CABG indicates coronary artery bypass graft surgery; CMS, Centers for Medicare & Medicaid Services; FY, fiscal year; MI, myocardial infarction; VA, Veterans Affairs.aPatients may have had more than 1 percutaneous coronary intervention (PCI) care episode over the 10-year study period.bPatients may have had more than 1 surgical episode in the 24 months after a PCI episode.
Left, plots created with R package GGPLOT222; smoothed trends were fitted using the loess algorithm. The lines represent the estimate of the rate and the shaded area around the line represents the 95% CI. Right, density plots of number of operations per day; y-axis intervals in blue indicate range from 0 to 50 operations/day. BMS indicates bare metal stent; DES, drug-eluting stent; MACE, major adverse cardiac event.
eAppendix 1. Variable Definitions
eAppendix 2. Comparison of Matched Cases to Matched Controls
eAppendix 3. Case Report Forms
eTable. Alternate Models of Perioperative MACE
eFigure. Adjusted Odds Ratios of MACE across Time by Stent Type
Hawn MT, Graham LA, Richman JS, Itani KMF, Henderson WG, Maddox TM. Risk of Major Adverse Cardiac Events Following Noncardiac Surgery in Patients With Coronary Stents. JAMA. 2013;310(14):1462-1472. doi:10.1001/jama.2013.278787
Guidelines recommend delaying noncardiac surgery in patients after coronary stent procedures for 1 year after drug-eluting stents (DES) and for 6 weeks after bare metal stents (BMS). The evidence underlying these recommendations is limited and conflicting.
To determine risk factors for adverse cardiac events in patients undergoing noncardiac surgery following coronary stent implantation.
Design, Setting, and Participants
A national, retrospective cohort study of 41 989 Veterans Affairs (VA) and non-VA operations occurring in the 24 months after a coronary stent implantation between 2000 and 2010. Nonlinear generalized additive models examined the association between timing of surgery and stent type with major adverse cardiac events (MACE) adjusting for patient, surgery, and cardiac risk factors. A nested case-control study assessed the association between perioperative antiplatelet cessation and MACE.
Main Outcomes and Measures
A composite 30-day MACE rate of all-cause mortality, myocardial infarction, and cardiac revascularization.
Within 24 months of 124 844 coronary stent implantations (47.6% DES, 52.4% BMS), 28 029 patients (22.5%; 95% CI, 22.2%-22.7%) underwent noncardiac operations resulting in 1980 MACE (4.7%; 95% CI, 4.5%-4.9%). Time between stent and surgery was associated with MACE (<6 weeks, 11.6%; 6 weeks to <6 months, 6.4%; 6-12 months, 4.2%; >12-24 months, 3.5%; P < .001). MACE rate by stent type was 5.1% for BMS and 4.3% for DES (P < .001). After adjustment, the 3 factors most strongly associated with MACE were nonelective surgical admission (adjusted odds ratio [AOR], 4.77; 95% CI, 4.07-5.59), history of myocardial infarction in the 6 months preceding surgery (AOR, 2.63; 95% CI, 2.32-2.98), and revised cardiac risk index greater than 2 (AOR, 2.13; 95% CI, 1.85-2.44). Of the 12 variables in the model, timing of surgery ranked fifth in explanatory importance measured by partial effects analysis. Stent type ranked last, and DES was not significantly associated with MACE (AOR, 0.91; 95% CI, 0.83-1.01). After both BMS and DES placement, the risk of MACE was stable at 6 months. A case-control analysis of 284 matched pairs found no association between antiplatelet cessation and MACE (OR, 0.86; 95% CI, 0.57-1.29).
Conclusions and Relevance
Among patients undergoing noncardiac surgery within 2 years of coronary stent placement, MACE were associated with emergency surgery and advanced cardiac disease but not stent type or timing of surgery beyond 6 months after stent implantation. Guideline emphasis on stent type and surgical timing for both DES and BMS should be reevaluated.
