Abstract
Objectives
Endovascular abdominal aortic aneurysm repair (EVAR) is increasingly used for emergent treatment of ruptured abdominal aortic aneurysm (rAAA). We sought to compare the perioperative and long-term mortality, procedure-related complications and rates of re-intervention of EVAR versus open aortic repair of rAAA in Medicare beneficiaries.
Methods
We examined perioperative and long-term mortality and complications after EVAR or open aortic repair performed for rAAA in all traditional Medicare beneficiaries discharged from a US hospital from 2001–2008. Patients were propensity score matched on baseline demographics, coexisting conditions, admission source, and hospital volume of rAAA repair and sensitivity analyses were performed to evaluate the impact of bias that might have resulted from unmeasured confounders
Results
Of 10,998 patients with repaired rAAA, 1126 underwent EVAR and 9872 underwent open repair. Propensity score matching yielded 1099 patient pairs. The average age was 78 years, and 72.4% were male. Perioperative mortality for EVAR and open repair were 33.8% and 47.7% respectively (p<0.001) and this difference persisted for more than four years. EVAR patients had higher rates of AAA-related reinterventions when compared with open repair patients (endovascular reintervention at 36 months 10.9% vs 1.5%, p<0.001), whereas open patients had more laparotomy related complications (incisional hernia repair at 36 months 1.8% vs. 6.2% p<0.001, all surgical complications at 36 months 4.4% vs. 9.1%, p<0.001). Use of EVAR for rAAA has increased from 6% of cases in 2001 to 31% of cases in 2008, while over the same time period overall 30-day mortality for admission for rAAA regardless of treatment has decreased from 55.8% to 50.9%.
Conclusions
EVAR for rAAA is associated with lower perioperative and long term mortality in Medicare beneficiaries. Increasing adoption of EVAR for rAAA is associated with an overall decrease in mortality of patients hospitalized for rAAA over the last decade.
INTRODUCTION
Despite better preventive practices and increasing rates of repair of intact abdominal aortic aneurysms (AAA) in older and higher risk populations1, ruptured abdominal aortic aneurysm (rAAA) continues to cause over 5,000 deaths annually in the United States.2,3 Autopsy data demonstrate that 50–70% of patients with ruptured AAA do not survive to hospital presentation.4 For those that do, the traditional treatment has been emergent open aortic repair, but mortality after open aortic repair remains over 40%.5–7 For intact aneurysms, endovascular aortic repair (EVAR) offers improved perioperative mortality, and speedier recovery versus open repair8–10 and EVAR has become the dominant treatment for intact AAA repair in the United States.11 Critically ill patients with ruptured AAA also may benefit from EVAR, but necessary preoperative imaging and specific anatomic requirements can make EVAR less well suited for emergent use. As of 2008, only 31% of rAAA repairs in the US were treated with EVAR while more than 85% of intact repairs were treated with EVAR.1,12
Successful use of EVAR for ruptured AAA was first reported in 1994. 13,14 Subsequent case series and observational studies suggest that for selected patients, EVAR offers improved mortality when compared to open repair.12,15–22 Conversely, small randomized controlled trials demonstrated no difference in perioperative mortality, 23,24 while other trials are ongoing.25,26 Over 76% of ruptured AAAs occur in those over age 65 enrolled in Medicare.4 Thus, experiences in Medicare provide the most comprehensive data on rAAA available. In this paper, we sought to compare the perioperative and long-term mortality and short- and long-term complications in patients receiving EVAR versus open repair for ruptured AAA in the Medicare population. We also examine trends in mortality for rAAA to estimate the overall impact of rising adoption of EVAR on survival after rAAA.
METHODS
Patients
We identified all Medicare beneficiaries age 67 or older who were admitted to a US hospital with a primary discharge diagnosis of ruptured abdominal aortic aneurysm (ICD-9 441.3) between 2001 and 2008. We excluded patients with concurrent diagnoses of thoracic aneurysm (441.1, 441.2), thoracoabdominal aneurysm (441.6 or 441.7) and aortic dissection (441.00–441.03) as well as those with procedural codes for repair of the thoracic aorta (38.35, 38.45, 39.73), and visceral or renal bypass (38.46, 39.24, 39.26).
