Demographic, Clinical, and Operative Factors Affecting Long-Term Revision Rates After Cervical Spine Arthrodesis

Derman, Peter B. MD, MBA; Lampe, Lukas P. MD; Hughes, Alexander P. MD; Pan, Ting Jung MPH; Kueper, Janina BS; Girardi, Federico P. MD; Albert, Todd J. MD; Lyman, Stephen PhD

Journal of Bone & Joint Surgery - American Volume:
doi: 10.2106/JBJS.15.00938
Scientific Articles
Abstract

Background: Limited data exist on long-term revision rates following cervical spine arthrodeses. The purposes of this study were to define reoperation rates after primary cervical arthrodeses and to identify risk factors for revisions.

Methods: New York State’s all-payer health-care database was queried to identify all primary subaxial cervical arthrodeses occurring in the 16 years from 1997 through 2012. A total of 87,042 patients were included in the study cohort. Demographic information was extracted. Patients’ preoperative medical comorbidities, surgical indications, and operative approaches were assembled using codes from the ICD-9-CM (International Classification of Diseases, Ninth Revision, Clinical Modification). The cohort was followed to revision surgical procedures, death, or the end of the study period. All subsequent contiguous spinal arthrodeses, including in the subaxial cervical spine, were considered revisions. The overall revision risk and the risk associated with various preoperative characteristics, surgical indications, and operative approaches were assessed using a Cox proportional hazard model.

Results: During the study period, 6,721 patients (7.7%) underwent revision. The median time to revision was 24.5 months. The probability of undergoing at least one revision by 192 months was 12.6%. Arthrodeses performed via anterior-only approaches had a significantly higher probability of revision (p < 0.001) at 13.4% (95% confidence interval [95% CI], 12.9% to 13.9%) than those performed via posterior approaches at 7.4% (95% CI, 6.6% to 8.4%) or circumferential (anterior and posterior) approaches at 5.2% (95% CI, 4.0% to 6.8%). This relationship persisted in multivariate analysis; compared with anterior surgical procedures, there was a significantly lower risk of revision (p < 0.001) for posterior surgical approaches at a hazard ratio of 0.76 (95% CI, 0.69 to 0.84) and circumferential approaches at a hazard ratio of 0.53 (95% CI, 0.42 to 0.66). Patient age of 18 to 34 years, white race, insurance status of Workers’ Compensation or Medicare, and surgical procedures for spinal stenosis, spondylosis, deformity, and neoplasm were associated with elevated revision risk. Arthrodeses spanning few levels and those performed for fractures had a lower revision risk.

Conclusions: Primary subaxial cervical spine arthrodeses had a probability of revision approaching 13% over a 16-year period, with elevated reoperation rates in patients undergoing anterior-only surgical procedures. Age, race, insurance status, surgical indication, and number of spinal levels included in the arthrodesis were also associated with reoperation risk.

Level of Evidence: Therapeutic Level III. See Instructions for Authors for a complete description of levels of evidence.

Author Information

1Hospital for Special Surgery, New York, NY

2Charité-Universitätmedizin, Berlin, Germany

E-mail address for P.B. Derman: dermanp@hss.edu

Article Outline

Cervical spine arthrodesis is utilized in the treatment of numerous conditions, including spondylosis, trauma, tumors, and deformity1-4. The number of primary and revision cervical arthrodeses has increased in the United States over time, outpacing population growth5. Reported revision rates vary widely and may depend on demographic, clinical, and operative factors such as patient age, sex, comorbidities, and surgical approach5-8. These revision surgical procedures can be technically demanding, costly, and less likely to produce optimal functional outcomes9-11.

The literature of which we are aware contains a limited number of large, database-driven studies documenting long-term reoperation rates after cervical spine arthrodesis8,9,12. These investigations are restricted to a limited number of diagnoses or surgical approaches. The purposes of this study were to better define the revision rate after primary subaxial cervical spine arthrodesis and to identify risk factors for these reoperations. This was accomplished by conducting a review of a New York State database from 1997 to 2012. This data source is valuable in that it contains nearly complete, longitudinal data on a large U.S. patient population.

Back to Top | Article Outline

Materials and Methods

Established in 1979, the New York Department of Health’s Statewide Planning and Research Cooperative System (SPARCS) database is an all-payer data-reporting system that collects patient-level information on all inpatient admissions and surgical procedures in non-federal, acute-care hospitals in the State of New York13. SPARCS annual reports reveal data completeness levels of 95% to 100% from 1997 to 2011 with a median completeness of 99%14. At the time of this writing, the completeness report for 2012 had not yet been published. Inpatient data in SPARCS are categorized by ICD-9-CM (International Classification of Diseases, Ninth Revision, Clinical Modification) diagnosis and procedural codes.

