Kim, Bobby D. MS*; Edelstein, Adam I. MD†; Hsu, Wellington K. MD†; Lim, Seokchun BS*; Kim, John Y. S. MD‡
Lumbar fusion is a commonly performed procedure for the treatment of numerous symptomatic degenerative conditions of the spine. Arthrodesis is used in a wide range of diagnostic indications including instability, deformity, stenosis, disc pathology, and chronic low back pain. The coevolution of surgical techniques, indications, and instrumentation has led to dramatic increases in the rate of lumbar fusions.1 Despite its ubiquity, overall complication rates of lumbar fusion have been reported to be as high as 13%.2 Discovery of evidence-based associations between risk factors and complication rates continues to be a central focus of research.
One potential source of variation in complication rates that has yet to be fully elucidated is surgeon specialty. The current American health care delivery model is such that patients in need of operative care of the spine may be treated by surgeons who have completed either an orthopedic or neurosurgical training programs. Numerous analyses from various general surgical subspecialty fields have revealed significant differences in the outcomes of surgery when stratified by surgeon training.3–6 However, we are aware of no studies that comprehensively investigate the impact of spine surgeon specialty on 30-day complication rates after lumbar fusion.
The American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) is a prospective, multicenter clinical database that tracks surgical outcomes data from more than 250 private sector hospitals in the United States. At each institution, more than 240 clinical variables are coded from a systematically sampled set of cases. Using this data set, we aimed to assess whether a surgeon specialty of orthopedic surgery (OS) versus neurosurgery (NS) is associated with increased rates of 30-day complication rates after single-level lumbar fusion.
MATERIALS AND METHODS
NSQIP Participant Use Data File
Instituted by the ACS in 2004, NSQIP is a nationwide, validated, prospectively maintained, risk-adjusted surgical outcomes registry that provides a wide range of clinical data for more than 1.7 million patients across the country. All data are tracked for a 30-day window after the index procedure. The details of the database, including case inclusion, exclusion, and systematic sampling strategies have been reported previously.7–9 In brief, trained clinical reviewers at each participating institution collect patient and operative variables using highly standardized definitions. The collected data are deidentified to comply with the NSQIP participant user agreement. This agreement stipulates that no site-specific or surgeon-specific data are included to maintain the privacy of the participating entities. Regular inter-rater reliability audits are performed to ensure data integrity. Recently, the interobserver disagreement rate was reported to be 1.96%.10
Study Population Selection
Using primary Current Procedural Terminology (CPT) codes 22533, 22558, 22612, 22630, and 22633, the 2006–2011 NSQIP Participant Use Data File was retrospectively queried to identify all patients who underwent single-level lumbar fusion procedure. Of note, the data file excludes trauma cases and pediatric patients. Patient cases containing concurrent CPT codes for multilevel lumbar fusion/instrumentation, revision fusion, or pelvic instrumentation were excluded in our analysis. Those cases containing CPT codes for procedures that were unrelated to the lumbar region (e.g., cervical, thoracic) were also identified and excluded. In addition, cases in which the primary team was not NS or OS were excluded. Finally, patients with missing body mass index (BMI), unknown sex, or any comorbidities recorded as “Null” were excluded. The final data set was then stratified into 2 groups by surgeon specialty: NS and OS.
Study Demographics and Outcomes
The data set was stratified by procedural type. Patient demographics, comorbidities, and operative characteristics were tracked to analyze baseline differences between the cohorts and to identify potential confounders. Demographics included age, BMI, race, and sex. Preoperative serum hematocrit was used to define anemia, which was based on the World Health Organization's sex-based definition of anemia.11 Medical comorbidities analyzed included anemia, diabetes, dyspnea, preoperative functional status, history of chronic obstructive pulmonary disease, current pneumonia, congestive heart failure within 30 days of operation, history of myocardial infarction, previous percutaneous coronary intervention and/or previous cardiac surgery, hypertension requiring medication, hemiplegia, paraplegia, history of transient ischemic attack, stroke, chronic steroid use, bleeding disorders, preoperative open wound and/or wound infection, more than 10% loss of body weight 6 months prior to operation, chemotherapy within 30 days of operation, radiotherapy within 90 days of operation, and previous operation within 30 days of the index operation. Alcohol use (>2 drinks/d for the 2 wk preceding the index procedure) and smoking history (within 1 yr of operation) were included as lifestyle variables. Operative details tracked included inpatient status, emergency status, total work relative value unit, wound classification, American Society of Anesthesiologists (ASA) physical status classification, and total operative duration.
