With a new biomedical informatics center, Brown is taking a deep dive into big health data.
On a chilly evening in downtown Providence, a gaunt man is rushed to the hospital, unconscious. He knows no one; no one knows him. Yet his doctor, whom he’s never met, already has an intimate knowledge of his health. His history of asthma. His heart problems. His struggle with opioid addiction.
This man is fictional, but his circumstances might well be real. New electronic health data systems are improving care for countless patients by sharing their data instantly—not just with physicians, but with pharmacists, public health researchers, academics, and scientists.
If harnessed correctly, circulating “big health data” among all these players may change the face of medical treatment. The information is useful not just in the clinic, but behind the scenes, helping to inform public health policy, to create new treatments or consumer products, and even to shed light on the mysteries of the human genome as they pertain to human health. In order to use it, though, you first need to build the right tools and safeguards to work with vast datasets.
Meeting that need is the driving force behind the new Brown Center for Biomedical Informatics (BCBI). Its founding director, Indra “Neil” Sarkar, PhD, MLIS, and associate director, Elizabeth Chen, PhD, who are also associate professors in both the Warren Alpert Medical School and the Brown School of Public Health, are creating ways to harness electronic health data securely from a wide range of sources—like electronic health records, gene sequences, or even wearable devices like a Fitbit.
“[The health care community] is collecting information that we can use to learn about our patients. We can use it to study populations, develop models, and predict if a patient is at risk for a particular illness,” Chen says. “We’re all about trying to improve health in general—using data, data science, information technology, or some combination of those.”
Through a unique blend of collaborative research, educational outreach, and community partnerships, she says, the work being done at BCBI could help create a new personalized standard of care in Rhode Island and beyond.
Statewide Experiments
Both Sarkar and Chen have celebrated careers working with biomedical and health data, and have been collaborating since graduate school—so much so, they’ve been called an “informatics power couple” by a close colleague. Before coming to Brown in 2015, the pair—both elected fellows of the American College of Medical Informatics—worked at the University of Vermont, where they developed projects that they’re now continuing at BCBI. Chen’s work has focused mainly on clinical informatics, finding ways to extract and analyze information in electronic health records; Sarkar has worked largely in translational bioinformatics, creating tools to sort through mountains of genetic and evolutionary data in the lab.
When the opportunity arose to start a new center at Brown, they say, it offered an ideal place to put their multifaceted research into practice. “Rhode Island is a perfect petri dish to do population studies and longitudinal studies with patients,” Sarkar says. “The state is the nation’s smallest geographically, but at least 60 or 70 percent of the people who are born here, stay here. And because we have only one medical school that partners with the largest civilian health care providers in the state—as well as the Department of Veterans Affairs—we can figure out a way to coordinate all those data into one spot, and can do very powerful population analysis.”
One of Sarkar and Chen’s first goals for the center is to improve access to health data across the state. “BCBI’s greatest advantage is our access to those data,” Chen says. “It can be challenging for researchers to get that, and there’s no systematic process for it.”
Right now, they are working to create partnerships to share data between health care systems like Lifespan, Care New England, and the Providence VA Medical Center, as well as with the Rhode Island Department of Health and Rhode Island Quality Institute (RIQI). The overall goal of this collaboration, Chen says, is to build a computing infrastructure at Brown that can intelligently comb through disparate systems and pull relevant data from patients’ health records—essentially creating an ultra-secure, “one-stop shop” for health information that reflects the state’s entire population.
Those new computing tools will build on work that the state is already doing, says Laura Adams, president of RIQI, a nonprofit that works to solve regional health issues. (The group also runs CurrentCare, Rhode Island’s health information exchange.) By spreading health information freely between all care providers, she says, the state is able to provide emergency departments with lifesaving information for unconscious patients, like our hypothetical stranger. Before CurrentCare, if the patient wasn’t in a particular hospital’s health records system, doctors had to guess at his medical history when his gurney was wheeled in the door—meaning they had no way of knowing in advance if the patient was deathly allergic to a certain drug, or taking medication that might interfere with treatment.
“[Doctors] had to make some clinical judgments on the basis of very limited data. In some cases, actual harm was done to the patient that could have been avoided if they had data from competing systems,” Adams says.
Letting both clinicians and researchers access this data will be essential for improving care, Chen adds. In addition to giving new insight in the clinic, she’s hopeful that CurrentCare will be able to inform new scientific studies.
“That’s one of our big initiatives,” she says. “We want researchers to be able to come to us if they’re designing a study and ask, ‘How many patients meet my inclusion/exclusion criteria? If I’m doing a study on asthma, how do we actually find those patients?’”
One of the challenges of doing this, she notes, will be creating software that can deal with many different formats of health records. Certain hospitals may use a records system that’s formatted in a different way than another’s. Even a clinician’s own note-taking habits come into play—some may record a patient’s care as a narrative paragraph, whereas others create something akin to an Excel spreadsheet—yet each version has the same valuable information scattered within it.
To make sense of these data, Chen says it’s necessary to develop computational tools that can analyze natural language and intelligently summarize data for researchers—not an easy task by any standard. According to Sarkar, though, it’s just part of the bargain when it comes to the work being done at BCBI.