All adult patients who underwent isolated primary subaxial cervical arthrodesis in the 16 years spanning from 1997 to 2012 were identified in the SPARCS database using ICD-9-CM procedure codes. The ICD-9-CM codes used to define the cohort and to identify their surgical indications, operative approaches, and revision surgical procedures are outlined in the Appendix. Patients undergoing concurrent lumbar or atlantoaxial arthrodesis were excluded when defining the initial cohort, as were those whose spinal arthrodesis levels were not specified. Cervical arthrodeses extending into the thoracic region were included so as not to eliminate cervical arthrodeses with some distal extension. As such, patients with subaxial cervical spine fusion procedure codes and concurrent thoracic fusion procedure codes were not excluded. This represented 1.1% of the study cohort. Nonresidents of New York and those for whom patient identifiers were unavailable were also excluded because of impaired ability to follow them over time. Diagnosis codes for patients meeting the above criteria were reviewed to further screen out revision cases from the index cohort; patients with a diagnosis of mechanical complication of an orthopaedic implant on admission were assumed to be undergoing revision procedures that were miscoded as primary surgical procedures. These patients were therefore excluded.

Patient demographic characteristics (age, sex, race or ethnicity, and insurance status) and comorbidities (Elixhauser comorbidity measures) were extracted. The Elixhauser measures are a collection of 30 comorbid conditions that are associated with inferior hospital outcomes; these can be identified in large administrative inpatient databases using defined ICD-9-CM diagnosis codes15. Four of these measures were not used: metastatic cancer (highly collinear with a metastasis surgical indication), chronic blood loss anemia (likely to occur because of the surgical procedure, so less informative), acquired immunodeficiency syndrome (AIDS) (excluded from the cohort), and complicated hypertension (which was combined with uncomplicated hypertension to form one index). Surgical indications and approach were also determined via ICD-9-CM diagnosis codes (see Appendix).

The study cohort was followed forward to the earliest of the following: revision surgical procedures, death for any reason, or follow-up through 2012. All SPARCS records were searched for death (i.e., using inpatient, emergency department, and ambulatory surgery center databases). Those patients without a record of death were censored at their predicted time of death on the basis of life tables published by the Centers for Disease Control and Prevention (CDC)16. After the index procedure, all contiguous spinal arthrodeses including the subaxial cervical spine were considered revision surgical procedures. For instance, a patient who had undergone a prior subaxial cervical spine arthrodesis who subsequently underwent concurrent arthrodesis of the atlantoaxial, cervical, thoracic, and lumbar spine would have been included in the revision group, but a patient who subsequently underwent isolated lumbar arthrodesis would not have been included. This investigation focused on the time to revision after primary cervical arthrodesis; revisions after the first revision were not included in the analysis.

On the basis of this methodology, 94,780 patients underwent primary subaxial cervical arthrodeses performed in New York from 1997 to 2012. Of these patients, 7,177 were not New York residents, 558 did not have an available patient identifier, and 3 were miscoded with death records dated prior to a cervical arthrodesis, leaving 87,042 for the final analysis (Table I). The median follow-up was 70 months, with an interquartile range of 28 to 120 months. Of these patients, 6,721 (7.7%) subsequently underwent revision during the study period.

A Cox multivariate proportional hazard model was used to calculate overall revision risk as well as risk associated with various preoperative characteristics, surgical indications, and operative approaches. Hazard ratios were adjusted for age, sex, race, insurance status, surgical indication, approach, year of index surgical procedure, and Elixhauser comorbidities. A secondary analysis by the number of levels fused and the interaction term between this variable and surgical approach was performed. However, this subanalysis spans only 2004 through 2012 (encompassing 56,146 patients) because ICD-9-CM codes specifying the number of levels fused were first introduced in October 2003. Hazard ratios in the subgroup analysis were adjusted for the same variables as the main analysis with the addition of the number of levels fused. The threshold for significance was set at p < 0.05. Hazard ratios and 95% confidence intervals (95% CIs) were also reported.