In order to accurately identify ruptures as distinct from intact AAAs, we analyzed both hospital and physician claims and only included patients for whom the diagnosis was consistent across both sources (see Appendix Figure 1 for further explanation). Overall mortality rates were consistent with those reported in the literature when requiring both the hospital and physician code to indicate rupture. Treatment type was determined by examining ICD-9 and CPT procedure codes for open surgical repair (ICD-9 38.44, 39.25, CPT 35081, 35082, 35102, 35103, 35646) and for endovascular repair (ICD-9 39.71, CPT 34800, 34802, 34803, 34804, 34805). When hospital and physician claims conflicted regarding repair type (open versus endovascular), physician claims were given priority.10 Patients who had both EVAR and open repair during the same hospitalization were classified as EVAR patients that had conversion to open repair. All patients had had at least two years of prior Medicare experience that we used to control for coexisting conditions that might influence outcomes. We excluded beneficiaries enrolled in a Medicare Advantage health plan prior to or during the index hospitalization.
Creating Matched Cohorts
To control for the non-random assignment of patients to treatment groups, we created matched cohorts of patients after estimating logistic regression models predicting the likelihood of EVAR (propensity score). As explanatory variables, we used baseline demographic and coexisting conditions identified using inpatient and outpatient claims from the two years prior to but not including the index hospitalization.10,27 We also included as covariates measures of hospital volume of AAA repair, admission through the emergency department, and transfer from outside hospital and calendar year. We matched each beneficiary who underwent EVAR to the one beneficiary who underwent open repair with the closest estimated propensity to undergo EVAR. We chose one to one matching, as opposed to many to one matching, to maximize balance between treatment groups. To ensure close matches we required the estimated log-odds of endovascular repair between a patient who underwent endovascular repair and one who underwent open repair to be within 0.60 standard deviations. This value removes approximately 90% of the bias in estimates of effects due to differences in covariate distributions between “treatment” and comparison groups.28
These methods, however, are not able to account for the presence of unmeasured factors that might have influenced the treatment decision. For instance, if hemodynamically unstable patients were preferentially treated with open versus endovascular repair, our results could be biased in favor or EVAR. To address this issue, we performed an additional sensitivity analysis, and an analysis of temporal trends in repair type and mortality.
Outcomes
Mortality
Perioperative mortality was defined as death during the index hospitalization or within 30 days of the procedure. Long-term mortality included all deaths during the entire follow-up period. Dates of death were obtained from Medicare enrollment data through 12/31/10. Claims-based outcomes were censored at 12/31/08.
Perioperative Outcomes and Complications
Perioperative complications were assessed by examining diagnosis and procedure codes for the index hospitalization. Complications of interest included conversion of endovascular repair to open repair, reoperation for bleeding, tracheostomy, embolectomy, wound dehiscence, mesenteric ischemia, bowel resection, myocardial infarction, pneumonia, and renal failure. We also examined length of stay of patients who survived to hospital discharge as well as discharge disposition (home vs. institutional facility). A full listing of diagnosis codes used to identify complications is included in the Appendix.
Long term-Complications
We identified subsequent AAA-related hospitalizations with diagnosis and procedure codes related to abdominal aortic aneurysm, including open surgical reinteventions (open repair of the aneurysm, conversion from endovascular to open repair, repair of a graft–enteric fistula or graft infection, axillobifemoral bypass), and endovascular reinterventions (repeat endovascular repair, stent–graft extension, embolization, aortic or iliac angioplasty, graft thrombectomy) and laparotomy related complications, including non-surgical bowel obstruction, surgical bowel obstruction (obstruction related lysis of adhesions or bowel resection) and repair of an incisional hernia.
Statistical Analysis
We compared the clinical characteristics of the treatment cohorts using chi-square tests for categorical variables and t-tests for continuous variables other than time to event variables. Survival time, freedom from rupture, and time to early and late complications were examined using Kaplan-Meier life table methods. Differences between groups were tested with the log rank test. All analyses were performed in SAS 9.3.
Temporal Trends
To further explore if selection played a role in differences in observed mortality for EVAR vs. open repair, we examined temporal trends in overall repair rates, and 30-day mortality of EVAR and open repair. We identified all hospitalizations with a discharge diagnosis code of rAAA (441.3) using just hospital claims. EVAR patients were defined as those who had a concomitant code for an endovascular repair (ICD9-CM 38.44, 39.25), while open repair patients were defined as those who had a claim for open aortic repair (ICD-9 CM 39.71, 39.90). Patients with a claim for neither repair were considered to be unrepaired. If more hemodynamically stable patients were offered EVAR, and more unstable patients rates were preferentially offered open repair as EVAR was adopted over time, we would expect to see increasing 30-day mortality rates in those treated with open repair.
Sensitivity Analysis
To examine the impact of unmeasured confounders, we jointly modeled perioperative mortality (conditional on EVAR) and treatment choice (EVAR vs. open repair) based on methods described in O’Malley et al.29 In this approach, any unmeasured confounders are absorbed in the errors terms of the patient’s underlying propensity to (i) undergo EVAR and (ii) suffer peri-operative death. By examining the correlation of the error terms in (i) and (ii), we can estimate whether there is a relationship between unmeasured factors that make EVAR more likely and unmeasured factors that make mortality more likely. More details on this approach and the results of the sensitivity analysis can be found in the statistical appendix.