Noncardiac surgery after recent coronary stent placement is associated with increased risk of adverse cardiac events. Consequently, it is desirable to delay elective surgery as long as possible after coronary stent placement. In 2004, drug-eluting stents (DES) were approved and overtook bare metal stents (BMS) as the preferred revascularization strategy.1 Reports of unanticipated late stent thrombosis after cessation of dual antiplatelet therapy (APT) and case reports of stent thrombosis in patients with DES undergoing noncardiac surgery led to a revision of the American Heart Association (AHA)/American College of Cardiology (ACC) guidelines in 2007.1- 8 The revised guidelines recommend continuing dual APT for all patients at least 1 year after DES implantation.9 For patients with DES undergoing noncardiac surgery, class IIa recommendations, based on level C evidence, state the following: (1) elective surgery after DES implantation should be delayed until completion of 1 year of dual APT, or (2) if the surgery is urgent, the surgery should be performed without cessation of APT. The guidelines for DES differ from those for BMS, which recommend a delay in surgery and temporary cessation of APT after 4 to 6 weeks from stent placement.10
Approximately 600 000 percutaneous coronary stent procedures are performed annually in the United States.11,12 Twelve percent to 23% of these patients undergo noncardiac surgery within 2 years of coronary stent placement.13- 17 Delaying necessary noncardiac surgery can pose a significant clinical dilemma for a large number of patients. The delays in surgery recommended by the guidelines are based on a limited and conflicting evidence base. Case series early in the DES experience suggested high rates of major adverse cardiac events (MACE) after noncardiac surgery. However, subsequent, larger multicenter cohort studies reported MACE rates similar to BMS MACE rates.13,14,18 Small series assessing perioperative APT management found no evidence that continued perioperative APT mitigates the risk of MACE. It is not clear whether the lower observed MACE rates in more recent studies are attributable to the effectiveness of guideline-driven delays of elective surgery together with continuing perioperative APT or reflect more reliable estimates of perioperative MACE rates in populations with stents, or both.
To better understand the relationship between stent type, APT, and MACE associated with noncardiac surgery after coronary stent placement, we evaluated a national cohort of Veterans Affairs (VA) patients who had either coronary BMS or DES placed between 2000 and 2010. We hypothesized that early surgery is associated with higher MACE rates after coronary stent placement, particularly in patients with DES, and that continued APT reduces the risk of postoperative MACE.
We conducted a retrospective cohort study of patients undergoing noncardiac surgery within 2 years after cardiac stent implantation to examine the relationship between stent type and time from stent to surgery with a composite adverse event of myocardial infarction (MI), revascularization, and all-cause mortality (MACE). We conducted 3 analyses to address the hypotheses. First, we constructed a multivariable regression model to determine risk factors for MACE and the strength of their association. Second, we assessed MACE rates as a function of time between stent and surgery and stent type. Third, we assessed the association of APT cessation with MACE. The study protocol was reviewed and approved by the local VA institutional review board of each coauthor with waiver of informed consent.
Cardiac stents were identified in the VA’s National Patient Care Databases (NPCD) and the VA Clinical Assessment, Reporting, and Tracking (CART) Program. Noncardiac surgery occurring in the VA was identified in the VA Surgical Quality Improvement Program database (VASQIP) and noncardiac surgery occurring outside of the VA was identified using Centers for Medicare & Medicaid Services (CMS) data for the 73% of veterans in the cohort who had dual VA-Medicare eligibility. Demographics and comorbidities were obtained from the VA NPCD or CMS inpatient, outpatient, and carrier base files. Death was obtained from the VA Vital Status File. Additional laboratory results and medication prescriptions were obtained from the VA Decision Support System.
For the nested case-control portion of this study, we abstracted data from the VA electronic health record. Chart abstraction began March 2012 and concluded in March 2013. Standardized data collection forms were developed, and all chart abstractors were trained in accordance with the procedure manual.