Back to Top | Article Outline

Results

The median time to revision was 24.5 months overall (25.9 months for the anterior approach, 14.1 months for the posterior approach, and 12.5 months for the circumferential approach). The majority of 1 or 2-level arthrodeses were performed via an anterior approach (90.9%), most arthrodeses involving ≥8 levels were performed via a posterior approach (73.9%) or a circumferential approach (13.7%), and arthrodeses spanning 3 to 7 levels were more evenly split between approaches (Table II). Overall, the probability of revision at a follow-up of 192 months (16 years) was 12.6% (95% CI, 12.2% to 13.1%) (Fig. 1).

At the time of the original procedure, there were significantly lower hazard ratios for revision (p < 0.001 for all) for patients who were 50 to 59 years of age, 60 to 69 years of age, or ≥70 years of age compared with patients who were 18 to 34 years, but patients who were 35 to 49 years of age did not differ significantly compared with patients who were 18 to 34 years of age (p = 0.611) (Table III). Sex was not found to be associated with revision risk. White patients were significantly more likely to undergo revision than Hispanic and Asian patients (p < 0.001 for both) but did not differ significantly from black patients (p = 0.709). Surgical procedures covered by Workers’ Compensation or Medicare had significantly higher revision risk than those paid for by private insurers (p < 0.001). The revision risk of operations reimbursed by Medicaid did not differ from those covered by private insurance. The hazard ratio was significantly higher in patients undergoing arthrodesis for spinal stenosis (1.12 [95% CI, 1.04 to 1.21]; p = 0.004), spondylosis (1.12 [95% CI, 1.06 to 1.20]; p < 0.001), deformity (1.32 [95% CI, 1.11 to 1.58]; p = 0.002), and neoplasm (1.61 [95% CI, 1.21 to 2.14]; p = 0.001); and the hazard ratio was lower for patients undergoing arthrodesis for fracture (0.69 [95% CI, 0.58 to 0.83]; p < 0.001) (Table III).

At the 192-month follow-up, patients undergoing arthrodesis performed via an anterior approach had a significantly higher probability of revision (p < 0.001) at 13.4% (95% CI, 12.9% to 13.9%) than those undergoing arthrodesis via the posterior approach at 7.4% (95% CI, 6.6% to 8.4%) or those undergoing arthrodesis via the circumferential approach at 5.2% (95% CI, 4.0% to 6.8%) (Fig. 2). Compared with an anterior surgical procedure, this translated to a significantly lower risk of revision (p < 0.001) for posterior approaches, with a hazard ratio of 0.76 (95% CI, 0.69 to 0.84), and for circumferential approaches, with a hazard ratio of 0.53 (95% CI, 0.42 to 0.66). These general relationships were maintained in subanalyses adjusting for the number of levels fused and the interaction term between approach and the number of levels fused (Table IV). Comparison of short arthrodeses (1 or 2 levels) with longer arthrodeses (≥3 levels) revealed that the longer arthrodeses exhibited a significantly higher risk of revision overall (p < 0.001), with a hazard ratio of 1.18 (95% CI, 1.07 to 1.30). However, among approaches, revision risk was significantly higher (p < 0.001) for longer arthrodesis lengths only in the setting of anterior surgical procedures, with a hazard ratio of 1.25 (95% CI, 1.13 to 1.39).

Back to Top | Article Outline

Discussion

These analyses indicate that primary subaxial cervical arthrodesis has a probability of revision approaching 13% at 16 years after the initial procedure. Multivariate analysis identified a number of demographic, clinical, and surgical factors that were independently associated with revision risk. For instance, younger patient age, Workers’ Compensation insurance status, and anterior approaches were associated with increased revision risk, but shorter arthrodeses and arthrodeses for a diagnosis of fracture had relatively lower revision risk.

The finding that longer arthrodeses, especially when performed via an anterior approach, are associated with higher risk of reoperation falls in line with current clinical teaching, which typically favors posterior or circumferential approaches in the setting of arthrodeses spanning ≥3 levels17. Elevated revision rates in the setting of longer arthrodeses may be due to increased risk of pseudarthrosis and/or increased biomechanical demands placed on the adjacent segments with resulting degeneration18-20. Expansive anterior arthrodeses may be particularly prone to failure because of the risk of dislodgement or migration of long strut grafts or mechanical failure of fixation devices21,22. However, the finding that anterior approaches for 1 and 2-level arthrodeses were associated with an elevated long-term rate of reoperation, even after controlling for surgical indication, age, comorbidities, and a variety of other factors, was unanticipated and warrants future investigation.

