Big data may change how orthopedists are paid

Big data are now a mainstay in orthopedics, may help personalize patient treatment.

In the last 20 years, big data have become increasingly important in orthopedics. This field began with the development of orthopedic registries in Europe and expanded to include multiple musculoskeletal registries and databases worldwide that contain data on the outcomes of total joint replacement, ligament reconstruction, spine surgery and the incidence of trauma, among information that pertains to a host of other orthopedic procedures.

“[Big data has] become certainly popularized more recently due to an interest in making our health care system perform more predictably for an individual. We must find ways to reduce variability and cost drivers,” Frank J. Schwab, MD, chief of the Spine Center at Hospital for Special Surgery, told Orthopedics Today.

Right now, the main difference between large-scale orthopedic registries and databases in countries like Australia, New Zealand, Sweden and the United Kingdom and those in the United States is that the United States does not yet have a 100% capture rate, according to sources interviewed.

Image: Robert Essel

“In the United States, [data collection is] truly voluntary with the American Joint Replacement Registry (AJRR), and so it is not 100%,” Ryan M. Nunley, MD, associate professor in the department of orthopedic surgery at Washington University School of Medicine in St. Louis, said.

He noted patients may receive a joint replacement at a hospital that participates in the AJRR, but they have a revision procedure at a hospital that does not participate in the AJRR. That gap in the data may skew the data.

However, that does not necessarily mean all U.S. registries have low capture rates, Nunley said.

“There are some health systems within the United States that do have much better registries, including the Kaiser [Permanente] system, the California Registry, [the] Michigan Registry, where they are becoming more sophisticated,” he said.

New collection methods

The technological evolution has also supported a higher rate of orthopedic data collection by moving away from paper surveys completed during office visits to electronic surveys that patients can complete on a computer or cell phone on their own time, according to James N. Weinstein, DO, MS, who led the 7-year, $21 million Spine Patient Outcomes Research Trial or SPORT.

SPORT was one of the first forays into big data, in this case to support appropriate clinical care for patients with various spinal conditions.

Changes in data collection methods for trials like SPORT and those conducted by the International Spine Study Group (ISSG) have helped make patients more able and willing to submit information, Schwab noted. He said it has also facilitated data collection in a more systematic fashion.

“The use of the smartphone allows for an easy capture system, and the use of tablets and iPads also allows for an easy capture system. Not only is that capture system convenient and in the palms of the individual, but it is also cost-effective,” Kurt P. Spindler, MD, vice chairman of research for the Orthopedics & Rheumatology Institute at Cleveland Clinic, said.

Spindler, who is an investigator with the big data Multicenter Orthopaedic Outcomes Network (MOON) study, said the internet has made data collection easier by allowing researchers to remotely connect to different, secure databanks.

However, some sources told Orthopedics Today that big data does not always mean the quality of the result is better and that a big data approach to answering scientific questions is not appropriate for all investigations of orthopedic treatments.

According to Schwab, some research questions need a lot of data elements collected over an extended period and others are better answered using narrow data of high quality.

“Big data is not needed for a lot of things and not necessary, and can be just confusing,” Schwab said. “But in some respects, it can be powerful and sometimes the only way to answer certain questions. We just need to reframe it that big data is not the thing for everything.”

Pitfalls of big data

Image: Tina Hoffman

According to Schwab, there are some obvious pitfalls with big data. It can be burdensome to collect a lot of data, especially if patients are not engaged or if researchers do not have a good data collection system for pertinent parameters.

“I think more of it is having [data] in a format that is useful, being able to be clear about what each data point means in the database, definitions, timing and frequency, and then being able to make sure you have similar kinds of data across similar populations to be able to study it and use it in the most meaningful way,” Weinstein, who is professor of orthopedics at Dartmouth-Hitchcock Medical Center, said.

Furthermore, Weinstein said big data can be difficult to analyze, and researchers may have the tendency to misinterpret, over-interpret or under-interpret the information.

Daniel J. Berry, MD, the L.Z. Gund Professor of Orthopedic Surgery at Mayo Clinic, said another pitfall of big data is researchers may “find associations between things that are not causally related or that have little likelihood of clinical relevance.”

