Patients undergoing orthopaedic surgery often have multiple coexisting medical conditions (comorbidities) [13, 21, 24, 28, 39, 45, 58]. Preoperative risk assessment can help with decision-making and management strategies [11, 37, 50]. Several models for estimating risk based on coded comorbidities are currently in use for orthopaedic patients, but there is no consensus regarding the optimal approach [22, 37, 49, 51, 56, 57]. Selecting appropriate risk adjustment models can help hospitals contain costs while ensuring high levels of quality. Furthermore, inadequate comorbidity risk-adjustment might penalize practitioners and hospitals that care for the sickest patients .
Large administrative inpatient databases have been used to assess the effect of baseline comorbidity status on surgical care outcomes targeted in pay-for-performance initiatives such as mortality, morbidity, and discharge disposition [14, 22, 32-34, 37-39]. The Charlson Comorbidity Index  encompasses 19 medical conditions and is the most widely used comorbidity risk adjustment model in orthopaedic surgery. First reported in 1987 to predict 1-year mortality , this index subsequently was adapted for use with administrative databases . The Elixhauser measure , a more recent model including 31 conditions, is believed to be a better predictor of mortality in patients with cardiac, gastrointestinal, hepatobiliary, and oncologic conditions [6, 18, 27, 52]. Several prevalent comorbidities such as hypertension, obesity, weight loss, and psychiatric disorders that are included in the Elixhauser model are not included in the Charlson model .
We therefore determined whether there was a difference in the accuracy of the Charlson and Elixhauser comorbidity-based measures in predicting (1) in-hospital mortality after major orthopaedic surgery, (2) in-hospital adverse events, and (3) nonroutine discharge.
Patients and Methods
All data were extracted from the National Hospital Discharge Survey (NHDS) database [4, 8]. The NHDS is an annual probability sample survey of discharges from nonfederal, general, and short-stay hospitals in the United States [8, 19]. Sample data then were weighted to produce annual estimates of inpatient care . A maximum of seven medical diagnoses and four procedures were gathered and coded with the use of ICD-9-CM codes. Patient demographic information, hospital characteristics, and inpatient outcomes such as discharge disposition and hospital length of stay also were collected. Recognizing its utility to answer valuable clinical questions, the NHDS has been used extensively to analyze data associated with a wide range of diagnoses and procedures across different medical specialties [3, 25, 33, 34, 38, 44, 48]. Because of prior adequate data deidentification, our study was exempt from institutional review board approval.
Patients with a procedure code (ICD-9-CM) for primary TKA (81.84), primary THA (81.51), or spinal fusion (81.00 to 81.08) were included in the sample. Patients who underwent hip fracture (820.x) surgery also were included in the analysis. From a database with more than 500,000,000 patients treated between 1990 to 2007, an estimated 14,007,813 patients were identified and included in the analysis. The mean age of the patients was 66 ± 15 years, and female patients accounted for 62% of the study sample (Table 1). Overall, an estimated 0.80% of the included patients died during hospitalization. Mortality rates ranged from 0.30% for total joint arthroplasty and spinal fusion to 2.6% for operative treatment of hip fracture (Table 1).
Comorbidity burden was quantified using validated Charlson (adaptation by Deyo et al. ) and Elixhauser coding algorithms available for ICD-9-CM codes [9, 12, 43]. Dichotomous variables indicating the presence or absence of each Charlson and Elixhauser comorbidity were created, and their associations with mortality were assessed in bivariate analysis using chi-square tests. In addition, the original Charlson and the weighted Elixhauser scores, developed by van Walraven et al. , were computed and further stratified into groups (0, 1-2, 3-4, ≥ 5 and < 0, 0, 1-4, ≥ 5, respectively). The Charlson weights assigned to each comorbidity range from +1 to +6, while the Elixhauser weights range from −7 to +12 [5, 55]. Comorbidity scores then can be calculated for each patient by summing the individual weights of all comorbidities. Continuous variables (age, days of care) of the stratified Charlson and Elixhauser groups were analyzed with ANOVA, and categorical data (sex, mortality, adverse events, discharge status) were analyzed with the chi-square test (Table 2). Chronic pulmonary disease (11%) and uncomplicated diabetes mellitus (11%) were the most frequently encountered comorbidities when using the Charlson/Deyo algorithm (Table 3). Among all 31 Elixhauser comorbidities, uncomplicated hypertension (38%) was the most prevalent condition, followed by chronic pulmonary disease (11%) and uncomplicated diabetes mellitus (11%) (Table 4).