We identified all coronary stents implanted in VA facilities between 2000-2010 using codes from the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) (36.06 for BMS or 36.07 for DES) and direct abstraction from the CART Program data files. Percutaneous coronary intervention (PCI) care episodes were defined as a single visit to the catheterization laboratory for a PCI procedure, where 1 or more stents were implanted. Noncardiac surgical procedures were defined using Current Procedural Terminology (CPT) codes 10000 to 32999 and 34000 to 69999. We excluded minor surgeries, such as endoscopic procedures (CPT 43200-43272, 45300-45392, 46600-46608), and minor musculoskeletal procedures, such as application of a cast and joint aspiration (29000-29750). Operations preceded by an intervening coronary artery bypass graft surgery or non-VA stent or occurring during the same hospitalization as the PCI were excluded (Figure 1). The unit of analysis was the first surgical procedure occurring during a hospitalization within 2 years after a coronary stent placement. Because outcomes were assessed over a 30-day period after surgery, any subsequent surgeries occurring within 30 days after the index procedure were excluded from the analysis. For patients with multiple PCI care episodes, the timing between stent and surgery was measured from the most recent PCI care episode prior to surgery. Further details on the construction of the study cohort have been published.17
The outcome variable for the study was MACE within 30 days of exposure to noncardiac surgery. MACE was a composite variable including death from any cause, MI (ICD-9-CM codes 410.xx or VASQIP nurse-abstracted MI), or coronary revascularization (ICD-9-CM 00.66, 36.01-36.09; CPT: 33510-33519, 33520-33523, 33530-33536, 92973-92984, 92995-92998).
Noncardiac surgery was categorized using the primary CPT code: integumentary, 10040-19999; musculoskeletal, 20000-29999 (except amputation classified under vascular); respiratory, 30000-32999; vascular, 34000-37799 plus 27290, 27295, 27598, 27880-27899, 28801-28825; digestive, 40000-49999; genitourinary, 50000-58999; nervous, 61000-64999; or eye/ear, 65000-69999. Procedures with CPT codes not listed here were categorized as “other.” Procedure complexity was estimated from 2011 CMS work relative value units for the primary CPT code.
A patient’s cardiac risk at the time of noncardiac surgery was estimated from the 6-point revised cardiac risk index (rCRI) using administrative diagnosis codes from the International Classification of Diseases, Ninth Revision (ICD-9). The rCRI was calculated from ICD-9 diagnostic codes for congestive heart failure (CHF), stroke, MI, and diabetes; CPT codes associated with high-risk surgery; and laboratory data identifying 1 or more serum creatinine values greater than 2 mg/dL in the year prior to surgery.19 An insulin prescription in the Decision Support System pharmacy data within 12 months of surgery was used to identify insulin-dependent diabetes in patients with an ICD-9 code for diabetes. The rCRI was analyzed as both an ordinal and categorical variable: low risk (1 point), moderate risk (2 points), or high risk (≥3 points). Additional comorbidities at the time of surgery were identified in the VA NPCD and CMS data using ICD-9 diagnosis codes (listed in eAppendix 1 in the Supplement).
The nested case-control subset was restricted to (1) VA operations (because these were the only records available for review), (2) MI or revascularization end points, and (3) surgeries occurring more than 6 weeks after stent placement. Operations that occurred in the first 6 weeks after stent placement and operations followed by death alone were excluded (Figure 1). After exclusions, we matched cardiac MACE by fiscal year of operation, CPT category, work relative value unit (within 6 units), stent type, rCRI, and time from stent to operation (within 2 weeks) using 24 691 potential controls from VA surgeries that were not followed by a MACE (eAppendix 2 in the Supplement). Separate abstraction forms were assigned for exposure (preoperative antiplatelet management) and outcome (MACE and bleeding) so that an abstractor did not assess both for the same patient (eAppendix 3 in the Supplement). Uncertainty of an exposure or outcome variable was adjudicated by 2 of the senior investigators (M.T.H., T.M.M.).
To determine factors associated with MACE, generalized additive models were used to determine the relationship between time from stent to surgery and MACE with adjustment for stent type, surgical characteristics, cardiac risk factors, and comorbid conditions. Generalized additive models were used to allow time between stent and surgery to be fit as a linear or nonlinear term in assessing the relationship between surgical timing and MACE.20 The approximate P values for spline terms are derived using a score test and algorithmically estimated degrees of freedom. To examine the relative contribution of variables in the adjusted models, we calculated the analysis of variance χ2 for each variable minus its degrees of freedom (χ2 − df).21 The statistical threshold for significance was set at P = .05 for a 2-tailed test. To account for confounding by indication in choice of stent type, we conducted analyses using propensity score quintiles and inverse propensity weighting. We restricted this analysis to patients with stents placed after fiscal year 2003, when DES were widely available for implantation. Inverse probability weights were divided into quintiles and incorporated into the models.