He added, “Another pitfall of using big data is that frequently the big data have been collected for a purpose other than clinical research. When the data have been collected for another purpose, it is possible the way questions have been asked or the way data are structured may lead to limitations or erroneous results.”

What is more, patient data gathered for various orthopedic databases and registries may not correspond with how the patients are doing, according to Nunley.

“Just because the implant has not been taken out does not mean it is performing quite as well as some of the others, if you look just purely at the survivorship,” Nunley said.

More accuracy in big data

Despite the downsides of big data, Weinstein noted data can lead to more accurate outcomes through the incorporation of different collection sources, including patient-reported outcomes (PROs), claims data from insurance companies and EMRs.

“The ability to aggregate data across multiple databases gives you a better ability to generate models that start to be able to prevent problems before they happen as you start to study larger and larger populations, especially those with rare diseases,” Weinstein said.

As more data elements may be collected, it is possible some causes of diseases or critical associated parameters that are not obvious can be identified, according to Schwab.

“If you do not collect something upfront, you may never know about something that is an important parameter,” Schwab said. “The fact that you can cast a wider net and have more data elements has helped us uncover a few interesting associations.”

Big data can also help orthopedic surgeons and hospitals determine utilization costs with certain procedures, Kern Singh, MD, professor in the department of orthopedic surgery and co-director of the Minimally Invasive Spine Institute at Rush University Medical Center, noted.

“If we know spine surgery is increasing at a certain rate, it tells us about instrumentation utilization; it tells us how long hospitalizations are for a particular disease process; it helps us identify on a higher level which disease process may be more susceptible or more likely to have complications so we can manage them preoperatively,” said Singh, who is founder and president of the Minimally Invasive Spine Study Group.

According to Schwab, the possibility of wider data collection allows physicians to personalize their understanding of certain diseases, which can lead to more effective treatments. This can be advantageous to the health care system and to patients.

“It is taking huge amounts of data and hopefully getting to a point where we can, in a personalized way for one individual, anticipate the best clinical option moving forward on how to treat that patient, have predictability in what the outcome of the treatment will be and have the ability to select a treatment approach that is aligned with what the patient is looking for and is the most efficient and cost-effective way to reach that goal,” Schwab said.

Personalized treatment

Examples of when big data have already helped choose the best patient treatment can be found in several orthopedic specialties. SPORT compared surgical and non-surgical treatments for three of the most common back conditions among about 2,500 patients treated at 13 U.S. sites.

“[SPORT] becomes a good decision tool for patients who are facing spine surgery,” Weinstein said. “In this age, we have the ability to use this data to help patients make decisions about care for their symptoms and their needs at the time that they are trying to make that decision, what we call informed choice or shared decision-making. But, in this case, we use data elements and questions from the patients themselves to help them decide and show them what their likely outcome will be depending on their choice.”

Similarly, MOON and the Multicenter ACL Revision Study (MARS), respectively, collected data for more than 3,500 and 1,200 enrolled patients. These projects used big data to help establish outcomes for patients who undergo an ACL reconstruction or revision, according to Spindler.

The Lower Extremity Assessment Project or LEAP was used to identify whether amputation or limb salvage was more appropriate for a patient after severe lower extremity trauma.

There are countless other examples of how big data have impacted the course of musculoskeletal treatments in the United States, sources noted.

Use caution

Despite the advantages associated with using big data in orthopedics, sources interviewed noted that researchers should remain cautious when they analyze big data because there is the risk it was not input correctly, for example, which can affect its accuracy.

“When you have these large administrative databases, you are relying on the accuracy of the data that is being input,” Nunley said. “You may have a lower-waged coder at one hospital who is not putting in the data accurately compared to another health system, which is a huge volume center that spends a lot of time on internal audits to assess accuracy and makes sure that it is quality data that is being put in.”

The output — and conclusions drawn from it — is fully dependent on the quality input of the data, he said.

Due to the possibility of incomplete or inaccurate big data, Singh does not recommend using big data for general purposes.

“There is value in seeing trends, but it is a little harder to say you should necessarily apply the data routinely for clinical practice,” Singh said.

However, he noted, the evolution and increased popularity of “surgeon-driven or surgeon-specific registries of de-identified patient databases” may lead to more accurate patient information being entered into databases.

