CHICAGO—Big data, also known as data science, is an exciting, emerging field of research, and the opportunities are expanding, according to panelists of Answering Clinically Relevant Questions Using Large Datasets, a session at the 2018 ACR/ARHP Annual Meeting.
In 2013, the National Institutes of Health (NIH) launched Big Data to Knowledge (BD2K), an initiative to facilitate the use of large biomedical data sets for research, the design of new research tools and methodologies, and training researchers. The initiative promotes discovery in this new landscape, said Kenneth J. Ottenbacher, PhD, OTR, Russell Shearn Moody Distinguished Chair in Neurological Rehabilitation at the University of Texas Medical Branch, Galveston. He is also a principal investigator at the Center for Large Data Research and Data Sharing in Rehabilitation (CLDR).
Accessible, Shareable Data
“Data need to be findable, accessible, interoperable and reusable. That means the data are out there and can be used again,” said Dr. Ottenbacher. “This [approach] is a new concept of data science within the NIH. There are so many new, different ways we can use these new, different types of data. We should be trying to take advantage of that.”
Big data in biomedical research includes the National Patient-Centered Clinical Research Network (PCORnet), a program designed to make clinical research faster, easier and cheaper by using large-scale data sets. PCORnet’s Open Door program provides training to researchers and access to clinical data.
“Another area of research that is gaining interest is data from electronic health records [EHRs],” said Dr. Ottenbacher. “Healthcare systems are expanding and merging, and paying attention to their data through Epic or other medical records systems. There is a lot of interest in how we can use those EHRs to not just plan for care, but also understand care better. We do a lot of work with Medicare data. In five to 10 years, most of the people who do what we do won’t be using Medicare data anymore. They’ll be doing their research with [EHRs].”1
Large data sets available to investigators include:
- All of Us, the cornerstone of the NIH’s Precision Medicine Initiative, which has a goal of recruiting 1 million volunteers to provide health data that will be available for clinical research;
- The National Institute of Child Health and Human Development’s Data and Specimen Hub (DASH), a bank of de-identified clinical data from its funded studies for use in secondary research;
- The Multicenter Osteoarthritis Study (MOST), a collection of patient case report data and knee joint imaging; and
- The Osteoarthritis Initiative (OAI), an NIH-sponsored observational osteoarthritis (OA) study that offers investigators a large bank of raw data and more than 20 million images; and
- The Arthritis Foundation and the OsteoArthritis Society International (OARSI) co-sponsored a 2016 study to demonstrate how to correctly classify OA patients using the national clinical repository of the U.S. Veterans Health Administration.2
Researchers may find relevant data sets through the NIH-supported bioCADDIE, a searchable index of data repositories available for sharing. The NIH is also funding grants to teach researchers how to archive data to encourage data sharing, said Dr. Ottenbacher, who works with CLDR, which funds pilot projects and archives data related to physical disability, rehabilitation and recovery.
“Many studies in rheumatology would qualify. We have data sharing, and people who want to archive data from their completed studies. We provide a way for them to do that through the Archive of Data on Disability to Enable Policy and Research (ADDEP),” he said. CLDR’s Visiting Scholars program supports investigators for up to six months to work with researchers who are experienced in using large data sets.
Function & Utilization Data
Soham Al Snih, MD, PhD, associate professor, Division of Rehabilitation Sciences, University of Texas Medical Branch, used large data sets from the Centers for Medicaid & Medicare Services (CMS) and the Hispanic Established Populations for the Epidemiological Study of the Elderly (H-EPESE) for a new analysis of patients with OA, rheumatoid arthritis (RA) and spondyloarthropathies (SpA).
In a study supported by the NIH National Institute of Aging (NIA), she and her co-investigators compared self-reported arthritis data with data from CMS files for diagnosis sensitivity and specificity. They also examined sociodemographics, clinical characteristics and functional outcomes, provider types and level of healthcare use by different patient groups. The H-EPESE database includes archived patient questionnaire data collected from Mexican-Americans aged 65 years and older since 1993. They also analyzed CMS beneficiary summary file data, including ICD-9 billing codes and fee-for-service claims. Out of 3,952 original participants, 1,518 were used in their final sample.
“When we looked at sociodemographic characteristics by arthritis groups, we could see there were no differences by age” between those with self-reported arthritis and patients who met the CMS Chronic Case Warehouse definitions of arthritis, Dr. Al Snih said. “The group with no arthritis tended to [include mostly men], but the other groups with arthritis were more likely to [include married women and were] more likely to [include people under] financial strain compared with those who had no arthritis. When we looked at the distribution of clinical characteristics, we found that the groups with arthritis were more likely to report pain, their BMI was higher, they reported more depression and had more comorbidities using the Charlson index.”