“One thing that is essential in informatics is to avoid getting in the way of a doctor being a doctor,” he says. “We want the technology to work behind the scenes—so we spend a lot of time studying the process, studying what a care provider is doing, and teaching computers cues to follow through based on that. If we’re truly successful, clinicians and patients don’t even know we’re there.”
By looking deeply into data from multiple hospitals, the BCBI team is also hoping to uncover trends in patient outcomes that hint at how well specific interventions actually work. Ideally, Sarkar says, these tools may help fine-tune best practices that clinicians follow in emergency situations—not just by improving them for the field as a whole, but by customizing them to the needs of patients in specific geographic areas.
Decisions, Decisions
Orthopedic surgery offers one example. Given the icy winters of New England, slip-and-fall accidents happen at a much different rate here than in, say, Miami. Surgical treatments for broken bones and torn ligaments are costly not only in terms of dollars, but in quality of life: recovery, especially for elderly patients, can be long and painful after going under the knife. So, Sarkar asks, in each location, what factors lead a surgeon to choose either a full hip replacement or care that would simply make a patient more comfortable?
“Medicine is very well informed, but most of the guidelines and standards are based on measures that were developed at a single institution—and that place may be nothing like your institution,” Sarkar says. “We have the opportunity to look at all the data produced by patients in a single region, and build a quality registry to guide physicians working there. That’s our grand vision.”
In addition to shaping best practices, those data could also help clinicians decide when to use specific procedures or treatments in complex cases, especially ones where time is of the essence. Powerful computer algorithms may help them narrow down and diagnose ailments that aren’t immediately obvious, and automatically suggest the best course of action to support their own judgment.
“As a young intern or resident, there is a massive amount of information that you are expected to be the conduit for,” says Cedric Priebe, MD. “You’re taking information from the laboratory, from diagnostic imaging, from bedside monitoring, the pharmacy, pulling that all together, synthesizing it, and then sharing it with your team and your supervisors.”
Priebe is a practicing pediatrician and chief information officer at Lifespan. He’s working with BCBI to share electronic health data for research purposes, and plans to use those data to develop and improve existing decision support tools. Along with shaping best practices in the emergency department, he adds, these sorts of tools are helping to spot dangerous ailments that develop in patients over several days, catching them while they’re still easily treatable.
In cases like sepsis, a severe immune reaction that attacks all organs of the body, improved decision support would be potentially useful. At its onset, the condition can be difficult to spot, and if it advances, it can become deadly. By looking for a combination of early symptoms—like elevated heart rate, low blood pressure, high temperature, and high levels of lactate in the blood stream—new algorithms could automatically flag information that a single physician might otherwise miss.
“The ultimate secret sauce would be an algorithm that you can use in cases where you don’t know a problem is going to develop,” Priebe says. “That’s the essence of predictive analytics—it needs big data, sophisticated math, and electronic systems that can make those decisions actionable by caregivers.” In short, it needs areas of expertise that BCBI is helping to develop.
Data in the Genome
Although the center is focused mainly on improving care in health care settings, it’s also working to provide data for researchers outside of the clinic. The same analytic tools and techniques used to comb through a patient record, Sarkar says, can also be adapted to spot patterns in genetic information. Our genome, after all, is just another form of coded data—a string of molecules laid out in a specific order—and finding patterns in data is one of the calling cards of biomedical informatics.
“We’re using informatics tools to look at the basic biology of an individual,” Sarkar says. “Understanding that biology, combined with understanding their lifestyle choices and environmental factors, will give us a precise description of an individual before we move them through the health care system. If we have enough information about those individuals at the genomic level, we might be able to better customize their treatment.”
Sarkar is also examining genetic data to understand how certain diseases develop in the first place. He’s continuing work he started at the University of Vermont, where he developed new techniques to examine genes associated with Alzheimer’s—a disease that is exceedingly difficult to study. No definitive diagnosis can be made until an autopsy examines brain tissue, and tracking its progression in living patients is a challenge.
It’s possible to simulate the disease in specially bred mice, but that’s by no means an ideal method, Sarkar says; first, researchers must “humanize” the mouse, altering it physically and genetically, by adding genes related to Alzheimer’s into its DNA. Only then can they follow the disease as it spreads in the mutant mouse.
Instead of going through all the work of adding a gene into a tiny mammal, Sarkar says, it may make more sense to study simpler organisms that have the gene already. By carefully analyzing the gene’s evolutionary history, he was able to trace it back to animals that have been around since the evolution of metazoans—creatures like sea squirts and anemones. In their DNA, he found, they carry a close relative to the genes associated with Alzheimer’s in humans, meaning that they could be ideal for understanding how the gene itself works, and could help identify the process that causes damaging plaques to form in the brain.
Eventually, this approach could lead to new treatments for a number of conditions. Sarkar is developing similar techniques to study parturition. Better understanding of the genes associated with childbearing, he says, may offer a better understanding of certain pregnancy complications, like preeclampsia or spontaneous preterm birth.