Prior studies have focused on the effect of surgical approach on the subsequent revision risk after cervical spine surgical procedures. A 60-month longitudinal investigation of the Washington State inpatient database published in 2009 suggested that nearly 10% of patients who had a surgical procedure for degenerative cervical spine disease underwent reoperation by 60 months from the index procedure, similar to that observed in the New York population9. A more recent analysis of the MarketScan database revealed that patients undergoing arthrodesis for multilevel degenerative cervical disease were less likely to undergo reoperation if the surgical procedure was performed from an anterior approach rather than from a posterior approach8. However, the mean follow-up for that cohort was 18.6 months, with a median time to revision of 20.7 months. These shorter-term findings mirror those observed in the SPARCS database, where there was an early trend toward fewer revisions in the anterior group, but the cumulative revision rate for anterior approaches began to exceed that of posterior approaches at longer follow-up. Lee et al. reported specifically on revision for adjacent-segment disease and found that arthrodeses performed from a posterior approach had a 3 times greater risk of revision compared with those performed from an anterior approach23. Because the mean follow-up was 48 months (range, 12 to 168 months), it was more difficult to reconcile these findings with our own, although limiting their end point to revision for adjacent-segment disease may have contributed to this difference. Meanwhile, a meta-analysis comparing anterior and posterior approaches for the treatment of multilevel cervical spondylotic myelopathy found a higher revision rate in the anterior group, but reoperation figures were founded solely on studies comparing anterior arthrodeses with laminoplasty in the absence of arthrodesis6.

There is discordance within the literature on the survivorship of single-level cervical spine arthrodeses. Veeravagu et al. observed a 10.7% rate of revision at 24 months for anterior cervical discectomy and fusion12, Delamarter and Zigler found a 14.5% revision rate at 60 months24, and Singh et al. reported a reoperation rate of only 7.6% at 144 months25. This variation may be due in part to differences in surgeon preferences and patient factors (for example, demographic characteristics, comorbidities, and attitudes) that impact the decision on whether or not to perform revision surgical procedures. As such, it is difficult to define a consistent figure for the long-term revision rate of cervical spine arthrodesis on the basis of the prior literature.

This study has several characteristics that distinguish it from previous publications. It is based on a large database that does not rely on population sampling, unlike many of the nationwide databases26. The MarketScan database, for instance, is a non-random convenience sample, which is subject to sampling bias27. Instead, SPARCS contains nearly 100% of the procedures performed annually in a large U.S. population. SPARCS has the added benefit of including unique patient identifiers that permit investigators to track patients over time; this is not the case for the National Inpatient Sample (NIS) or the National Surgical Quality Improvement Program (NSQIP), two widely utilized nationwide databases26,28,29. As mentioned earlier, Washington State data were used in the past by King et al. to investigate similar subject matter9. The study presented herein differs from and improves upon that investigation in a number of key aspects. First, New York is one of the most populous states in the country, with a diverse population exceeding 2.5 times the size of the State of Washington30. Furthermore, analysis of 16 years of New York data (1997 to 2012) compared with 5 years of Washington data (1998 to 2002) allowed for substantially longer longitudinal tracking of the New York cohort and inclusion of more recent procedures. King et al. examined cervical spine operations for degenerative disease only; patients with diagnoses of infection, inflammatory spondyloarthropathy, malignancy, and trauma were excluded. Incorporating these diagnoses into the New York investigation allowed for a more comprehensive analysis of surgery for cervical spine disease. Finally, the use of life tables to censor New York patients at their predicted time of death is an improvement upon the methodology utilized in the prior study.

A number of limitations were present in this study, many of which are inherent to population-based database research, where the rigor of the final analysis is dependent on the quality of the initial administrative coding and database methodology. ICD-9-CM codes lack the granularity necessary to make some clinical distinctions (for example, arthrodesis technique and precise levels involved) that would have been interesting to consider had it been possible. In the future, studies employing ICD-10 (International Statistical Classification of Diseases and Related Health Problems, Tenth Revision) data may be capable of better addressing these questions and others. Reliance on administrative data also necessitated the use of revision arthrodesis as a proxy for surgical durability. However, in reality, the decision on whether or not to perform these revision operations is affected by factors such as radiographic findings, symptom severity, patient preferences, and surgeon discretion, which could not be captured in the available data. It would have also been useful to have been able to adjust for radiographic findings (for example, kyphosis) prior to the index surgical procedure, as this can affect the choice of surgical approach. As a state database, SPARCS only contains information on events taking place within medical institutions in New York; any deaths or revision procedures occurring elsewhere are not captured. The study population was limited to New York residents (ostensibly less likely to seek care outside the state border) and life spans were adjusted using life tables to help to address this issue; however, some residual bias toward underestimating revision probabilities may remain. Furthermore, unique patient identifiers for individuals with a diagnosis of AIDS or a previously terminated pregnancy are not released in the SPARCS data set to protect patient privacy. As such, 553 patients with AIDS and 5 with a history of pregnancy termination had to be excluded from our analysis. These individuals would have represented <0.6% of the study cohort, but their absence did limit the generalizability of our results toward patients with these two characteristics. Finally, this study considered a relatively large number of variables and therefore runs the risk of over-fitting. However, the large sample size of the cohort was sufficient to provide robust estimates for the purpose of exploratory analysis.

