't expect for predicting the probability of a person's readmission to the hospital.

"The interesting things we found were their living status and were they in assisted living situations, and whether they had drug and alcohol abuse,'' Jesser says. "Those were some social factors that were only found in unstructured data that emerged as important predictors in the model."

If ICPA turns the data into information they can act on, "we get lower costs and a much faster way to get to information we provide,'' says Froehlich. "That's where our big hope is ... improving care without spending as much."

Working differently

"We're at the cusp of a whole new way of [looking at] medical research,'' says UOIT's McGregor. "Data is multidimensional ... we have new types of data in just the neonatal setting we're looking to collect,'' such as brain activity and drugs being infused. Beyond looking at infection and apnea in preemies, they're also starting to look at other conditions, like hemorrhage of the brain in babies and adults. "We're targeting the most life-threatening conditions ... where we can make a significant difference."

IDC's Feldman acknowledges that "the technology itself is never going to be perfect... but computers, unlike people, are consistent." People can make judgments, though, and computers can't, so "if you combine the two, the outcome is more powerful" than relying on one or the other alone. Computers can "boost a physician's understanding of a patient" by crunching through more information than a human possibly could, and by then finding patterns in that information.

She sees predictive analytics and structured data making serious inroads within health care in the next five years, which will result in reduced costs and fewer adverse situations. The ability to use information better, says Feldman, "will substantially alter the future of health care."

Esther Shein is a freelance writer and editor. She can be reached at eshein@shein.net .

Read more about healthcare it in Computerworld's Healthcare IT Topic Center.

This story, "Data analytics driving medical breakthroughs" was originally published by Computerworld.

Top 10 Hot Internet of Things Startups
Join the discussion
Be the first to comment on this article. Our Commenting Policies