All univariable and bivariable statistics were calculated using SAS version 9.2 (SAS Institute) and generalized additive models used R package MGCV for spline models. Plots of unadjusted data were created with R package GGPLOT222 and smoothed trends were fitted using the loess algorithm. For the nested case-control study, univariable and bivariable statistics were calculated to examine differences in medication management by MACE. Odds ratios (ORs) were calculated with conditional logistic regression to account for matched pairs.23
Of the 124 844 PCI episodes of care occurring in 2000-2010, a total of 28 029 patients (22.5%) met study inclusion criteria and underwent 41 989 surgical procedures within 24 months (22.5%; 95% CI, 22.2%-22.7%) (Figure 1). Patient demographics and comorbidities along with stent and surgical characteristics are shown in Table 1 and Table 2. A total of 1980 MACE (4.7%) occurred within 30 days of surgery: 1170 MI or repeat revascularization without death, 141 MI or repeated revascularization with death, and 669 death alone. In unadjusted analyses, MACE rates differed significantly by stent type: BMS (5.1%) vs DES (4.3%, P < .001). Markers of ischemic heart disease were associated with MACE, including MI or CHF in the past 6 months (13.6% and 12.0%, respectively), and rCRI score (Table 1). In addition, operations occurring after publication of the 2007 ACC/AHA guidelines were associated with lower MACE rates (3.5%) compared with before the guidelines’ publication (5.1%, P < .001).
The results of the generalized additive models of MACE assessing time from stent to surgery as a continuous linear or nonlinear term and the relative contribution of model covariates (χ2 − df) to MACE are shown in Table 3. In the overall model of MACE, nonelective presentation for the surgical hospitalization was the most explanatory determinant, followed by conditions associated with ischemic cardiac disease, including recent MI or CHF, and higher rCRI score, whereas stent type was not significantly associated with MACE and was ranked 12th in explanatory importance of the 12 variables in the model. There was no significant interaction between stent type and time to surgery (P = .56 for BMS and P = .20 for DES). The plot of the adjusted OR over time by stent type is provided in the eFigure in the Supplement. Because of the possibility of multicolinearity between variables included in the rCRI and as independent variables in the model (ie, history of coronary artery disease and recent MI), we assessed maximum variance inflation factors for all rCRI component variables and found it to be less than 1.1 for all variables assessed. In addition, a comparison of the model output excluding variables that are also considered in rCRI (operation type, MI in past 6 months, CHF admission in past 6 months, chronic kidney disease) is provided in the eTable in the Supplement, and the estimates for rCRI and stent type did not change substantively.
Time from stent to surgery was correlated with MACE, with higher rates observed for surgery closer to stent implantation (Figure 2A), nonelective admission source (Figure 2B), rCRI category (Figure 2C), and recent MI (Figure 2D). After adjustment, the odds of a MACE for surgery between 6 weeks and 6 months after DES placement was lower than for BMS (adjusted OR [AOR], 0.75; 95% CI, 0.62-0.91) and not significantly different for surgery less than 6 weeks (AOR, 1.1; 95% CI, 0.8-1.5) or more than 6 months after stent implantation (AOR, 0.92; 95% CI, 0.82-1.05). In the propensity analysis, stent type was significant (P = .001) with lower odds of MACE for surgery after DES placement (AOR, 0.87; 95% CI, 0.80-0.94) (eTable in the Supplement). Because the direction of the estimate did not rectify concern for confounding by indication for stent type, and given the need to truncate the cohort, we elected to not pursue modeling with propensity for DES. A prior study has also found limited value of propensity adjustment over multivariable regression modeling for outcomes by stent type.24
To investigate the association between APT management around the time of surgery and MACE, we performed a case-control study on the subset of VA surgical procedures. Of the 369 abstracted VA cases, a MACE was confirmed in 284 (77.0%). There was no significant difference in the likelihood of receiving dual APT prior to surgery (59.9% cases vs 55.6% controls; P = .43) or completely stopping APT for at least 5 days (22.9% cases vs 25.4% controls; P = .49) (Table 4). In matched analyses, there was no association between complete APT cessation and adverse cardiac events (OR, 0.86; 95% CI, 0.57-1.29). Post hoc power analyses indicated that the cohort had 80% power to detect an OR of 1.68 with α of .05.