In conclusion, this study demonstrates that primary subaxial cervical spine arthrodesis has a probability of revision approaching 13% over a period of 16 years, with elevated reoperation rates in patients undergoing an anterior-only surgical procedure. Patient age, race, insurance status, surgical indication, and arthrodesis length were associated with reoperation risk. Knowledge of these factors should help to promote exploration of strategies to reduce the prevalence of revision surgical procedures and to facilitate more accurate preoperative counseling of patients and their relatives.

Back to Top | Article Outline

Appendix Cited Here...

Tables showing the ICD-9-CM codes utilized to identify the study cohort, operative approaches, revision surgical procedures, and surgical indications are available with the online version of this article as a data supplement at jbjs.org.

Investigation performed at the Hospital for Special Surgery, New York, NY

Disclosure: This study was conducted without an external source of funding. On the Disclosure of Potential Conflicts of Interest forms, which are provided with the online version of the article, one or more of the authors checked “yes” to indicate that the author had a relevant financial relationship in the biomedical arena outside the submitted work.

Back to Top | Article Outline

References

1. Kwon BK, Vaccaro AR, Grauer JN, Beiner JM. The use of rigid internal fixation in the surgical management of cervical spondylosis. Neurosurgery. 2007 ;60(1)(Supp1 1):S118–29.
2. Kwon BK, Vaccaro AR, Grauer JN, Fisher CG, Dvorak MF. Subaxial cervical spine trauma. J Am Acad Orthop Surg. 2006 ;14(2):78–89.
3. Mesfin A, Buchowski JM, Gokaslan ZL, Bird JE. Management of metastatic cervical spine tumors. J Am Acad Orthop Surg. 2015 ;23(1):38–46.
4. Hann S, Chalouhi N, Madineni R, Vaccaro AR, Albert TJ, Harrop J, Heller JE. An algorithmic strategy for selecting a surgical approach in cervical deformity correction. Neurosurg Focus. 2014 ;36(5):E5.
5. Rajaee SS, Kanim LEA, Bae HW. National trends in revision spinal fusion in the USA: patient characteristics and complications. Bone Joint J. 2014 ;96-B(6):807–16.
6. Zhu B, Xu Y, Liu X, Liu Z, Dang G. Anterior approach versus posterior approach for the treatment of multilevel cervical spondylotic myelopathy: a systemic review and meta-analysis. Eur Spine J. 2013 ;22(7):1583–93. Epub 2013 May 9.
7. Liu X, Wang H, Zhou Z, Jin A. Anterior decompression and fusion versus posterior laminoplasty for multilevel cervical compressive myelopathy. Orthopedics. 2014 ;37(2):e117–22.
8. Cole T, Veeravagu A, Zhang M, Azad TD, Desai A, Ratliff JK. Anterior versus posterior approach for multilevel degenerative cervical disease: a retrospective propensity score-matched study of the MarketScan database. Spine (Phila Pa 1976). 2015 ;40(13):1033–8.
9. King JT Jr, Abbed KM, Gould GC, Benzel EC, Ghogawala Z. Cervical spine reoperation rates and hospital resource utilization after initial surgery for degenerative cervical spine disease in 12,338 patients in Washington State. Neurosurgery. 2009 ;65(6):1011–22; discussion 1022-3.
10. Tian W, Yan K, Han X, Yu J, Jin P, Han X. Comparison of the clinical and radiographic results between cervical artificial disc replacement and anterior cervical fusion: a six-year prospective non-randomized comparative study. J Spinal Disord Tech. 2014 . [Epub ahead of print.]
11. Liu HW, Chen L, Xu NW, Yang HL, Gu Y. Outcomes of secondary laminoplasty for patients with unsatisfactory results after anterior multilevel cervical surgery. J Korean Neurosurg Soc. 2015 ;57(1):36–41. Epub 2015 Jan 31.
