Big Data: What Every Rheumatologist Should Know

With each passing year, big data has taken on a larger role in medicine and more specifically, in rheumatology.

“Here is the thing about big data, it is not new. Health plans have been using big data for a long time,” Catherine MacLean, MD, PHD, from Hospital for Special Surgery, told Healio Rheumatology. “What is new in big data is that these sources of data are now becoming more available.”

To better understand big data, Healio Rheumatology spoke with leaders of the Corrona, ArthritisPower, BRASS, ImmPort and RISE registries and discussed what these registries can provide for rheumatologists. In addition, they discussed the benefits of big data in general for rheumatology.

Corrona Registry

The Corrona Registry collects data from both physician and patients at the time of a clinic visit and does not require a physician or patient to recall this information later, according to its website. It also includes provider reasons for treatment changes and targeted adverse event forms, which are assessed by clinical experts for each adverse event.

“The Corrona Registry is doing well,” Joel Kremer, MD, FACP, from the Center for Rheumatology in New York, told Healio Rheumatology. “We are the only registry in the United States that collects data from clinical sites from both physicians and patients from a large geographic representation of different sites across the United States.”

He added, “We have more than 80 peer-review publications, full manuscripts [and] about 300 abstracts, and we just launched an inflammatory bowel disease registry.”

In addition to casting a wide net, Kremer mentioned that Corrona has incorporated a variety of disease types.

“We have registries now in rheumatoid arthritis (RA), psoriatic arthritis (PsA), spondyloarthropathies [and] psoriasis in collaboration with the National Psoriasis Foundation. Since our onset, we have involved 45,000 people. We have had 345,000 individuals in the registry. We had a total of 154,000 patient-years of data in the Corrona Registry. We are, by far, the biggest registry of RA on the planet.”

Kremer also mentioned the use of claims data, which is an advantage for the Corrona Registry.

“We are also using claims data, so we are the only registry in the world that can triangulate that,” he said.

In addition to standard data, Kremer discussed the collection of quality of life data through new wearable technology.

“We can also collect quality of life data, absenteeism and all kinds of standard outcomes from patients using wearables,” Kremer said. “We then feed this data back to the patients so they can see where they stand vs. others with similar disease.”

“We have invested almost $80 million in our registry over 15 years,” he added. “When it comes to big data, you get what you pay for.”

ArthritisPower

ArthritisPower is a registry that is funded by the non-profit Patient Centered Outcomes Research Institute (PCORI) and uses a free smartphone app or web interface to collect data on multiple validated disease-specific and generic patient-reported outcomes (PROs), symptoms and treatment during a snapshot of days, weeks or months from patients with rheumatic diseases, according to the app’s website. This information can be sent to physicians and researchers. As of June 2017, ArthritisPower has enrolled more than 6,000 patients, predominantly with RA, PsA and other types of spondyloarthritis.

“ArthritisPower is a patient-centered registry,” Jeffrey R. Curtis, MD, MS, MPH, from the University of Alabama at Birmingham, told Healio Rheumatology. “It is governed by patients predominantly with arthritis and other rheumatic conditions across the United States, representing different ethnic, geographic, racial and social diversity who serve on the steering committee, and who work with researchers to set the research agenda and the direction of the registry.”

Curtis discussed the ability of the app to use electronic health record (EHR) data and claims data.

“The data are able to be linked with and exchanged to other kinds of data to capture more than just patient generated information. Data sources we are linking to include lab results (eg, multi-biomarker disease activity test scores) and electronic health record data. There are well over 80 million people’s electronic health record data in this system (PCORnet),” Curtis said. “They also are starting to bring new data sources, including administrative claims data, into the mix.”

Curtis mentioned the advantages of real-world data, including patient data, vs. clinical trial data.

“The data that we have from clinical trials is typically focused on metrics that doctors recognize, but are not necessarily things physicians would measure in the real world,” he said. “For example, the most common outcome measure in a RA clinical trial are ACR20, ACR50, and ACR70 scores. If you then ask, ‘What fraction of U.S. rheumatologists measure ACR scores in clinical practice?’ The answer is nobody. We have as an outcome measure something that is well accepted, highly valid, but it is so impractical in the real world that nobody uses it.”

He added, “Even for tried and true measures like the DAS28, patients do not care a lot about that. They care much more about pain and physical function and others, like how they are sleeping, fatigue and whether they can go to work regularly. Can they participate with their family doing leisure activities? Those sorts of patient-reported outcomes are somewhat new to the rheumatology space. Just in the last 3 [years] to 5 years have both clinicians and researches started to take notice of the fact that we historically have measured a lot of things that patients do not care much about. Our failure to place enough importance on what matters most to patients sometimes results in challenges in implementing treatment strategies, such as treat-to-target, which may set a target (eg, DAS28 remission) using an outcome measure that patients don’t understand or may find somewhat irrelevant.”

BRASS

The Brigham and Women’s Hospital Rheumatoid Arthritis Sequential Study (BRASS) is a prospective observational study of more than 1,300 patients diagnosed with RA at the Robert Breck Brigham Arthritis Clinic in Boston. According to its website, the goals of the study are to determine biomarkers predictive of drug response and toxicity, as well as biomarkers predictive of disease activity and prognosis. In addition, the study aims to evaluate the natural history of RA through measurement of clinical, functional, psychosocial and economic outcomes.

