When the Charlson comorbidity score was used to predict complication occurrence, the AUC was 0.609 (95% CI = 0.604 to 0.614) for the anterior cervical cohort, 0.616 (95% CI = 0.607 to 0.625) for the posterior cervical cohort, 0.614 (95% CI = 0.606 to 0.622) for the anterior thoracolumbar cohort, and 0.604 (95% CI = 0.601 to 0.608) for the posterior thoracolumbar cohort. The AUC for the entire cohort was 0.607 (95% CI = 0.604 to 0.610). All of the AUCs from the Charlson score were significantly less than their counterparts based on logistic regression (p < 0.01).
For predicting the specific type of complication that occurred, the analysis focused on the 44,507 occurrences of complications in the patient cohort. Predictive accuracy was poor for identifying which type of complication would occur.
Initially, the research group developed a working definition of “complications” after spine surgery using a survey of patients and surgeons4,16,30. A prospective analysis was then conducted with use of this patient-centered definition of perioperative complications. There was a clear trend toward an increased incidence of complications in patients who had undergone surgery for a neoplasm or infection. The complication incidence correlated with the surgical approach10,11,14,31,32. A higher number of patient comorbidities correlated with an increased risk of a major complication, a minor complication, and any complication. Some comorbidities independently predicted the occurrence of complications11,14.
An ICD-9-CM model of the prospectively developed comorbidity scoring approach was developed and applied to an administrative database. The predictive value of the score was compared with that of the Charlson measure. An increasing score correlated linearly with an increasing complication incidence, at least as well as the Charlson measure33.
Longitudinal data provide a more robust assessment of complication incidence than incident-only or admission-only data sets. An analysis of the MarketScan longitudinal database revealed that patient follow-up extending to 30 days substantially improved the sensitivity of our analyses in capturing adverse-event occurrence when compared with the sensitivity when adverse events were identified only as those occurring during a patient’s initial admission34. The overall increase in complication recognition was 35.8% with inclusion of longitudinal data, which provides one explanation for the poor accuracy of administrative reports that did not utilize longitudinal data for analysis10. Longitudinal data may be used to assess readmissions and reoperations and enable a more accurate assessment of perioperative adverse events27.
The overall retrospective complication incidences captured in this assessment were comparable with those obtained with similar prospective patient assessments14,31-33,35. Our results revealed higher complication incidences than have been noted in other, single-admission or non-longitudinal databases8,36-42.
Initially, we did not plan to incorporate use of BMP as a distinct predictor of adverse events. However, we included it after initial analyses revealed that use of BMP was associated with the incidence of adverse events in cervical spine surgery (OR = 1.34, 95% CI = 1.2 to 1.5, p < 0.0001)27.
Our assessment of patients who had undergone spine surgery showed an adverse-event rate of 13.95%, well within norms found in the literature and superior to the rates found in studies that used non-longitudinal data7,23. We built prediction models using different methods, primarily logistic regression with additive main effects and logistic regression with main effects as well as all 2 and 3-factor interactions (for subgroups of proportions >0.3%). Correlations between predictive scores derived with the various methods were high; for example, the correlation between LASSO and boosted additive trees was 0.81.
The model was less effective overall at predicting the type of complication, with AUCs of 0.60 to 0.62 for all complication types other than pulmonary complications. The model was effective at predicting pulmonary complications, with an AUC of 0.72.
Hence, our approach allowed us to reliably predict complication occurrence for given patients on the basis of operative, diagnostic, and patient-specific factors. A description of the use of the score and derivation of adverse-event probability is included in the Appendix.
There are important limitations of our model. A substantial weakness is the patient ages within the MarketScan database. Although the elderly constitute one of the most rapidly growing groups of patients receiving spine surgery, a plot of patient ages in the MarketScan database shows nearly no Medicare-aged patients (Fig. 2). While the incidence of spine surgery procedures is growing, the majority of the growth is in Medicare-aged patients, primarily because of the effects of degenerative disease on the aging spine1-3.
Administrative database studies rely on the accuracy of coding of both index procedures and comorbidities; errors may lead to inappropriate overreporting or underreporting. As is the case in all retrospective data analyses, there is a risk that confounders that were not assessed biased the study results. Although we accounted for these differences using both multivariate logistic regression and propensity score analysis, retrospective data cannot replace a prospective trial with these covariates balanced at the study outset.
The definition of comorbidities chosen for this analysis is broad, incorporating smoking status, history of psychiatric disorders, alcohol abuse, and other factors not used by other authors28,29,43,44. This broad approach may incorporate elements of limited clinical impact. Further development of the measure, and validation across multiple databases, will hopefully limit this risk and ensure that the measure appropriately models patient experience.
