This article is more than 5 years old. Information may no longer be current.
What challenges have you seen or do you foresee in combining, drawing conclusions from large administrative databases?
Click Here to Manage Email Alerts
Click here to read the Cover Story, "Big data may change how orthopedists are paid."
Must develop resources, relationships
Health care has seen an explosion of information that is approaching the order of yottabytes (1,024 gigabytes). Although randomized controlled trials are the only true way to establish causation, the use of observational big data can be less expensive, can evaluate the heterogeneity of treatment effect and can give feedback about ongoing processes. Other specialties (cardiology, transplant surgery, etc.) have established robust registries to record and evaluate patients’ outcomes, but orthopedic surgery has mostly lagged, save for some arthroplasty registries (Michigan Arthroplasty Registry Collaborative Quality Initiative, AJRR), which have shown promising results. Most health databases queried today (Medicare, National Inpatient Sample, etc.) use administrative data, which are not specific to orthopedic surgery, and thus lack laterality, implant type and specific orthopedic diagnostic/classification. To make robust conclusions and recommendations which the public demands, we, as a specialty, need to take ownership of not only the data, but also its evaluation. There has been good initial work by Saleh and Shaha looking at existing big data through an orthopedic lens. In “big data” articles by Shaha and colleagues and Anoushiravani and colleagues, the benefits and challenges of using these data were highlighted, as well as the potential to improve both health care quality and value through the development of national orthopedic registries. To accomplish this, we must collaborate not only with those outside of our field, such as in economics, social sciences and other medical specialties, but also collaborate across traditional institutional lines to develop resources and relationships if we are to be successful in this important enterprise going forward.
- References:
- Anoushiravani AA, et al. Orthop Clin North Am. 2016;doi:10.1016/j.ocl.2016.05.008.
- Shaha SH, et al. Orthop Clin North Am. 2016;doi:10.1016/j.ocl.2016.05.009.
Bryant W. Oliphant, MD, MBA, is an orthopedic surgeon affiliated with the University of Michigan and the Detroit Medical Center.
Disclosure: Oliphant reports no relevant financial disclosures.
Distinguish source, methods
Information analysis from large databases can complement other methods of orthopedic outcomes research. The sheer number of registered patients in administrative databases increases the power of the data, but conclusions from the analysis should be carefully interpreted.
It is important to distinguish the source and methods of “big data.” Mining insurance company records is much different than well-designed multicenter orthopedic outcomes research, such as the MOON and MARS studies. The identification of specific diagnoses or procedures from hospital or insurance ICD-9 or CPT codes can produce data that are: nebulous (ie, do not specify surgical or non-surgical extremity); flawed (ie, recorded by non-medical personnel); biased (ie, designed for billing and not medical research); inconsistent (ie, variable recording practices by individual or registry); or incomplete (ie, terminology changes, new diagnoses/procedures over time).
Conclusions drawn from large database investigations need to be validated by other research strategies. Statistical differences from a large sample size may not reflect clinical significance. Linkage to a medical records system makes it possible to conduct population-based descriptive, case-control, prospective cohort and cross-sectional research studies. Improvements in database methodology to ensure more accurate recording of information combined with rigorous analysis will improve the clinical utility for orthopedic surgeons.
Michael J. Stuart, MD, is an Orthopedics Today Editorial Board member who practices at the Mayo Clinic.
Disclosure: Stuart reports he is a consultant for and receives royalties from Arthrex, and he receives research support from Stryker and the USA Hockey Foundation.
Does not account for clinical issues
Research involving medical databases can offer some unique opportunities in orthopedics. These large databases can, at a low cost, allow analysis of large patient groups. Frequently, their best use is to identify trends in medical care or outcomes that can then be more carefully evaluated by traditional outcomes research. These databases can help determine rare or unusual findings due to their size. Unfortunately, there are several potentially significant drawbacks for use of the databases. Concerns include the fundamental reason the data is collected, which is for financial administrative use, and thus diagnostic and procedural coding may be influenced by reimbursement issues. Studies have demonstrated the challenge of avoiding clerical mistakes in data entry for these databases. The most significant concerns revolve around clinical issues not accounted for in the database and minimal assessment of outcomes that do not involve additional procedures or treatment. Preoperative issues regarding indications and clinical parameters cannot be identified. No PROs are obtained and any type of structural outcome involving physical exam or imaging will be lacking. Big data is potentially powerful in identifying trends, but clinicians must understand the limitations in applying any clinical findings.
Rick W. Wright, MD, is the Jerome J. Gilden Distinguished Professor of Orthopaedic Surgery at Washington University School of Medicine in St. Louis.
Disclosure: Wright reports no relevant financial disclosures.