“BRASS is a registry that is 15 years old. It is a single-center registry of patients with RA and the median follow-up of patients is approximately 5 years,” Daniel H. Solomon, MD, MPH, from Brigham and Women’s Hospital, told Healio Rheumatology. “We have long-term follow-up on hundreds of patients with linked electronic medical records from Brigham Women’s Hospital.”

Solomon added, “We have published approximately 70 papers and a similar number of abstracts over the last decade,” he said. “There is a huge databank that is stored for interrogation. Different papers have examined different aspects of RA, including outcomes, treatment patterns, biomarker profiles, genetic epidemiology, comparative effectiveness, and patient-centered outcomes. It is incredible and runs the gambit.”

However, there have been challenges. One has been patient follow-up over the extended duration of BRASS, which Solomon said is common for big data registries.

“Like every registry, one of the biggest challenges is follow-up with patients. Because BRASS patients are seen clinically at Brigham,” Solomon said, “we have high retention, about 70% retention, which is good for a registry. How do you follow patients? How do you keep them engaged and wanting to give you information at every visit? The other issue is when patients migrate out of our system and decide to stop following up with the doctors at Brigham, then it is hard to continue follow-up.”

In comparison to the other registries, Solomon said that although BRASS does not have as many patients as the Corrona Registry, it contains more types of data.

“BRASS is smaller in total numbers, but it has hundreds of variables and all kinds of biospecimens available for each patient,” Solomon said. “BRASS is a different but complementary model. Big data for many basic scientists requires having multi-dimensional data: genetics, transcriptomics, expression profiling, as well as clinical data. BRASS has that.”

ImmPort

The Immunology Database and Analysis Portal (ImmPort) is a centralized immunology data repository developed by the Northrop Grumman Information Technology Health Solutions team for the Division of Allergy, Immunology, and Transplantation (DAIT) at the National Institute of Allergy and Infectious Diseases (NIAID). It is made possible by partnerships with researchers at the University of California at San Francisco, Stanford University, University of Buffalo and the Technion – Israel Institute of Technology.

“The goals of ImmPort are to create an open access platform for research and clinical data sharing, to provide an integrated environment that brings in the usefulness of scientific data and advances the data-driven research, and extends the value of the scientific data in all areas of immunology research,” Sanchita Bhattacharya, ImmPort Science Project Lead, the University of California at San Francisco, told Healio Rheumatology. “The data shared through ImmPort are provided primarily by NIAID-funded clinical trial programs and major immunological research consortiums, and also by other research organizations.”

Bhattacharya described the creation of ImmPort. She said the data dissemination portal for clinical trials and immunology research data came into existence more than a decade ago with the intent of data sharing to the broader research community. Since then, ImmPort data contributors have shared 255 studies with a total of 49,319 patients.

“Once the data are deposited in ImmPort, our data curation team conducts data sanity checks and applies data standards and ontologies before the data are shared for research and knowledge dissemination. ImmPort promotes the open-access initiative so anyone in the world can download and analyze the data,” she said.

“ImmPort provides comprehensive information about each study including experiment or clinical trial design schema, study protocols, clinical assessments, meta-data and raw data from mechanistic studies for secondary analysis or repurposing of these datasets. We are developing tools that allow researchers to integrate multiple studies from different sources to answer clinical research questions that are untenable to address from a single study.” Bhattacharya said.

The database provides participant level de-identified information on demographics, disease types, interventions, vaccination responses, transplant outcome, allergies, concomitant medications, lab test values and beyond.

“Our group has successfully demonstrated the secondary analysis of an ANCA-Associated Vasculitis (RAVE) trial where we identified distinct subsets of granulocytes, as novel early markers, distinguishing patients achieving remission at 6 months following rituximab or cyclophosphamide treatment from those for whom treatment failed. This analysis provided novel study insight and discovery which may lead to more successful trials and therapeutic courses in AAV,” she said.

“Unlike other clinical trial data repositories, ImmPort does not require project proposal submissions for data access. A valid email address from your institution is sufficient enough to download the rich data source, in plain tab-delimited or structured query language, a structured format or using open-source tools, like RimmPort,” Bhattacharya said.

“If you are a physician or data enthusiast, budding bioinformaticians or seasoned computational biologist you can get access to individual and cohort level information on patients from ImmPort — a valuable resource for generating a reference cohort,” she said.

RISE

The Rheumatology Informatics System for Effectiveness (RISE) is a national electronic health record (EHR) registry from the American College of Rheumatology that passively collects data from practices. Participation in RISE can fulfill national quality-reporting requirements for the CMS Qualified Clinical Data Registry, as well as facilitating performance improvement by benchmarking quality metrics against regional and national standards. It also allows for data collection without individual patient consent.