“In many ways, we’re detectives, trying to piece together [disease]genes’ stories using informatics,” Sarkar says. “But the techniques that we use to identify which organisms to look at—data mining, machine learning—all those things are identical whether we’re looking at a clinical context or a biological context.”
From sea squirts to humans, BCBI is working to share genetic information with researchers across the nation—and just like electronic health records, Chen and Sarkar say it’s essential to set up a secure and standardized infrastructure to house those data. It may be several years before they’re ready to open up genetic and health information for national use, but right now, they’re already starting to experiment locally. Advance Clinical and Translational Research (Advance-CTR), an NIH-funded partnership between Brown, the University of Rhode Island, and several state health care systems, is already using local data to develop real-world treatments for patients in Rhode Island.
Getting the Word Out
When it comes to using biomedical informatics, however, both Sarkar and Chen feel that the next big improvement will come not through any particular research initiative, but through education. No matter how good data tools become, they’re useless unless future academicians, clinicians, scientists, and even patients know how to use them. In that sense, Chen says, promoting data literacy will be key.
“I think we have a lot of work to do when it comes to both training medical and other students about informatics and working more closely with patients,” she says. “We can develop all these tools and all these systems, but are they really useful to the patient and clinician? We have to do more about involving them, educating them, and developing solutions that are directly useful to them.”
Sarkar echoes that sentiment wholeheartedly. One of his major educational initiatives at BCBI is expanding not only a scholarly concentration for Warren Alpert medical students, but eventually creating master’s and PhD programs and hosting classes for undergraduates.
“Brown’s academic culture of letting students take their own path is a big part of why our work is thriving here,” Sarkar says. “It allows a biology major to take a computer science class, and no one raises an eyebrow. It allows a medical student to explore computer stuff. That’s baked right into the Brown culture.” These students are also incredibly self-motivated, he adds. “We showed up in July 2015, and immediately four medical students came knocking at our door, wanting to work with us. In a very short time, the original four grew to 10 students. We were just flooded.”
The immediate interest, he thinks, may stem from the fact that informatics is an interdisciplinary field, so a growing number of students’ interests—be they policy, computer science, biology, or public health—overlap widely with the work being done at BCBI.
In a student hackathon cosponsored by the center in 2016, Sarkar and Chen worked closely with students in a variety of disciplines to create new health care innovations over the course of one weekend. The ideas varied widely, from custom breast prostheses to apps that provide information to surgical patients.
Sachin Pendse ’17, an international relations and computer science concentrator, worked on software that could help patients identify the side effects of medications and offer suggestions to cope with them, like deep breathing, meditation, or alternative medicine. Informatics, he says, would play a big role in improving the app’s content. By collecting data on what helps patients in need (and what doesn’t), its code could be tweaked to make better suggestions, or to shift its focus to more successful interventions.
“Neil was an amazing mentor for that sort of thing,” Pendse says. “He ended up giving us his personal cell phone number and saying, if you want any information about how biomedical informatics works in the real world, just text me. I’ve never had that sort of access before.”
Pathway to Informatics
Until BCBI was founded, students at Brown had no direct pathway to biomedical informatics, and had to find their own way in a growing discipline. Biomedical informatics, according to the American Medical Informatics Association (AMIA), “studies and pursues the effective uses of biomedical data, information, and knowledge for scientific inquiry, problem solving, and decision making, motivated by efforts to improve human health.” BCBI is working with partners across Brown, including the Center for Computational Molecular Biology, School of Engineering, School of Public Health, and the Data Science Initiative, to develop complementary educational offerings for students at all levels. Many current giants in the field—like Isaac Kohane, MD, PhD ’81, inaugural chair of the Department of Biomedical Informatics at Harvard Medical School—started out by studying unrelated disciplines at Brown, and entered a career in informatics without the benefits of direct mentorship.
“I think having a center like this when I was a student would have caused me to focus much more rapidly on applications to medicine,” says Kohane, who studied biology as an undergraduate. “There are still only a small number of schools that have excellence in both biology and in computing in the way that Brown does, along with leaders in informatics like Neil and Liz to train people in the discipline. I’m really pleased to see that this center has been established, and I’m confident it means the next generation of leaders in biomedical informatics will be from Brown.”
That’s a goal the BCBI team is working toward. In addition to teaching and mentoring within the greater Brown community, Sarkar and Chen are also focusing on students before they even reach the undergraduate level. They’ve created summer programs and workshops for high school audiences, where they aim to seed an interest in informatics from an early age.
“Neil and Liz are pioneers in that area,” says Doug Fridsma, MD, PhD, president of the AMIA. “At first, we all thought it was a little crazy, but if you think about it, if you’re a kid interested in technology or computers, and you also want to do something that helps people and contributes to society, informatics provides a pathway to combine those interests.”
For both Sarkar and Chen, it doesn’t matter how old or young a student is when they enter the field, so long as they share a passion for positively impacting human health. Using data, they say, can unite expertise from a number of different fields, and have a large-scale impact on patients.
“We’re all on the same team here,” Sarkar says. “In the end, it’s all about providing better patient care, with the best available data, in the most efficient way, whether it be through inspiring students or working directly with researchers and clinicians.”