Multivariable binary logistic regression analyses were performed to assess the contributions of the individual Charlson (Table 5) and Elixhauser (Table 6) comorbidities to predicted in-hospital mortality, our primary response variable. Charlson and Elixhauser comorbidities with a p value less than 0.10 in bivariate analysis and present in at least 0.2 % of the population were included in the logistic regression modeling. Two main models were constructed; each of these regression models encompassed one of the two comorbidity-based scores, and age, sex, and year of surgery, as independent variables. A base model that included only age, sex, and year of surgery also was evaluated .
To determine which model best predicted inpatient mortality, receiver operating characteristic (ROC) curves were plotted and the regression models were compared on the basis of the area under the ROC curve (AUC) and its 95% CI [20, 54]. The AUC quantifies the ability of our models to assign a high probability of mortality to the patients who died . Values range from 0.50 to 1.0, with 0.50 indicating no ability to discriminate and 1.0 indicating perfect discrimination. In general, values less than 0.70 are considered to show poor discrimination, values between 0.70 and 0.80 can be considered acceptable, and between 0.80 and 0.90 excellent. In addition to the absolute improvement in predictive performance, we calculated the difference between two AUCs in percent beyond the predictive power of the base model including age, sex, and year of surgery . For instance, a difference in AUC between the Charlson and Elixhauser comorbidity scores of 0.80 and 0.90, when the baseline AUC is 0.75, corresponds to a 67% relative increase in AUC: 0.90-0.75 - 0.80-0.75/0.90-0.75 = 0.67. We calculated the relative improvement in predictive performance of the Elixhauser score to the Charlson score. Global model performance also was compared using the Nagelkerke pseudo R-square measure . Our secondary outcome variables, in-hospital adverse events and nonroutine discharge, were analyzed in an analogous fashion to in-hospital mortality. In-hospital adverse events were defined by ICD-9 codes (Appendix), following a coding approach used in other studies [34, 38]. A nonroutine discharge was defined as discharge to a skilled nursing facility or rehabilitation center. So as to set stricter standards owing to the large weighted sample size, a p value less than 0.001 was used to define significance in all analyses.
Elixhauser comorbidity adjustment provided better prediction of in hospital case-mortality (AUC, 0.86; 95% CI, 0.86-0.86) compared with the Charlson model (AUC, 0.83; 95% CI, 0.83-0.84) and the base model with no comorbidities (AUC, 0.81; 95% CI, 0.81-0.81) (Table 7). In terms of relative improvement in predictive ability, the Elixhauser model performed 60% better than the Charlson model. The base model already showed excellent discrimination and accounted for 12% of the variation (Nagelkerke R-square = 0.12). The further addition of the Charlson comorbidity score to the base model led to a 2% increase in the amount of variation explained. The inclusion of the Elixhauser score to the base model achieved greater discrimination than the base model alone and the model incorporating the Charlson index and explained 18% of the variability (Nagelkerke R-square = 0.18). The rates of in-hospital death, together with those of adverse events and nonroutine disposition, increased steadily with the number of comorbidities and index scores In multivariable logistic regression analysis, chronic renal failure (odds ratio [OR], 4.3; 95% CI, 4.2-4.4; p < 0.001), complicated diabetes mellitus (OR, 4.3; 95% CI, 4.2-4.5; p < 0.001), and myocardial infarction (OR, 3.9; 95% CI, 3.9-4.0; p < 0.001) were the Charlson conditions associated with the greatest odds of in-hospital death (Table 5). Weight loss (OR, 5.0; 95% CI, 4.8-5.1; p < 0.001), pulmonary circulation disorders (OR, 4.5; 95% CI, 4.4-4.7; p < 0.001), and chronic renal failure (OR, 4.4; 95% CI, 4.3-4.6; p < 0.001) had the highest adjusted odds of inpatient mortality in the Elixhauser algorithm (Table 6).