Finally, the predictive ability of the approach, while superior to that of the commonly used Charlson comorbidity score, still has room for improvement. We took a broad approach to assessing as many different factors as could be reliably extracted from a robust longitudinal administrative data set. While the overall complication incidence based on the administrative data appears representative, our ability to predict complication occurrence in the testing set only reached an AUC of 0.70. We believe that this may be improved by incorporating larger data sets and probably by expanding our assessment to Medicare-aged patients. However, this predictive accuracy may also reflect the inherent weaknesses of administrative data.
In conclusion, we present a novel data modeling effort that predicts the occurrence of complications in spine surgery on the basis of administrative data, with less accurate prediction of complication type. We believe that the development of a predictive modeling tool illustrating the risk of complication occurrence after spine surgery will aid in patient counseling and improve the accuracy of risk modeling strategies.
A description of the LASSO regularization and an example of adverse-event risk score calculation are available with the online version of this article as a data supplement at jbjs.org.
Investigation performed at Stanford University School of Medicine, Stanford, California
1. Deyo RA, Martin BI, Ching A, Tosteson ANA, Jarvik JG, Kreuter W, Mirza SK. Interspinous spacers compared with decompression or fusion for lumbar stenosis: complications and repeat operations in the Medicare population. Spine (Phila Pa 1976). 2013 ;38(10):865–72.
2. Deyo RA, Mirza SK. Trends and variations in the use of spine surgery. Clin Orthop Relat Res. 2006 ;443(443):139–46.
3. Deyo RA, Mirza SK, Martin BI, Kreuter W, Goodman DC, Jarvik JG. Trends, major medical complications, and charges associated with surgery for lumbar spinal stenosis in older adults. JAMA. 2010 ;303(13):1259–65.
4. Nasser R, Yadla S, Maltenfort MG, Harrop JS, Anderson DG, Vaccaro AR, Sharan AD, Ratliff JK. Complications in spine surgery. J Neurosurg Spine. 2010 ;13(2):144–57.
5. Patel N, Bagan B, Vadera S, Maltenfort MG, Deutsch H, Vaccaro AR, Harrop J, Sharan A, Ratliff JK. Obesity and spine surgery: relation to perioperative complications. J Neurosurg Spine. 2007 ;6(4):291–7.
6. Patil CG, Lad SP, Santarelli J, Boakye M. National inpatient complications and outcomes after surgery for spinal metastasis from 1993-2002. Cancer. 2007 ;110(3):625–30.
7. Whitmore RG, Stephen JH, Vernick C, Campbell PG, Yadla S, Ghobrial GM, Maltenfort MG, Ratliff JK. ASA grade and Charlson Comorbidity Index of spinal surgery patients: correlation with complications and societal costs. Spine J. 2014 ;14(1):31–8. Epub 2013 Apr 17.
8. Boakye M, Patil CG, Ho C, Lad SP. Cervical corpectomy: complications and outcomes. Neurosurgery. 2008 ;63(4)(Suppl 2):295–301, discussion :301-2.
9. Boakye M, Patil CG, Santarelli J, Ho C, Tian W, Lad SP. Cervical spondylotic myelopathy: complications and outcomes after spinal fusion. Neurosurgery. 2008 ;62(2):455–61, discussion :461-2.
10. Campbell PG, Malone J, Yadla S, Chitale R, Nasser R, Maltenfort MG, Vaccaro A, Ratliff JK. Comparison of ICD-9-based, retrospective, and prospective assessments of perioperative complications: assessment of accuracy in reporting. J Neurosurg Spine. 2011 ;14(1):16–22. Epub 2010 Dec 10.
11. Campbell PG, Yadla S, Nasser R, Malone J, Maltenfort MG, Ratliff JK. Patient comorbidity score predicting the incidence of perioperative complications: assessing the impact of comorbidities on complications in spine surgery. J Neurosurg Spine. 2012 ;16(1):37–43. Epub 2011 Oct 28.
12. Carreon LY, Puno RM, Dimar JR 2nd, Glassman SD, Johnson JR. Perioperative complications of posterior lumbar decompression and arthrodesis in older adults. J Bone Joint Surg Am. 2003 ;85-A(11):2089–92.
13. Lovecchio F, Hsu WK, Smith TR, Cybulski G, Kim B, Kim JY. Predictors of thirty-day readmission after anterior cervical fusion. Spine (Phila Pa 1976). 2014 ;39(2):127–33.