“RISE takes identified data out of many rheumatologists’ electronic health records for the purposes of understanding treatment patterns and practice variation, improving care and satisfying some of the quality-reporting measures,” Curtis said. “RISE is also in its infancy and growing. The information contained in an EHR is what the doctor needed to put there to take care of patients and to at least minimally satisfy some of the quality metrics. That is for billing, documentation and quality reporting purposes. However, while data from a rheumatologist’s EHR has great value in and of itself, comprehensive capture of adverse events or efficient patient data capture will generally be lacking in an EHR-only data source.”

Joan M. Von Feldt, MD, MSEd, from the University of Pennsylvania, described the potential of this registry.

“Our expectation will be somewhere around 25% to 30% of practices around the country [will be] connected to RISE,” Von Feldt, told Healio Rheumatology. “That would be a tremendous source of data.”

She also described the basic premise of the data collection.

“We are capturing what rheumatologists are doing,” she said. “We are not mandating any tests. We do not tell them what to do. We do not tell them how to do it. We do not tell them how to record it. We just record what they are doing. This is just a snapshot of what a rheumatologist is doing in terms of care of the patients.”

Von Feldt said that RISE captures data on every patient seen in the practice.

“Therefore, RISE has quickly become one of the largest registries of rheumatic diseases for not just RA, but orphan diseases like systemic lupus erythematosus and scleroderma,” she said.

Clinical Implications

As big data will have long-term impacts for many specialties, it also holds potential for rheumatology, MacLean said.

“With regard to long-term benefits of using big data in rheumatology, I think the big wins are that there will be data on more people. For rheumatology, that is particularly important because there are some fairly low-prevalence, low-incidence diseases, like scleroderma for example,” she said.

MacLean also said big data can transform clinical practice into a form of personalized medicine.

“I think that long term what we are going to see is more incorporation of big data into care delivery and the formation of clinical diagnoses and decisions,” she said. “I think that is going to be routine. Something that I and many others think about is in this realm of personalized medicine.”

Another impact of big data will be to better understand subgroups of patients in clinical trials, MacLean said.

“If you look at a study treatment X vs. placebo and on average let us say treatment X was not any better than placebo in that population, but a lot of times within these studies there are some patients who get better, there is a great interest in those people who did get better even though overall people did not,” MacLean said. “That is another place where big data is going to be helpful, particularly in rheumatology. It is going to give us the ability, over the long term, to look at those additional data elements so we can do a better job sub-setting out particular patient groups who have combinations of characteristics that we are not thinking about now.”

According to MacLean, there are two subspecialties in rheumatology that can most benefit from big data.

“Big data may help us better understand rheumatic diseases that we now bucket into big groups, for example, RA or systemic lupus,” she said. “There is a huge disease spectrum within each of those diseases, I would say more so with lupus, and big data may help us to parse out specific clinical subsets of patients [who] might respond differently to different types of therapy.”

Curtis said an exciting trend on the horizon with the use of big data is the ability to integrate multiple different data sources to facilitate predictive analytics. Interoperability to link EHR data, health tracker devices, biosensors, and pharmacy and medical claims data may allow rheumatologists to predict future health events or possibly triage problems before the problems become full-blown.

“This notion that data could be made interoperable is now starting to be ready for prime time. I think the trend to bring all this together — if not under a single roof, then at least accessible to a common analytic platform — is exciting to facilitate fantastic science and to improve the quality of care.” – by Will A. Offit

• References:
• https://arthritispower.creakyjoints.org
• Nasrallah, M, et al. Arthritis Res Ther; 2015;doi: 10.1186/s13075-015-0778-z.
• www.brassstudy.org
• www.corrona.org
• www.immport.org/immport-open/public/home/home
• Yazdany J, et al. Arthritis Care Res (Hoboken). 2016;doi:10.1002/acr.23089.

• For more information:
• Sanchita Bhattacharya, can be reached at University of California at San Francisco, 550 16th St., 5068, San Francisco, CA 94158; email: sanchita.bhattacharya@ucsf.edu.
• Jeffrey R. Curtis, MD, MS, MPH, can be reached at Faculty Office Tower, Room 802, 510 20th St., South Birmingham, AL 35294; email: jcurtis@uab.edu.
• Joel Kremer, MD, FACP, can be reached at 4 Tower Pl., Albany, NY 12203; email: jkremer@corrona.org.
• Catherine MacLean, MD, PhD, can be reached at Hospital of Special Surgery, 535 E. 70th St., New York, NY 10021; email: randolphs@hss.edu.
• Daniel H. Solomon, MD, MPH, can be reached at Brigham and Women’s Hospital, Building for Transformative Medicine, 60 Fernwood Rd., Boston, MA 02115; email: dsolomon@bwh.harvard.edu.
• Joan M. Von Feldt, MD, MSEd, can be reached at the University of Pennsylvania, Perelman Center South, 3400 Civic Center Blvd., Philadelphia, PA 19104; email: vonfeldt@upenn.edu.

Disclosures: Bhattacharya, MacLean, Solomon and Von Feldt report no relevant financial disclosures. Kremer reports he is an officer in Corrona, is a part-time employee of Corrona and has equity interest in Corrona. Curtis reports he is the co-principal investigator for ArthritisPower.