Although the Elixhauser measure (AUC, 0.65; 95% CI, 0.65-0.65) was 100% more accurate than the Charlson measure (AUC, 0.64; 95% CI, 0.64-0.64) in predicting adverse events in terms of relative improvement, the discriminative ability of the model was poor and the difference in the absolute improvement in predictive power between the two models (AUC, 0.01) is of dubious clinical importance (Table 7).
The Elixhauser and Charlson models showed the same degree of discrimination for nonroutine discharge prediction after major orthopaedic surgery (AUC, 0.81; 95% CI, 0.81-0.81 for both scales) (Table 7).
Surgical mortality and morbidity rates are important parameters of in-hospital quality of care [11, 37, 50]. Given the increasing age and complexity of patients undergoing orthopaedic surgery, it is necessary to appropriately adjust for patient risk, recognizing that the underlying nature of some patients’ conditions may make them more likely than others to experience poor outcomes. We therefore assessed and compared the two most commonly used comorbidity risk adjustment models in orthopaedic surgery, the Charlson and Elixhauser measures, regarding their ability to predict in-hospital death, adverse events, and nonroutine discharge.
Our results should be interpreted after taking into account numerous factors. Despite access to large numbers and associated power, administrative databases have several recognized limitations [14, 23, 29]. First, the NHDS dataset is based on billing data from ICD-9-CM codes, and such a coding system may not fully capture the patient population of interest . In particular, it has been suggested that administrative databases tend to underreport chronic medical conditions that are considered less acute in the perioperative orthopaedic surgery setting [16, 31, 33, 34, 39]. Second, the possibility of errors in coding of the diagnoses and procedures cannot be avoided ; however, misclassification mistakes distribute evenly in large-scale studies . Third, the NHDS database does not include data regarding the timing of diagnoses, which hinders the differentiation of baseline comorbidities from complications . Analyses of risk-adjusted mortality rates should adjust mortality rates only for baseline comorbid diseases, not complications that arise from surgery . The degree to which this issue influenced our results is unclear, although it has been reported that the majority of common diagnoses are comorbidities rather than adverse events [15, 30, 36]. Fourth, the NHDS enabled only ascertainment of inpatient outcomes, and thus postdischarge complications and readmissions were not captured. Fifth, we performed risk-adjustment using administrative data only; inclusion of clinically relevant variables such as the American Society of Anesthesiologists (ASA) score or the Frailty Index developed by the Canadian Study of Health and Aging may have improved model performance [24, 40, 58]. Finally, we did not perform any clinical data abstraction from medical records, which is considered the gold standard risk adjustment method in these comparisons ; thus, we were able to compare the Charlson and Elixhauser measures only for their relative performance.
We found that the Elixhauser comorbidity risk adjustment model performed numerically better than the Charlson model in predicting in-hospital mortality after major orthopaedic surgery. Although differences in the AUC values between the two comorbidity-based measures were small, it has been noted that even slight improvements in the AUC for such indexes can translate into quantifiable reductions in confounding bias . Overall, the AUC values for inpatient mortality for the Charlson and Elixhauser comorbidity-based measures in our study were comparable to or slightly higher than those described in other patient populations [6, 18, 27, 52]. Consistent with a study by Nikkel et al.  in patients with hip fractures, the Elixhauser weight loss or malnutrition comorbidity was the major factor influencing mortality. We found some comorbidities (eg, hypothyroidism, obesity, uncomplicated diabetes, hypertension) to be associated with decreased odds of inpatient mortality. It is counterintuitive that these comorbidities would protect against inpatient death. It may be that these comorbidities are most common in patients with less overall infirmity compared with the average orthopaedic inpatient.
The occurrence of in-hospital adverse events after major orthopaedic surgery was slightly more accurately predicted with the Elixhauser comorbidity index. However, in line with the study by Gordon et al.  looking at the influence of the Elixhauser and Charlson measures on reoperations after THA, the predictive accuracy of both models to detect adverse events was poor (AUC values < 0.70). An AUC value approximating 0.70 is considered acceptable for discrimination and validation of methods for ongoing use ; we therefore could not validate the Charlson or Elixhauser measures in terms of predicting perioperative complications after major orthopaedic surgery. There may be something beyond measurable comorbidities that is yet unaccounted for in orthopaedic inpatient morbidity.