People with arthritis had higher healthcare use compared with those without the disease, including 97.4% with one or more outpatient visits in a year, she said.
When investigators analyzed data specific to the types of arthritis, they found patients with RA and SpA had more hospital admissions and emergency department visits than OA patients. Additionally, more than 80% of patients in each arthritis patient group reported at least one functional limitation according to the Instrumental Activities of Daily Living and Activities of Daily Living (IADL-ADL) index. In the study, 32.3% of RA patients had been treated by a rheumatologist, compared with 5.6% of the OA patients and none of the patients with SpA. Also, 47% of the SpA patients were treated by a physical or occupational therapist based on the claims data.
In the future, the investigators plan to link their data with area resource files on locally available healthcare services, “because this is a population with high rates of disability, and we can see that a majority of the participants are seen by primary care providers and very few by rheumatologists,” said Dr. Al Snih.
Carpal Tunnel Clinical Care
“Large data sets are extremely powerful. We can use them to describe epidemics, to determine our diets, to estimate risk, to define the typical and, therefore, the atypical—and all of these things are extremely useful to us as clinicians in understanding who we treat and how we treat them,” said Nancy A. Baker, ScD, MPH, OT, associate professor of occupational therapy, Tufts University, Boston. “There are numerous questions that can be asked related to large data sets. We can ask prevalence and incidence questions. We can ask characteristic questions related to specific populations and, maybe, how those differ from general populations. We can look at risk factors and how diseases, treatments or outcomes change over time.”
Dr. Baker studied patterns of clinical care among patients with carpal tunnel syndrome using the Clinformatics Data Mart, a repository of claims reimbursement data from a large U.S. health insurer. Treatment guidelines for carpal tunnel syndrome were published in 2014, recommending surgery for patients with severe disease and conservative care, such as splinting, for others. “But what hasn’t been defined well are what conservative treatments really seem to work well for people with carpal tunnel syndrome,” she said.3,4
Dr. Baker and her co-investigators searched the database for patients diagnosed for the first time with carpal tunnel syndrome from 2010–2012. They tracked each patient over a three-year period after their initial diagnosis to see what treatments they received based on ICD-9 codes.
Out of 19 million eligible patients in the database over 2010–2012, investigators identified 200,532 (1%) patients with a carpal tunnel syndrome diagnosis and filtered this group to a final sample of 24,931 carpal tunnel syndrome patients. These patients each had four years of continuous insurance coverage and a confirmed initial provider visit date.
Often, carpal tunnel syndrome is considered a work-related illness in the U.S. In past studies, treatment seemed to vary greatly depending on a patient’s insurance coverage, said Dr. Baker. Out of six possible treatments—surgery, positioning (splints), corticosteroids, heat-based treatment, manual treatment and therapeutic exercises—this study found 53.5% of patients diagnosed with carpal tunnel syndrome received no treatment at all, and 28.4% received one treatment. Overall, the most common treatment prescribed was surgery, followed by positioning and steroids. Patients who had only a diagnostic office visit were most likely to receive a splint, and patients with two or more visits were more likely to have surgery, she said.
“Patients weren’t really receiving treatments in a three-year period after a diagnosis of [carpal tunnel syndrome], and they probably weren’t getting worse, because they weren’t receiving more treatments,” said Dr. Baker. Positioning may have been under-reported because patients could have purchased over-the-counter splints, she added.
Dr. Baker followed this study with a new, not-yet-published study on the potential influence of a patient’s insurance type on the carpal tunnel syndrome treatments prescribed, using data from the National Ambulatory Medical Care Survey. Out of 23,236,000 total carpal tunnel syndrome treatments prescribed from 2005–2014, 51.2% were splints, she said.
Susan Bernstein is a freelance medical journalist based in Atlanta.
- Coorevits P, Sundgren M, Klein GO, et al. Electronic health records: new opportunities for clinical research. J Intern Med. 2013 Dec;274(6):547–560.
- Grasso MA, Yesha Y, Rishe N, et al. A big data approach for selection of a large osteoarthritis cohort. Osteoarthr Cartil. 2016 Apr;24(suppl 1):S208-S209.
- Baker NA, Stevans JM, Terhorst L, et al. What types of treatment are provided for patients with carpal tunnel syndrome? A retrospective analysis of commercial insurance. PM R. 2018 Aug;10(8):826–835.
- Huisstede BM, Frieden J, Coert JH, et al. Carpal tunnel syndrome: Hand surgeons, hand therapists, and physical medicine and rehabilitation physicians agree on a multidisciplinary treatment guideline—results from the European HANDGUIDE study. Arch Phys Med Rehabil. 2014 Dec;95(12):2253–2263.