12. Veeravagu A, Cole T, Jiang B, Ratliff JK. Revision rates and complication incidence in single- and multilevel anterior cervical discectomy and fusion procedures: an administrative database study. Spine J. 2014 ;14(7):1125–31. Epub 2013 Oct 11.
13. New York State Department of Health. Statewide Planning and Research Cooperative System (SPARCS): overview. 2015 Feb. https://www.health.ny.gov/statistics/sparcs/. Accessed 2016 May 27.
14. New York State Department of Health. Statewide Planning and Research Cooperative System (SPARCS): annual reports. 2013 Jul. https://www.health.ny.gov/statistics/sparcs/annual/ip/index.htm. Accessed 2016 May 27.
15. Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care. 1998 ;36(1):8–27.
16. Centers for Disease Control and Prevention. Life tables. 2015 . http://www.cdc.gov/nchs/products/life_tables.htm. Accessed 2016 May 27.
17. Emery SE. Anterior approaches for cervical spondylotic myelopathy: which? When? How? Eur Spine J. 2015 ;24(Suppl 2):150–9. Epub 2015 Feb 5.
18. Shen HX, Buchowski JM, Yeom JS, Liu G, Lin N, Riew KD. Pseudarthrosis in multilevel anterior cervical fusion with rhBMP-2 and allograft: analysis of one hundred twenty-seven cases with minimum two-year follow-up. Spine (Phila Pa 1976). 2010 ;35(7):747–53.
19. Fraser JF, Härtl R. Anterior approaches to fusion of the cervical spine: a metaanalysis of fusion rates. J Neurosurg Spine. 2007 ;6(4):298–303.
20. Dang ABC, Hu SS, Tay BKB. Biomechanics of the anterior longitudinal ligament during 8 g whiplash simulation following single- and contiguous two-level fusion: a finite element study. Spine (Phila Pa 1976). 2008 ;33(6):607–11.
21. Vaccaro AR, Falatyn SP, Scuderi GJ, Eismont FJ, McGuire RA, Singh K, Garfin SR. Early failure of long segment anterior cervical plate fixation. J Spinal Disord. 1998 ;11(5):410–5.
22. Sasso RC, Ruggiero RA Jr, Reilly TM, Hall PV. Early reconstruction failures after multilevel cervical corpectomy. Spine (Phila Pa 1976). 2003 ;28(2):140–2.
23. Lee JC, Lee SH, Peters C, Riew KD. Risk-factor analysis of adjacent-segment pathology requiring surgery following anterior, posterior, fusion, and nonfusion cervical spine operations: survivorship analysis of 1358 patients. J Bone Joint Surg Am. 2014 ;96(21):1761–7.
24. Delamarter RB, Zigler J. Five-year reoperation rates, cervical total disc replacement versus fusion, results of a prospective randomized clinical trial. Spine (Phila Pa 1976). 2013 ;38(9):711–7.
25. Singh K, Phillips FM, Park DK, Pelton MA, An HS, Goldberg EJ. Factors affecting reoperations after anterior cervical discectomy and fusion within and outside of a Federal Drug Administration investigational device exemption cervical disc replacement trial. Spine J. 2012 ;12(5):372–8. Epub 2012 Mar 16.
26. Murphy M, Alavi K, Maykel J. Working with existing databases. Clin Colon Rectal Surg. 2013 ;26(1):5–11.
27. Hansen LG, Chang S. Health research data for the real world: the MarketScan databases. Truven Health Analytics. 2011 Jul. http://truvenhealth.com/portals/0/assets/PH_11238_0612_TEMP_MarketScan_WP_FINAL.pdf. Accessed 2016 May 27.
28. Agency for Healthcare Research and Quality Healthcare Cost and Utilization Project (HCUP). Overview of the National (Nationwide) Inpatient Sample (NIS). 2016 Feb. https://www.hcup-us.ahrq.gov/nisoverview.jsp. Accessed 2016 May 27.
29. American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP). About ACS NSQIP. https://www.facs.org/quality-programs/acs-nsqip/about. Accessed 2016 May 27.
30. U.S. Census Bureau, Data Integration Division. Annual estimates of the resident population for the United States, regions, states, and Puerto Rico: April 1, 2010 to July 1, 2015. https://www.census.gov/popest/data/state/totals/2015/index.html. Accessed 2016 May 27.
Copyright 2016 by The Journal of Bone and Joint Surgery, Incorporated