14. Campbell PG, Yadla S, Malone J, Zussman B, Maltenfort MG, Sharan AD, Harrop JS, Ratliff JK. Early complications related to approach in cervical spine surgery: single-center prospective study. World Neurosurg. 2010 ;74(2-3):363–8.
15. Rampersaud YR, Moro ER, Neary MA, White K, Lewis SJ, Massicotte EM, Fehlings MG. Intraoperative adverse events and related postoperative complications in spine surgery: implications for enhancing patient safety founded on evidence-based protocols. Spine (Phila Pa 1976). 2006 ;31(13):1503–10.
16. Lebude B, Yadla S, Albert T, Anderson DG, Harrop JS, Hilibrand A, Maltenfort M, Sharan A, Vaccaro AR, Ratliff JK. Defining “complications” in spine surgery: neurosurgery and orthopedic spine surgeons’ survey. J Spinal Disord Tech. 2010 ;23(8):493–500.
17. Wang MC, Chan L, Maiman DJ, Kreuter W, Deyo RA. Complications and mortality associated with cervical spine surgery for degenerative disease in the United States. Spine (Phila Pa 1976). 2007 ;32(3):342–7.
18. Bernheim SM, Grady JN, Lin Z, Wang Y, Wang Y, Savage SV, Bhat KR, Ross JS, Desai MM, Merrill AR, Han LF, Rapp MT, Drye EE, Normand SL, Krumholz HM. National patterns of risk-standardized mortality and readmission for acute myocardial infarction and heart failure. Update on publicly reported outcomes measures based on the 2010 release. Circ Cardiovasc Qual Outcomes. 2010 ;3(5):459–67. Epub 2010 Aug 24.
19. Bradley EH, Herrin J, Curry L, Cherlin EJ, Wang Y, Webster TR, Drye EE, Normand SL, Krumholz HM. Variation in hospital mortality rates for patients with acute myocardial infarction. Am J Cardiol. 2010 ;106(8):1108–12.
20. Bradley EH, Herrin J, Elbel B, McNamara RL, Magid DJ, Nallamothu BK, Wang Y, Normand SL, Spertus JA, Krumholz HM. Hospital quality for acute myocardial infarction: correlation among process measures and relationship with short-term mortality. JAMA. 2006 ;296(1):72–8.
21. Bratzler DW, Normand SL, Wang Y, O’Donnell WJ, Metersky M, Han LF, Rapp MT, Krumholz HM. An administrative claims model for profiling hospital 30-day mortality rates for pneumonia patients. PLoS One. 2011;6(4):e17401. Epub 2011 Apr 12.
22. Drye EE, Normand SL, Wang Y, Ross JS, Schreiner GC, Han L, Rapp M, Krumholz HM. Comparison of hospital risk-standardized mortality rates calculated by using in-hospital and 30-day models: an observational study with implications for hospital profiling. Ann Intern Med. 2012 ;156(1 Pt 1):19–26.
23. Keenan PS, Normand SL, Lin Z, Drye EE, Bhat KR, Ross JS, Schuur JD, Stauffer BD, Bernheim SM, Epstein AJ, Wang Y, Herrin J, Chen J, Federer JJ, Mattera JA, Wang Y, Krumholz HM. An administrative claims measure suitable for profiling hospital performance on the basis of 30-day all-cause readmission rates among patients with heart failure. Circ Cardiovasc Qual Outcomes. 2008 ;1(1):29–37.
24. Krumholz HM, Wang Y, Mattera JA, Wang Y, Han LF, Ingber MJ, Roman S, Normand SL. An administrative claims model suitable for profiling hospital performance based on 30-day mortality rates among patients with an acute myocardial infarction. Circulation. 2006 ;113(13):1683–92. Epub 2006 Mar 20.
25. Krumholz HM, Wang Y, Mattera JA, Wang Y, Han LF, Ingber MJ, Roman S, Normand SL. An administrative claims model suitable for profiling hospital performance based on 30-day mortality rates among patients with heart failure. Circulation. 2006 ;113(13):1693–701. Epub 2006 Mar 20.
26. Lindenauer PK, Bernheim SM, Grady JN, Lin Z, Wang Y, Wang Y, Merrill AR, Han LF, Rapp MT, Drye EE, Normand SL, Krumholz HM. The performance of US hospitals as reflected in risk-standardized 30-day mortality and readmission rates for medicare beneficiaries with pneumonia. J Hosp Med. 2010 ;5(6):E12–8.
27. Cole T, Veeravagu A, Jiang B, Ratliff JK. Usage of recombinant human bone morphogenetic protein in cervical spine procedures: analysis of the MarketScan longitudinal database. J Bone Joint Surg Am. 2014 ;96(17):1409–16.