Thirty-day postoperative complications were categorized as overall, surgical, and medical complications. Overall complications were defined as incurring any surgical and/or medical complications. Surgical complications included surgical site wound infection by anatomical location (superficial, deep, and/or organ/space) and wound dehiscence. Medical complications were defined as complications related to medical conditions that developed from nonsurgical causes and included: (1) cardiovascular (cardiac arrest, myocardial infarction, or stroke); (2) pulmonary (pneumonia, unplanned reintubation, or ventilator-assisted respiration for >48 hr); (3) coma or peripheral nerve damage; (4) renal (progressive renal insufficiency or acute renal failure); (5) thromboembolic (deep vein thrombosis or pulmonary embolism); (6) septic (sepsis or septic shock); and (7) urinary tract infection. The bleeding transfusion variable underwent a definition change in 2009; to avoid this inconsistency in the data, this variable was excluded from the analysis. Reoperation is defined by NSQIP as a return to the operating room within 30 days of the index procedure for intervention of any kind. Major outcomes of interest in this study included overall, surgical/medical complications, and reoperation.
IBM SPSS Statistics version 20 (IBM Corp., Armonk, NY) was used to perform all descriptive and comparative statistics. We first compared unadjusted variables between the OS and NS groups. Continuous variables were compared using Student t test or the Mann-Whitney U test. Categorical variables were analyzed using the Pearson χ2 test or Fisher exact test where appropriate. We then performed an adjusted analysis. To account for the nonrandom assignment of patients between the 2 cohorts, a propensity score matching analysis was used. Propensity score matching allows for an improved estimate of the treatment effect by balancing observed covariates simultaneously between groups. The details and description of propensity score matching in its applied context are published elsewhere.12–14 Briefly, we created a nonparsimonious logistic regression model through which the propensity scores could be derived. The matching algorithm used for this study was 1:1, nearest neighbor, and computerized greedy matching without replacement. This indicates that 1 OS case was matched to 1 control (NS) case based on nearest propensity scores. If the differences in the propensity scores were not within the designated limit (caliper width), then the algorithm used the next NS case until the match was found. The NS case was not reused once matched. Units outside of common support were discarded for improved balance of covariates. The optimal caliper width was determined by calculating 0.2 of the standard deviation of the logit of the propensity score.15 This procedure yielded 1264 well-matched NS and OS pairs. Model adequacy was validated by comparing the significance and standardized mean difference between the pre- and postmatched sample. A total of 33 variables were balanced, which included all of the aforementioned preoperative variables with the exception of preoperative hematocrit. The measured covariates and outcomes in the matched data set were analyzed using the paired t test or Wilcoxon signed rank test for continuous variables and the McNemar exact test for categorical variables. We then performed 2 independent series of multivariate logistic regression analyses using both the unadjusted and the propensity-matched data set. Candidate variables for each regression model were identified via univariate screening, from which any variables with P < 0.2, and 10 or more occurrences were selected. The candidate variables as well as surgeon specialty and procedure type were included in the regression models to estimate the impact of spine surgeon specialty on outcomes. The NS group served as reference in the regression. For all tests, significance was defined at P < 0.05.
