Health

Health Informatics Researchers Tackle Health Care Data Challenges

By Josh Ambrose

Janusz Wojtusiak

Old hard-drives line the top of bookshelves in Mason computer scientist Janusz Wojtusiak’s office; some as small as a deck of cards, others as big as the final Harry Potter tome. You might not know it to look at them, but those ubiquitous beige boxes represent both the history and the future of one of the simplest, yet most daunting challenges for modern health care—the sharing of information via technology.

“At its most basic level, we’re studying the intersection of technology and health information but not to the exclusion of either,” says Wojtusiak, who is the director of Mason’s Center for Discovery Science and Health Informatics. “We’re studying the way that important data are communicated in health care, the way that information flows, and ways in which technology can become smarter to help in the process.”

But surely doctors and nurses share information with each other all the time, right? Would that were so, asserts Wojtusiak’s colleague Hua Min. Even when health care professionals do share information, there are a number of challenges in that communication—not the least of which is the actual meaning of the words themselves.

“Let’s say you overhear a conversation between two people, and they keep using the word apple. Are they talking about fruit or about computers?” asks Min. “If you’re on the street, usually the context isn’t that difficult to parse. But when it comes to sprawling databases of medical records…well, it’s a lot more complicated.”

Hua Min

A long-time collaborator with the National Cancer Institute, Min says researchers in health informatics see this problem often. She uses cancer drug trials as an example. “Researchers in one lab may come up with new terms to describe a phenomenon that’s already been recorded with different terms in a different clinical trial just a year or two earlier,” says Min who studied medicine in her native China before coming to the United States. “To help researchers make progress in the most expedient way possible, we simply have to ensure that we keep finding ways to link the old and the new.”

It is basic communication difficulties such as these that are also surprisingly common outside of clinical research, according to Priya Nambisan, an assistant professor of health informatics and health care management in the College of Health and Human Services. “It’s part of what makes it difficult for the ER doctor you consulted when you broke your foot on vacation to pass on the information to the new family physician you recently began visiting back home. Both their terminology and their technology might be completely different. The solution is that health care providers need to agree on national standards.”

Min agrees. “This is why different doctors are always insisting on administering the same test themselves. We need standards for the standards. Right now, in an already maxed-out health care provider system, doctors are increasingly failing to consult the labyrinth of prior physician-entered electronic medical records at hospitals—let alone patient-provided personal health records (PHRs)—because of the overabundance of information and the difficulties in deciphering them.”

In addition to repeating tests, physicians rely on the patients’ oral accounts. “Often the fastest way to proceed is to simply ask patients basic questions like, are you allergic to anything? before proceeding. It’s a process fraught with the possibility of miscommunication or inadvertent oversight,” says Min.

Priya Nambisan

The sheer amount of information recorded in today’s medical environment is an equally daunting challenge.  Mason’s Phan Giang points out that “billions of pages of clinical text” are generated each year in the United States. Yet, as the average clinician has less than seven minutes to consult patient records, they often only skim the surface of the recorded data. What if they had computer tools to help them analyze the data?

“It’s relatively easy for computers to report unambiguous information—structure data—such as height and weight,” Giang says. “Where it gets tricky is with diagnoses and other physician-variable terms—clinical text.”

This text, as Min shared, can contain a great variety in terminology. To successfully analyze clinical text is hugely challenging and requires the intelligent analysis of many contextual terminologies. “One of the goals here at the [health informatics] program is to bring as much of that clinical text into the structure data, so that our software models can more easily and accurately interpret the data.”

Textual processing isn’t the only kind of health informatics that interests Giang. He is also researching “decision making and uncertainty” as they relate to scenarios such as the depression and suicide rates of veterans from Iraq and Afghanistan, which are alarmingly high.

Phan Giang

“How do doctors address this challenge? Primarily, at the moment, by prescribing antidepressants,” says Giang. “Where the uncertainty in this scenario comes in, however, is  that these medications haven’t necessarily been tested on that population. For example, an African American soldier may react rather differently from the suburban white woman on whom a medication may have been tested. We have the efficacy and knowledge for one population but not for the other.” In response, Giang’s studies attempt to analyze and rate the uncertain factors so that patients and their health care providers can more efficiently decide on the most appropriate course of action.

“The need for research and innovation in the way we share information via technology is growing by the week,” says Wojtusiak. “Health care information is the most diverse and complex data out there.”

It’s in response to that challenge that he runs Mason’s Machine Learning and Inference Laboratory (MLI). Of course, the patient information that can now be entered is far more than just disease codes on a patient’s electronic record keyed into a computer terminal—more and more unstructured data sets, such as physician notes and images, are also recorded. At the MLI,  Wojtusiak works with graduate students to develop learning algorithms to analyze complex health care data culled from computer databases and new methodologies to compare data from published studies with the latest patient information.

“Our goal is to create methods that use all available information, not only clean data sets collected in controlled environments,” says Wojtusiak. “Information comes from clinical trials, patient records, published studies, public health data sets, existing terminologies, and so on. The challenge is how to use all of that.”

One such study is examining the feasibility of predicting payments of medical claims by analyzing patterns of how insurance companies (which often determine the affordability of treatment) respond to specific billing information.

Nambisan, on the other hand, has worked for years on analyzing the ways that patients communicate with health care providers online, from forums to social media interaction. One of the projects that she is currently leading at Mason is research on improving the way that PHRs are handled. With a team of students, she is working on the development of a software program to allow patients to carry their own PHRs with them everywhere on their mobile electronic devices.

“With an easily available app, patients could now carry extensive, but private, password-protected medical histories with them everywhere they went,” she says as she shows off the beta interface design of the program, noting its familiar search capabilities, cross-indexing, and term definitions. An app could revolutionize the way both patients and physicians interact with health care records so the interest in such a tool is high.

“Companies as big as Google and Microsoft have offered up new programs for PHRs; however, no one has fully harnessed the mobile electronics as of yet,” she adds. “And more important, they haven’t offered anything in the way of standardizing the terminology in the records.”

This article originally appeared in a slightly different form on the College of Health and Human Services website.

To read more stories about Mason, check out the university’s News site.

Meet Mason Students

Volgenau School’s Engineers Day

Keith Sullivan

Mason PhD computer science student Keith Sullivan leads Mason’s robotics team, the RoboPatriots, who compete internationally each summer. Right now, Sullivan and the team are preparing to compete in the 2012 RoboCup in Mexico City, Mexico. The RoboPatriots compete in the kid-size humanoid league with robots that are 30 to 60 centimeters tall. The robots […]

Meet more Mason Students!