Both comorbidity indexes provided clinically relevant insight for estimating nonroutine discharge after orthopaedic surgery; the Elixhauser score was no better than the Charlson score. This finding suggests that the Charlson and Elixhauser indexes are valid prediction tools for healthcare resource use risk adjustment, and researchers should choose between them based more on their availability and comfort with the method .
The Elixhauser measure has not been introduced in orthopaedic surgery research until recently [17, 26, 39, 57, 59, 60], perhaps because of scarce reported comparisons with the Charlson model  and concerns regarding the inclusion of too many explanatory variables (31 variables), therefore requiring a relatively large sample size . The main attractiveness of the use of large administrative databases for medical research lies in the possibility of studying rare occurrence events, such as inpatient mortality, that otherwise would be difficult to investigate in small population studies .
Testing comorbidity risk adjustment measure performance in orthopaedic surgery is worthy of future study. Further research comparing the Charlson and Elixhauser methods with the less accessible (costwise) risk adjustment methods of Disease Staging (Thomson Medstat Inc, Ann Arbor, MI, USA) and All Patient Refined Diagnosis Related Groups (3M Health Information Systems, Wallingford, CT, USA) is warranted . In addition, we currently are testing and validating a specific comorbidity-based measure for outcome prediction after orthopaedic surgery. The Elixhauser comorbidity measure outperformed the widely used Charlson measure in predicting inpatient mortality and morbidity after major orthopaedic surgery, and its more extensive use in claims-based studies should be considered. Future research assessing and comparing the performance of these measures in predicting long-term outcomes would be of value.
1. Alosh H, Li D, Riley LH 3rd, Skolasky RL. Health care burden of anterior cervical spine surgery: national trends in hospital charges and length of stay, 2000 to 2009. J Spinal Disord Tech
. 2013 October 16 [Epub ahead of print].
2. Baldwin LM, Klabunde CN, Green P, Barlow W, Wright G. In search of the perfect comorbidity measure for use with administrative claims data: does it exist? Med Care.
3. Bhattacharyya T, Iorio R, Healy WL. Rate of and risk factors for acute inpatient mortality after orthopaedic surgery. J Bone Joint Surg Am.
4. Centers for Disease Control and Prevention. National Hospital Discharge Survey. Available at: http://www.cdc.gov/nchs/nhds.htm
. Accessed May 6, 2014.
5. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis.
6. Chu YT, Ng YY, Wu SC. Comparison of different comorbidity measures for use with administrative data in predicting short- and long-term mortality. BMC Health Serv Res.
7. DeFrances CJ, Lucas CA, Buie VC, Golosinskiy A. 2006 National Hospital Discharge Survey. Natl Health Stat Report.
8. Dennison C, Pokras R. Design and operation of the National Hospital Discharge Survey: 1988 redesign. Vital Health Stat
9. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol.
10. Dimick JB. How should we risk-adjust hospital outcome comparisons? Arch Surg.
11. Dimick JB, Welch HG, Birkmeyer JD. Surgical mortality as an indicator of hospital quality: the problem with small sample size. JAMA.
12. Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care.
13. Everhart JS, Altneu E, Calhoun JH. Medical comorbidities are independent preoperative risk factors for surgical infection after total joint arthroplasty. Clin Orthop Relat Res.
14. Fleischut PM, Mazumdar M, Memtsoudis SG. Perioperative database research: possibilities and pitfalls. Br J Anaesth.
15. Ghali WA, Quan H, Brant R. Risk adjustment using administrative data: impact of a diagnosis-type indicator. J Gen Intern Med.
16. Gonzalez Della Valle A, Chiu YL, Ma Y, Mazumdar M, Memtsoudis SG. The metabolic syndrome in patients undergoing knee and hip arthroplasty: trends and in-hospital outcomes in the United States. J Arthroplasty
. 2012;27:1743-1749 e1741.
17. Gordon M, Stark A, Skoldenberg OG, Karrholm J, Garellick G. The influence of comorbidity scores on re-operations following primary total hip replacement: comparison and validation of three comorbidity measures. Bone Joint J.