Unadjusted Patient Population
The initial query of the 2006–2011 NSQIP registry by single-level lumbar fusion CPT codes yielded 6952 patients, of which 2970 patients were included for analysis (Figure 1). The NS cohort included 1565 patients, whereas the OS cohort had 1405 patients. A total of 727 patients (24.5%) were instrumented, of which 406 (25.9%) were in the NS group and 321 (22.8%) were in the OS group. The demographic profiles of the cohorts were similar with the exception of race (Table 1). The OS cohort had a lower proportion of Caucasians and a greater proportion of “other/unspecified” race (P < 0.05). Comorbidities were also similar between the groups with the exception of a significantly higher rate of current smoking status in the NS group (P < 0.05). Compared with the OS cohort, the NS cohort had significantly longer total operative duration and a greater proportion of patients assigned to ASA class 3 or higher (P < 0.05). In addition, the NS group had a greater percentage of cases classified as emergent (P = 0.065).
Surgical Procedure Type
TABLE 1-b. Patient C...Image Tools
Posterior or posterolateral and posterior interbody techniques were the 2 most commonly performed procedures in both the unadjusted and the propensity-matched populations (Table 2). In the propensity-matched population, the greatest difference in procedure type was for posterior interbody fusion, comprising 56.2% of the NS group compared with 43.8% of the OS group. The surgical procedure type was further controlled for in our multivariate analysis.
Thirty-Day Outcomes of Unadjusted Population
Overall complication rates of the unadjusted data set were 7.3% and 7.1% for the NS and OS cohort, respectively (P = 0.861) (Table 3). There were a total of 5 deaths (0.3%) in the NS group and 3 deaths (0.2%) in the OS group (P = 0.729). There was no statistically significant difference in any of the remaining outcomes of interest, including medical complications (P = 0.707), surgical complications (P = 0.428), or reoperation (P = 0.748).
Propensity Score–Matched Patient Population
A multivariate logistic regression model was used to calculate propensity scores for all patients from which a matching algorithm created 1264 pairs of well-matched NS and OS cohorts (Table 1). The C statistic of the regression model was 0.61. The mean standard difference of propensity scores before and after matching was 0.41 and 0.04, respectively. The caliper width used for the matching procedure was 0.08. After matching, no variables were statistically significantly different.
Thirty-Day Outcomes of Matched Population
Comparison of 30-day outcomes between the NS and OS cohorts is listed in Table 3. Univariate analyses of matched data revealed no statistical differences in any of the outcomes that were tracked. These include our main outcomes of interest: overall complications (P = 0.693), medical complications (P = 0.786), surgical complications (P = 0.374), and reoperation (P = 0.707). There were a total of 5 deaths (0.4%) in the NS and 3 deaths (0.2%) in the OS group.
Multivariate Analysis of Matched Population
To estimate the impact of spine surgeon specialty as an independent risk factor for postoperative complications, a series of multivariate logistic regression models were employed, which controlled for potential confounding variables (Table 4). Our multivariate analysis revealed that after controlling for 15 potential confounders, surgeon specialty was not a risk factor for overall complications (odds ratio [OR], 0.95; 95% confidence interval [CI], 0.69–1.30; P = 0.740). Of the candidate variables included, age, BMI, dependent functional status, prior stroke, ASA class more than 2, and total operative duration significantly predicted development of any complications (P < 0.05). Male sex was protective of any complication (OR, 0.65; P = 0.012). Similarly, further regression models for each outcome of interest revealed that surgeon specialty was not an independent risk factor for medical complications (OR, 1.11; 95% CI, 0.77–1.60; P = 0.583), surgical complications (OR, 0.76; 95% CI, 0.46–1.26; P = 0.287), or reoperation (OR, 1.10; 95% CI, 0.76–1.60; P = 0.618). The regression models exhibited adequate calibrations and discriminations with Hosmer-Lemeshow statistics of 0.151 to 0.946 and C-indices of 0.695 to 0.717, respectively.
Lumbar fusion has become an increasingly popular option for selected patients seeking relief from numerous degenerative conditions of the lumbar spine. This procedure is associated, however, with a non-negligible risk of perioperative morbidity and mortality. As such, study of the factors that may portend increased risk, and subsequent optimization of those factors, provides an attractive avenue to improve outcomes of surgery. This retrospective population-based study focused on the impact of spine surgeon specialty on 30-day rates of postoperative complications for patients undergoing single-level arthrodesis of the lumbar spine. We find that after controlling for other confounding variables, surgeon specialty carried no association with increased risk for any of the 30-day complications studied.