Hand-in-hand with the accuracy of data, researchers must be cautious of the quality of the data and how they quantitate the information found, Weinstein said.

“Whatever we extract from data needs to be pragmatic and it needs to be realistic,” Schwab said. “That sounds odd, but if you collect a lot of data, you may find some statistical association between two things that have nothing to do with each other and have no pragmatic application, but tend to be a fluke of statistics. One needs to be cautious not to let the statistical analysis of big data drive our behavior or treatment if it does not make good sense and if it is not pragmatic and if it is not heavily vetted by looking at multiple datasets and having rigorous statistical approaches,” he said.

Changing orthopedic practice

When big data are collected accurately and analyzed well, it can change the way an orthopedic surgeon runs his or her practice, Spindler said.

“[With] the use of big data, you can operationalize a system where you are collecting the patient-relevant outcomes and judging whether there is an improvement in patient-relevant outcome based upon certain procedures or techniques or treatment doings,” Spindler said. “I think it will potentially revolutionize the way that we get paid and paid for performance and so forth.”

One way to improve big data collection and analysis in orthopedic surgery is by reviewing how other medical specialties use big data. Orthopedics can learn a lot about personalizing patient care that way, Schwab said.

“Oncology, infectious disease and areas of internal medicine ... have used big data to come up with targeted therapies, and I would say the areas of pharmacology ... have been rather successful,” Schwab said. “We should learn from that to recognize the pathologies we treat in orthopedics. While it is convenient to standardize the description and the classification of the disease, all the other layers that are relevant to one individual need to be taken into account and can offer us much more targeted treatments with more predictable outcomes,” he said.

Orthopedics should also look to cardiology, which is considered a gold standard when it comes to high-level, high-number, high-powered studies, Singh said.

The Framingham Heart Study is an example of the power of big data in the cardiology field.

“[Cardiology has] done large-scale trials, large-scale databases. They are able to evaluate drug vs. placebo trials all the time,” he said. “You cannot do that necessarily in orthopedics because you cannot make a surgery or implant placebo-effect or placebo-like [devices], but I think cardiology is probably one of the gold standards” in the area of big data research.

Overall, big data will continue to be important for orthopedic research and is an area that must remain open to discussion and interpretation, according to Spindler.

“Big data will be an inevitable part of the future of orthopedics,” Berry said. “We need to prioritize creating and supporting registries that are purpose-built for collection of big data, so that the data we collect is as relevant as possible to the questions that are of most importance to our patients.” – by Casey Tingle

• References:
• Higgins TF, et al. Orthop Clin North Am. 2010;doi:10.1016/j.ocl.2009.12.006.
• SPORT: The Spine Patient Outcomes Research Trial. Available at: http://www.dartmouth.edu/sport-trial/. Accessed Aug. 4, 2017.

• For more information:
• Daniel J. Berry, MD, can be reached at 200 1st St. SW, Rochester, MN 55902; email: madson.rhoda@mayo.edu.
• Ryan M. Nunley, MD, can be reached at Campus Box 8233, 660 Euclid Ave., Saint Louis, MO 63110; email: williamsdia@wustl.edu.
• Frank J. Schwab, MD, can be reached at 523 E. 72nd St., New York, NY 10021; email: ironsm@hss.edu.
• Kern Singh, MD, can be reached at 1611 W. Harrison St., Chicago, IL 60612; email: kern.singh@rushortho.com.
• Kurt P. Spindler, MD, can be reached at Cleveland Clinic Sports Health Center, Mail Code SH02, 5555 Transportation Blvd., Garfield Heights, OH 44125; email: homrocj2@ccf.org.
• James N. Weinstein, DO, MS, can be reached at The Dartmouth Institute, 1 Medical Center Dr., Lebanon, NH 03756; email: faith.e.johnston@hitchcock.org.

Disclosures: Berry reports he is the chairman of the board of directors for the AJRR; receives royalties from DePuy Synthes; and receives stock from and is a consultant for Bodycad. Schwab reports he is involved in the ISSG. Spindler reports he receives funding from the NIH for the MOON project. Weinstein reports he is the developer of ImagineCare. Nunley and Singh report no relevant financial disclosures.

Click here to read the , “What challenges have you seen or do you foresee in combining, drawing conclusions from large administrative databases?”