18. Grendar J, Shaheen AA, Myers RP, Parker R, Vollmer CM Jr, Ball CG, Quan ML, Kaplan GG, Al-Manasra T, Dixon E. Predicting in-hospital mortality in patients undergoing complex gastrointestinal surgery: determining the optimal risk adjustment method. Arch Surg.
19. Hall MJ, DeFrances CJ, Williams SN, Golosinskiy A, Schwartzman A. National Hospital Discharge Survey: 2007 summary. Natl Health Stat Report.
20. Harrell FE Jr, Lee KL, Califf RM, Pryor DB, Rosati RA. Regression modelling strategies for improved prognostic prediction. Stat Med.
21. 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.
22. Humphries W, Jain N, Pietrobon R, Socolowski F, Cook C, Higgins L. Effect of the Deyo score on outcomes and costs in shoulder arthroplasty patients. J Orthop Surg (Hong Kong).
23. Iezzoni LI. Assessing quality using administrative data. Ann Intern Med.
24. Johnson CC, Sodha S, Garzon-Muvdi J, Petersen SA, McFarland EG. Does preoperative American Society of Anesthesiologists score relate to complications after total shoulder arthroplasty? Clin Orthop Relat Res.
25. Khanna S, Keddis MT, Noheria A, Baddour LM, Pardi DS. Acute kidney injury is an independent marker of severity in Clostridium difficile infection: a nationwide survey. J Clin Gastroenterol.
26. Krupic F, Eisler T, Eliasson T, Garellick G, Gordon M, Karrholm J. No influence of immigrant background on the outcome of total hip arthroplasty: 140,299 patients born in Sweden and 11,539 immigrants in the Swedish Hip Arthroplasty Register. Acta Orthop.
27. Lieffers JR, Baracos VE, Winget M, Fassbender K. A comparison of Charlson and Elixhauser comorbidity measures to predict colorectal cancer survival using administrative health data. Cancer.
28. Maradit Kremers H, Visscher SL, Kremers WK, Naessens JM, Lewallen DG. Obesity increases length of stay and direct medical costs in total hip arthroplasty. Clin Orthop Relat Res
29. Memtsoudis SG. Limitations associated with the analysis of data from administrative databases. Anesthesiology.
2009;111:449; author reply 450-451.
30. Memtsoudis SG, Gonzalez Della Valle A, Besculides MC, Gaber L, Sculco TP. In-hospital complications and mortality of unilateral, bilateral, and revision TKA: based on an estimate of 4,159,661 discharges. Clin Orthop Relat Res.
31. Memtsoudis SG, Kirksey M, Ma Y, Chiu YL, Mazumdar M, Pumberger M, Girardi FP. Metabolic syndrome and lumbar spine fusion surgery: epidemiology and perioperative outcomes. Spine (Phila Pa 1976
32. Memtsoudis SG, Stundner O, Rasul R, Sun X, Chiu YL, Fleischut P, Danninger T, Mazumdar M. Sleep apnea and total joint arthroplasty under various types of anesthesia: a population-based study of perioperative outcomes. Reg Anesth Pain Med.
33. Menendez ME, Neuhaus V, Bot AG, Ring D, Cha TD. Psychiatric disorders and major spine surgery: epidemiology and perioperative outcomes. Spine (Phila Pa 1976).
34. Menendez ME, Neuhaus V, Bot AG, Vrahas MS, Ring D. Do psychiatric comorbidities influence inpatient death, adverse events, and discharge after lower extremity fractures? Clin Orthop Relat Res.
35. Mittlbock M, Heinzl H. A note on R2 measures for Poisson and logistic regression models when both models are applicable. J Clin Epidemiol.
36. Myers RP, Quan H, Hubbard JN, Shaheen AA, Kaplan GG. Predicting in-hospital mortality in patients with cirrhosis: results differ across risk adjustment methods. Hepatology.
37. Neuhaus V, King J, Hageman MG, Ring DC. Charlson comorbidity indices and in-hospital deaths in patients with hip fractures. Clin Orthop Relat Res.
38. Neuhaus V, Swellengrebel CH, Bossen JK, Ring D. What are the factors influencing outcome among patients admitted to a hospital with a proximal humeral fracture? Clin Orthop Relat Res.