The pathway to becoming a practicing spine surgeon varies considerably between those who undergo training in the orthopedic and neurosurgical paradigms. These differences include the duration of training, the realm of clinical entities encountered, and the role of postresidency fellowship training. As a result, the clinical decisions made by spine surgeons can vary significantly depending on background.16–20 Accordingly, orthopedic specialty has been associated with a higher rate of fusions than neurosurgical specialty in unadjusted analyses (OR = 12.5).21 Furthermore, in some countries, the spine care can be often dominated by one subspecialty over another. Some of these cultural differences may be attributable to the perceived complication rate and clinical outcomes that are dependent upon training.
The idea that surgeon factors may impact the outcomes of surgery has been a topic of study in the past. Research in the general surgery subspecialty literature has found that surgeons who have completed additional fellowship training have lower unadjusted complication rates than non–fellowship-trained surgeons in a wide array of procedures including major pulmonary, esophageal operations, and carotid endarterectomy.22–25 Literature specific to spine surgery has found that higher surgeon or hospital volume correlates with lower complication rates for a variety of procedures.26–31
Bederman et al21 performed a retrospective database analysis of 6128 patients from a single province in Canada who underwent surgery for degenerative disease of the lumbar spine. Included in their analysis was a comparison of the probability of reoperation-free survival at up to 10 years after surgery stratified by surgeon specialty. They found no significant differences between orthopedic and neurosurgical training cohorts.
Our study adds to the existing literature by comparing a wide array of complication rates from a nationwide database for a single procedure. This approach gives us an excellent avenue to detect a difference in perioperative complication rates, if one were to exist. The overall complication rates in our unadjusted data was 7.2%, which is lower than previously reported 13% in administrative data from 1993–2002.2 Our regression analysis of propensity-matched cohorts further finds that no significant difference could be found for any of the complications studied. This result lends support to the current model of surgical care provision for patients with lumbar spine pathology requiring arthrodesis.
In addition to spine surgeon specialty, we found some of the covariates included in the regression model to significantly predict overall complication. These include age, BMI, dependent functional status, prior stroke, ASA class greater than 2, and total operative duration. With the exception of dependent functional status, these factors have been implicated as adverse prognostic indicators in spine surgery.32–35 In particular, our recent analysis revealed that increasing operative duration independently predicted a wide array of complications in patients undergoing single-level lumbar fusion.36 Preoperative functional health status had been shown to be adversely associated with general surgery outcomes.37
This study is not without several limitations. The NSQIP database is not designed to capture procedure-specific or specialty-specific outcome variables, and thus factors such as quality of life measures or radiographical parameters are not available for comparison. The 30-day duration of follow-up prevents us from tracking any complications or reoperations that may occur outside of this timeframe. Our risk-adjusted models are limited to the variables tracked in the database and the potential for residual confounding cannot be fully eliminated. The database also does not capture surgeon- or hospital-level data for inclusion in the regression. Because of the database coding protocol, we were unable to detect or control for the possible involvement of multiple surgical specialties in a given case. Information about referral patterns is unavailable and thus a selection bias may be introduced to the analysis.
This is the first population-based study that comprehensively investigates the effect of surgeon specialty in the realm of spinal surgery. Our analysis reveals that surgeon specialty is not a risk factor for any of the studied 30-day complications in patients undergoing single-level lumbar fusion. This helps to support the currently dichotomous paradigm of training for spine surgeons. Further research is warranted to validate this relationship in other spine procedures and for other outcomes.
Deidentified patient information is freely available to all institutional members who comply with the ACS-NSQIP Data Use Agreement. The Data Use Agreement implements the protections afforded by the Health Insurance Portability and Accountability Act of 1996 and the ACS-NSQIP Hospital Participation Agreement.
The ACS-NSQIP and the hospitals participating in the ACS-NSQIP are the source of the data used herein; they have not verified and are not responsible for the statistical validity of the data analysis or the conclusions derived by the authors.