To assess the robustness of these findings, we conducted several sensitivity analyses. First, to understand the association of the 2007 perioperative guidelines with the findings and its potential relationship with stent selection, we restricted the cohort to the 32 102 operations occurring prior to 2007 and observed no association between DES and higher MACE rates prior to publication of the ACC/AHA guidelines (AOR, 0.97; 95% CI, 0.86-1.09 compared with BMS). Second, to understand MACE rates among elective operations only, we restricted the cohort to only elective and outpatient procedures and obtained similar results (DES AOR, 0.90; 95% CI, 0.81-1.01 compared with BMS). Third, we restricted the end points to MI or revascularization and MI or death and observed no difference in the estimate for DES (DES AOR, 0.91; 95% CI, 0.81-1.02 and DES AOR, 0.90; 95% CI, 0.82-1.00 compared with BMS, respectively) (eTable in the Supplement).
This study assessing the risk of major adverse cardiac events after noncardiac surgery in patients with recent coronary stenting identified several factors, principally acuity of clinical presentation for surgery and several markers of advanced cardiac disease. Although the time from stent placement to surgery was associated with MACE, this was principally observed for surgery in the first 6 months after the stent procedure, whereas timing of surgery more than 6 months after the stent procedure was not significantly associated with MACE. While the data suggest that the risk of surgery after DES placement may stabilize earlier, the potential confounding and nonrandomized nature of this observational study does not allow for direct comparison of outcomes by stent type. Stent type was not significantly associated with MACE for surgeries more than 6 months after stent placement, and we did not observe an association between APT cessation with MACE.
Of the 600 000 coronary stent procedures performed annually, nearly 20% are followed by at least 1 surgical procedure in the ensuing 2 years.11- 17 The present findings suggest that underlying surgical and cardiac risk, rather than stent type, are the primary factors associated with perioperative MACE; that event rates stabilize by 6 months; and that APT continuation does not substantially mitigate risk. Accordingly, the current focus of the guidelines on differential timing recommendations by stent type may warrant reconsideration, and greater concentration may need to be placed on assessing and optimizing cardiac risk.
The antiproliferative properties of DES protect against neointimal hyperplasia and the subsequent in-stent restenosis, but this benefit results in delayed endothelialization of DES, compared with BMS, leading to increased risk for stent thrombosis.2 A meta-analysis of 4 randomized clinical trials showed an increased rate of stent thrombosis 1 year after the implantation of DES compared with BMS.25 In response to this concern, the 2007 revised ACC/AHA guidelines specifically emphasized both timing and antiplatelet management for patients with DES undergoing noncardiac surgery. These recommendations were based largely on limited evidence of case series reporting stent thrombosis in surgical patients and reports of stent thrombosis after dual APT cessation within 1 year of DES implantation.
The differential MACE rate based on timing of noncardiac surgery by stent type is supported by limited and conflicting evidence. A report from the CREDO-Kyoto registry on 1878 patients (17%) who underwent a noncardiac surgery within 2 years of stent placement observed an overall MACE rate of 3.2% and similar rates between BMS (3.5%) and DES (2.9%).25 Similarly, a study of the Ontario stent registry cohort with linked administrative data reported on 2725 patients undergoing surgery within 2 years of stent placement. They reported the optimal time of surgery as 46 to 180 days for BMS and after 180 days for DES, with the only statistically significant difference by stent type being higher MACE rate for DES when surgery was less than 45 days and for BMS when surgery was between 181 and 365 days.14 We observed higher MACE rates for BMS compared with DES, particularly in a window where it was thought safe to proceed with surgery for patients with BMS (45-180 days) but not DES.
These prior studies and the current analysis are observational, meaning that neither stent type nor surgery timing was randomized and other factors could be confounding the results. Considering the current findings in the context of the prior studies, we recommend future prospective studies to assess the safety of noncardiac surgery at 6 months after DES implantation. In addition, the findings challenge the current focus on stent type and timing of surgery as the primary decision points of perioperative risk assessment in patients with prior coronary stents. Additional cardiac risk factors of recent MI, higher rCRI, and recent CHF exacerbation warrant more attention in the algorithms for risk stratification in patients with stents.