28. 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.
29. Veeravagu A, Cole TS, Jiang B, Ratliff JK, Gidwani RA. The use of bone morphogenetic protein in thoracolumbar spine procedures: analysis of the MarketScan longitudinal database. Spine J. 2014 ;14(12):2929–37. Epub 2014 May 16.
30. Ratliff JK, Lebude B, Albert T, Anene-Maidoh T, Anderson G, Dagostino P, Maltenfort M, Hilibrand A, Sharan A, Vaccaro AR. Complications in spinal surgery: comparative survey of spine surgeons and patients who underwent spinal surgery. J Neurosurg Spine. 2009 ;10(6):578–84.
31. Yadla S, Malone J, Campbell PG, Maltenfort MG, Harrop JS, Sharan AD, Ratliff JK. Early complications in spine surgery and relation to preoperative diagnosis: a single-center prospective study. J Neurosurg Spine. 2010 ;13(3):360–6.
32. Yadla S, Malone J, Campbell PG, Nasser R, Maltenfort MG, Harrop JS, Sharan AD, Ratliff JK. Incidence of early complications in cervical spine surgery and relation to preoperative diagnosis: a single-center prospective study. J Spinal Disord Tech. 2011 ;24(1):50–4.
33. Chitale R, Campbell PG, Yadla S, Whitmore RG, Maltenfort MG, Ratliff JK. International classification of disease clinical modification 9 modeling of a patient comorbidity score predicts incidence of perioperative complications in a nationwide inpatient sample assessment of complications in spine surgery. J Spinal Disord Tech. 2015 ;28(4):126–33.
34. Veeravagu A, Cole TS, Azad TD, Ratliff JK. Improved capture of adverse events after spinal surgery procedures with a longitudinal administrative database. J Neurosurg Spine. 2015 ;23(3):374–82. Epub 2015 Jun 12.
35. Campbell PG, Malone J, Yadla S, Maltenfort MG, Harrop JS, Sharan AD, Ratliff JK. Early complications related to approach in thoracic and lumbar spine surgery: a single center prospective study. World Neurosurg. 2010 ;73(4):395–401.
36. Boakye M, Patil CG, Santarelli J, Ho C, Tian W, Lad SP. Laminectomy and fusion after spinal cord injury: national inpatient complications and outcomes. J Neurotrauma. 2008 ;25(3):173–83.
37. Kalanithi PS, Patil CG, Boakye M. National complication rates and disposition after posterior lumbar fusion for acquired spondylolisthesis. Spine (Phila Pa 1976). 2009 ;34(18):1963–9.
38. Lad SP, Nathan JK, Boakye M. Trends in the use of bone morphogenetic protein as a substitute to autologous iliac crest bone grafting for spinal fusion procedures in the United States. Spine (Phila Pa 1976). 2011 ;36(4):E274–81.
39. Lad SP, Patil CG, Berta S, Santarelli JG, Ho C, Boakye M. National trends in spinal fusion for cervical spondylotic myelopathy. Surg Neurol. 2009 ;71(1):66–9, discussion :69. Epub 2008 Jun 02.
40. Li G, Patil CG, Lad SP, Ho C, Tian W, Boakye M. Effects of age and comorbidities on complication rates and adverse outcomes after lumbar laminectomy in elderly patients. Spine (Phila Pa 1976). 2008 ;33(11):1250–5.
41. Patil CG, Santarelli J, Lad SP, Ho C, Tian W, Boakye M. Inpatient complications, mortality, and discharge disposition after surgical correction of idiopathic scoliosis: a national perspective. Spine J. 2008 ;8(6):904–10. Epub 2008 Mar 20.
42. Veeravagu A, Patil CG, Lad SP, Boakye M. Risk factors for postoperative spinal wound infections after spinal decompression and fusion surgeries. Spine (Phila Pa 1976). 2009 ;34(17):1869–72.
43. Harris MB, Reichmann WM, Bono CM, Bouchard K, Corbett KL, Warholic N, Simon JB, Schoenfeld AJ, Maciolek L, Corsello P, Losina E, Katz JN. Mortality in elderly patients after cervical spine fractures. J Bone Joint Surg Am. 2010 ;92(3):567–74.
44. McHenry TP, Mirza SK, Wang J, Wade CE, O’Keefe GE, Dailey AT, Schreiber MA, Chapman JR. Risk factors for respiratory failure following operative stabilization of thoracic and lumbar spine fractures. J Bone Joint Surg Am. 2006 ;88(5):997–1005.