* It is currently unknown whether spine surgeon specialty has an impact on 30-day complication rates in patients undergoing single-level lumbar fusion.
* To examine whether such relationship exists, we reviewed 2960 surgical cases using the NSQIP database while using propensity score matching technique to improve the estimate of the effect of surgeon specialty on outcomes.
* In multivariate analysis that controlled for numerous potential confounders, spine surgeon specialty was not a factor for increased risk of overall complication, medical complications, surgical complications, or reoperation.
* In summary, the spine surgeon specialty is not a risk factor for any of the reported 30-day complication rates in patients undergoing single-level lumbar fusion.
1. Deyo RA, Gray DT, Kreuter W, et al. United States trends in lumbar fusion surgery for degenerative conditions. Spine 2005;30:1441–5.
2. Kalanithi PS, Patil CG, Boakye M. National complication rates and disposition after posterior lumbar fusion for acquired spondylolisthesis. Spine 2009;34:1963–9.
3. Feasby TE, Quan H, Ghali WA. Hospital and surgeon determinants of carotid endarterectomy outcomes. Arch Neurol 2002;59:1877–81.
4. Pearce WH, Parker MA, Feinglass J, et al. The importance of surgeon volume and training in outcomes for vascular surgical procedures. J Vasc Surg 1999;29:768–76.
5. Goodney PP, Lucas FL, Stukel TA, et al. Surgeon specialty and operative mortality with lung resection. Ann Surg 2005;241:179–84.
6. Kingsmore D, Hole D, Gillis C. Why does specialist treatment of breast cancer improve survival? The role of surgical management. Br J Cancer 2004;90:1920–5.
7. Khuri SF, Daley J, Henderson W, et al. Risk adjustment of the postoperative mortality rate for the comparative assessment of the quality of surgical care: results of the National Veterans Affairs Surgical Risk Study. J Am Coll Surg 1997;185:315–27.
8. Khuri SF, Henderson WG, Daley J, et al. Successful implementation of the department of Veteran Affair's National Surgical Quality Improvement Program in the private sector: The Patient Safety In Surgery Study. Ann Surg 2008;248:329–36.
9. Birkmeyer JK, Shahian DM, Dimick JB, et al. Blueprint for a new American College of Surgeons: National Surgical Quality Improvement Program. J Am Coll Surg 2008;5:777–82.
10. Shiloach M, Frencher SK Jr, Steeger JE, et al. Toward robust information: data quality and inter-rater reliability in the American College of Surgeons National Surgical Quality Improvement Program. J Am Coll Surg 2010;210:6–16.
11. McLean E, Cogswell M, Egli I, et al. Worldwide prevalence of anemia, WHO Vitamin and Mineral Nutrition Information System, 1993–2005. Public Health Nutr 2009;12:444–45.
12. Rosenbaum PR, Rubin DB. Constructing a control group using multivariate-matched sampling methods that incorporate the propensity score. Am Stat 1985;39:33–38.
13. Thoemmes F, Kim ES. A systematic review of propensity score methods in the social sciences. Multivariate Behav Res 2011;46:90–118.
14. Stukel TA, Fisher ES, Wennberg DE, et al. Analysis of observational studies in the presence of treatment selection bias: effects of invasive cardiac management on AMI survival using propensity score and instrumental variable methods. JAMA 2007;297:278–85.
15. Austin PC. Optimal caliper widths for propensity-score matching when estimating differences in means and differences in proportions in observational studies. Pharm Stat 2011;10:150–61.
16. Irwin ZN, Hilibrand AS, Gustavel M, et al. Variation in surgical decision making for degenerative spinal disorders. Part I: lumbar spine. Spine 2005;30:2208–13.
17. Glaser JA, Jaworski BA, Cuddy BG, et al. Variation in surgical opinion regarding management of selected cervical spine injuries. Spine 1998;23:975–83.
18. Morganti C, Sweeny CA, Albanese SA, et al. Return to play after cervical spinal injury. Spine 2001;26:1131–6.
19. Grauer JN, Vaccaro AR, Beiner JM, et al. Similarities and differences in the treatment of spine trauma between surgical specialties and location of practice. Spine 2004;29:685–96.