The efficacy of APT in reducing perioperative ischemic cardiac events is established.26 However, the effectiveness of continued APT agents in reducing perioperative MACE events in patients with coronary stents is less clear. In the CREDO-Kyoto registry, 2398 patients had a surgical procedure within 3 years of stent implantation. They found that 30-day MACE rates were 4.9% for dual APT, 1.1% for single APT, and 2.3% for no APT, although the results were not significant.25 Other studies have also reported higher rates of MACE after surgery with continued dual APT compared with single APT.27,28 These observational studies, including the present study, are likely confounded by the fact that patients with the highest cardiac risk are most likely to both be taking dual APT and have it continued perioperatively. Nonetheless, matched-pair analysis did not find an association between continued APT and MACE. One potential explanation behind this finding may be that the anti-ischemic properties of APT are offset by a higher risk of bleeding. A prospective study of 103 patients undergoing noncardiac surgery reported a cardiac related mortality of 5% despite continued APT, and bleeding events occurred more frequently among those with MACE.29 Thus, bleeding events and their sequelae may be in the causal pathway of MACE and confound potential protective effects of continued APT.
Several considerations need to be given to the present findings. First, the study sample comprised primarily older male patients, thus limiting the generalizability to women or younger men. Second, the clinical decision-making factors that influenced stent selection were largely unavailable to us and limited the ability to account for them in the models. Accordingly, the results could be confounded by those factors. Third, many patients underwent more than 1 PCI procedure during the dates of the study cohort, which could result in misclassification bias for time from stent placement to surgery. However, based on these and others’ data, the PCI care episode closest to the surgery likely possesses the highest risk. Fourth, the surgical population by design is heterogeneous, with procedures ranging from minor outpatient to emergent inpatient operations. Although this improves the generalizability, it limits the ability to make recommendations regarding specific surgical populations or clinical scenarios. Fifth, we relied on administrative data to determine the end points, which could result in misclassification bias. Sixth, the case-control analysis of APT management had limited power to detect a true association. Seventh, the observational nature of the cohort and its inherent selection bias in stent type and surgery renders the findings as hypothesis generating only. As such, it suggests important areas for inquiry, ideally with randomized trials, to improve the evidence base supporting guideline recommendations.
Predominant risk factors for MACE after noncardiac surgery in patients with recent coronary stent implantation included nonelective surgical presentation and conditions associated with advanced ischemic cardiac disease. The time between coronary stent implantation and noncardiac surgery provided less explanatory importance. Stent type among those patients undergoing surgery more than 6 months after stent placement was not significantly associated with MACE. Complete APT cessation in the perioperative period was also not associated with MACE. Guidelines recommending prolonged delay and continued use of APT for patients with DES should be reevaluated.
Corresponding Author: Mary T. Hawn, MD, MPH, Department of Surgery, Section of Gastrointestinal Surgery, University of Alabama at Birmingham, 1922 Seventh Ave S, KB 428, Birmingham, AL 35294-0016 (firstname.lastname@example.org)
Published Online: October 7, 2013. doi:10.1001/jama.2013.278787.
Author Contributions: Ms Graham and Dr Richman had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Hawn, Itani, Henderson, Maddox.
Acquisition of data: Hawn, Graham, Henderson.
Analysis and interpretation of data: Hawn, Graham, Richman, Henderson, Maddox.
Drafting of the manuscript: Hawn, Graham, Itani, Maddox.
Critical revision of the manuscript for important intellectual content: Hawn, Richman, Itani, Henderson, Maddox.
Statistical analysis: Graham, Richman, Henderson.
Obtained funding: Hawn, Graham, Henderson, Maddox.
Administrative, technical, or material support: Graham.
Study supervision: Hawn, Maddox.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Itani reported having received research support from Merck and Cubist. No other disclosures were reported.
Funding/Support: This study is supported by a VA Health Services Research & Development grant (No. IIR 09-347). In addition, Drs Maddox and Richman are supported by VA Career Development Awards.
Role of the Sponsor: The sponsor had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. The VA National Surgery Office and the CART Program approved the manuscript for adherence to the data use agreements.
Disclaimer: The opinions expressed are those of the authors and not necessarily those of the Department of Veterans Affairs or the US government.
Additional Contributions: We acknowledge the VA Surgical Quality Data Use Group (SQDUG) for its role as scientific advisors and for the critical review of data use and analysis presented in this article. The advisors did not receive compensation for their contribution besides their salaries.
Correction: This article was corrected online January 8, 2014, for errors in Table 1.