39. Nikkel LE, Fox EJ, Black KP, Davis C, Andersen L, Hollenbeak CS. Impact of comorbidities on hospitalization costs following hip fracture. J Bone Joint Surg Am.
40. Patel KV, Brennan KL, Brennan ML, Jupiter DC, Shar A, Davis ML. Association of a modified frailty index with mortality after femoral neck fracture in patients aged 60 years and older. Clin Orthop Relat Res.
41. Pine M, Jordan HS, Elixhauser A, Fry DE, Hoaglin DC, Jones B, Meimban R, Warner D, Gonzales J. Enhancement of claims data to improve risk adjustment of hospital mortality. JAMA.
42. Quan H, Li B, Couris CM, Fushimi K, Graham P, Hider P, Januel JM, Sundararajan V. Updating and validating the Charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries. Am J Epidemiol.
43. Quan H, Sundararajan V, Halfon P, Fong A, Burnand B, Luthi JC, Saunders LD, Beck CA, Feasby TE, Ghali WA. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care.
44. Rincon F, Rossenwasser RH, Dumont A. The epidemiology of admissions of nontraumatic subarachnoid hemorrhage in the United States. Neurosurgery.
45. Schairer WW, Vail TP, Bozic KJ. What are the rates and causes of hospital readmission after total knee arthroplasty? Clin Orthop Relat Res.
46. Schneeweiss S, Seeger JD, Maclure M, Wang PS, Avorn J, Glynn RJ. Performance of comorbidity scores to control for confounding in epidemiologic studies using claims data. Am J Epidemiol.
47. Schneeweiss S, Wang PS, Avorn J, Glynn RJ. Improved comorbidity adjustment for predicting mortality in Medicare populations. Health Serv Res.
48. Scott FI, Osterman MT, Mahmoud NN, Lewis JD. Secular trends in small-bowel obstruction and adhesiolysis in the United States: 1988-2007. Am J Surg.
49. Singh JA, Sperling JW, Cofield RH. Ninety day mortality and its predictors after primary shoulder arthroplasty: an analysis of 4,019 patients from 1976-2008. BMC Musculoskelet Disord.
50. Sloan FA, Perrin JM, Valvona J. In-hospital mortality of surgical patients: is there an empiric basis for standard setting? Surgery.
51. Soohoo NF, Farng E, Lieberman JR, Chambers L, Zingmond DS. Factors that predict short-term complication rates after total hip arthroplasty. Clin Orthop Relat Res.
52. Southern DA, Quan H, Ghali WA. Comparison of the Elixhauser and Charlson/Deyo methods of comorbidity measurement in administrative data. Med Care.
53. Tseng VL, Yu F, Lum F, Coleman AL. Risk of fractures following cataract surgery in Medicare beneficiaries. JAMA.
54. Uno H, Cai T, Pencina MJ, D'Agostino RB, Wei LJ. On the C-statistics for evaluating overall adequacy of risk prediction procedures with censored survival data. Stat Med.
55. Walraven C, Austin PC, Jennings A, Quan H, Forster AJ. A modification of the Elixhauser comorbidity measures into a point system for hospital death using administrative data. Med Care.
56. Voskuijl T, Hageman M, Ring D. Higher Charlson comorbidity index scores are associated with readmission after orthopaedic surgery. Clin Orthop Relat Res.
57. Vulcano E, Lee YY, Yamany T, Lyman S, Valle AG. Obese patients undergoing total knee arthroplasty have distinct preoperative characteristics: an institutional study of 4718 patients. J Arthroplasty.
58. Wasielewski RC, Weed H, Prezioso C, Nicholson C, Puri RD. Patient comorbidity: relationship to outcomes of total knee arthroplasty. Clin Orthop Relat Res.
59. Yoshihara H, Yoneoka D. Trends in the surgical treatment for spinal metastasis and the in-hospital patient outcomes in the United States from 2000 to 2009. Spine J
. 2013 pii: S1529-9430(13)01844-5.
60. Yoshihara H, Yoneoka D. Incidental dural tear in spine surgery: analysis of a nationwide database. Eur Spine J.
List of ICD-9 codes included to identify adverse events