20. Rampersaud YR, Fisher C, Wilsey J, et al. Agreement between orthopedic surgeons and neurosurgeons regarding a new algorithm for the treatment of thoracolumbar injuries: a multicenter reliability study. J Spinal Disord Tech 2006;19:477–82.
21. Bederman SS, Kreder HJ, Weller I, et al. The who, what, and when of surgery for the degenerative lumbar spine: a population-based study of surgeon factors, surgical procedures, recent trends and reoperation rates. Can J Surg 2009;52:283–90.
22. Ferraris VA, Saha SP, Davenport DL, et al. Thoracic surgery in the real world: does surgical specialty affect outcomes in patients having general thoracic operations? Ann Thorac Surg 2012;93:1041–7.
23. Schipper PH, Diggs BS, Ungerleider RM, et al. The influence of surgeon specialty on outcomes in general thoracic surgery: a national sample 1996 to 2005. Ann Thorac Surg 2009;88:1566–73.
24. Hollenbeak CS, Bowman AR, Harbaugh RE, et al. The impact of surgical specialty on outcomes for carotid endarterectomy. J Surg Res 2010;159:595–602.
25. Dimick JB, Goodney PP, Orringer MB, et al. Specialty training and mortality after esophageal cancer resection. Ann Thorac Surg 2005;80:282–6.
26. Taylor HD, Dennis DA, Crane HS. Relationship between mortality rates and hospital patient volume for Medicare patients undergoing major orthopedic surgery of the hip, knee, spine, and femur. J Arthroplasty 1997;12:235–42.
27. Vitale MA, Arons RR, Hyman JE, et al. The contribution of hospital volume, payer status, and other factors on the surgical outcomes of scoliosis patients: a review of 3606 cases in the state of CA. J Pediatr Orthop 2005;25:393–9.
28. Dasenbrock HH, Clarke MJ, Witham TF, et al. The impact of provider volume on the outcomes after surgery for lumbar spinal stenosis. Neurosurgery 2012;70:1346–53.
29. Farjoodi P, Skolasky RL, Riley LH. The effects of hospital and surgeon volume on postoperative complications after lumbar spine surgery. Spine 2011;36:2069–75.
30. Martin BI, Mirza SK, Franklin GM, et al. Hospital and surgeon variation in complications and repeat surgery following incident lumbar fusion for common degenerative diagnoses. Health Serv Res 2013;48:1–25.
31. Martin BI, Mirza SK, Flum DR, et al. Repeat surgery after lumbar decompression for herniated disc: the quality implications of hospital and surgeon variation. Spine J 2012;12:89–97.
32. Fineberg SJ, Kurd MF, Patel AA, et al. Incidence and risk factors for gastrointestinal hemorrhage after lumbar fusion. Spine 2013;38:1584–9.
33. Bekelis K, Desai A, Bakhoum SF, et al. A predictive model of complications after spine surgery: the National Surgical Quality Improvement Program (NSQIP) 2005–2010. [published online ahead of print October 4, 2013] Spine 2013.
34. Gruskay JA, Fu M, Bohl DD, et al. Factors affecting length of stay following elective posterior lumbar spine surgery: A multivariate analysis. [published online ahead of print October 31, 2013] Spine J 2013.
35. Proietti L, Scaramuzzo L, Schiro GR, et al. Complications in lumbar spine surgery: a retrospective analysis. Indian J Orthop 2013;47:340–5.
36. Kim BD, Hsu WK, Saha S, et al. Operative duration as an independent risk factor for postoperative complications in single-level lumbar fusion: a review of 4588 surgical cases. Spine 2014;39:510–20.
37. Albright EL, Davenport DL, Roth JS. Preoperative functional health status impacts outcomes after ventral hernia repair. Am Surg 2012;78:230–4.
single-level lumbar fusion; arthrodesis; spine surgeon specialty; postoperative complications; overall complications; medical complications; surgical complications; reoperation; ACS-NSQIP; propensity score matching