The study was approved by the Harvard Medical School Institutional Review Board.
RESULTS
We identified 10,998 patients who underwent repair for ruptured AAA from January 2001 to December 2008. Of these 9,872 patients underwent open repair and 1,126 underwent EVAR. Baseline characteristics of the patients and their coexisting conditions are listed in table 1. Prior to matching, patients receiving EVAR tended to be older (78.2 v. 77.2 years, p<0.001), have more medical diagnoses (e.g., 19.0% with CHF versus 12.2%, p<0.001,), and were more likely to have a pre-existing diagnosis of AAA (25.6% v. 16.4%, p<0.001). In contrast, patients receiving open repair were more likely to have been admitted through the emergency department (80.5% v. 63.2%, p<0.001), and were less likely have been transferred (2.4% v. 6.0%, p<0.001). Propensity score matching yielded 1099 patient pairs. With the exception of cardiac arrhythmias (16.3% v. 13.1%, p=.02) and history of peripheral vascular disease (16.3% v. 13.1%, p=.04), there were no remaining statistically significant differences between the populations.
Table 1.
Patient Characteristics
| Unmatched | Matched | |||||
|---|---|---|---|---|---|---|
|
| ||||||
| EVAR (N=1126) % | Open (N=9872) % | p | EVAR (N=1099) % | Open (N=1099) % | p | |
|
Demographic Characteristics
| ||||||
| Age (mean, SD) | 78.2 (6.6) | 77.2 (6.3) | <0.001 | 78.2 (6.6) | 78.2 (6.6) | 0.75 |
| Age 67–69 | 10.39 | 11.97 | <0.001 | 10.65 | 10.1 | 0.98 |
| Age 70–74 | 23.18 | 25.32 | 23.29 | 22.84 | ||
| Age 75–79 | 27.35 | 31.98 | 27.48 | 27.48 | ||
| Age 80–84 | 21.31 | 16.68 | 21.11 | 22.11 | ||
| Age 85+ | 17.76 | 14.04 | 17.47 | 17.47 | ||
| Male | 74.2 | 74.2 | 0.97 | 74.8 | 75.1 | 0.88 |
| Race | ||||||
| Black race | 6.3 | 3.8 | <0.001 | 6.1 | 6.6 | 0.60 |
| Other race | 3.1 | 2.0 | 0.01 | 2.8 | 2.3 | 0.42 |
| Source of Admission | ||||||
| Emergency Department | 63.2 | 80.5 | <0.001 | 64.2 | 63.4 | 0.72 |
| Outside Hospital Transfer | 6.0 | 2.4 | <0.001 | 5.3 | 5.4 | 0.92 |
|
| ||||||
|
Coexisting conditions
| ||||||
| Myocardial Infarction in past 6 months | 1.1 | 0.9 | 0.66 | 0.9 | 0.9 | 1 |
| Myocardial Infarction in past 7–24 months | 7.1 | 3.8 | <0.001 | 7.1 | 7.1 | 1 |
| Congestive Heart Failure | 19.0 | 12.2 | <0.001 | 18.4 | 16.7 | 0.31 |
| Cardiac Arrhythmias | 24.0 | 15.6 | <0.001 | 23.4 | 19.3 | 0.02 |
| Valvular Disease | 11.0 | 6.9 | <0.001 | 10.6 | 10.7 | 0.94 |
| Peripheral Vascular Disease | 16.7 | 11.4 | <0.001 | 16.3 | 13.1 | 0.04 |
| Hypertension | 61.1 | 54.4 | <0.001 | 60.7 | 60.2 | 0.83 |
| COPD | 31.0 | 25.8 | <0.001 | 30.8 | 30.2 | 0.78 |
| Uncomplicated DM | 16.2 | 11.8 | <0.001 | 15.6 | 15.8 | 0.86 |
| Complicated DM | 3.9 | 2.2 | <0.001 | 3.6 | 2.6 | 0.14 |
| Renal Disease | 8.2 | 5.4 | 0.02 | 8.2 | 8.7 | 0.65 |
| Metastatic Cancer | 2.7 | 0.9 | <0.001 | 2.6 | 1.8 | 0.24 |
| Solid tumor | 16.0 | 11.9 | <0.001 | 15.9 | 15.3 | 0.68 |
| Obesity | 2.0 | 1.6 | 0.29 | 2.0 | 1.9 | 0.88 |
| ESRD | 1.6 | 0.6 | <0.001 | 1.5 | 1 | 0.33 |
| Cerebrovascular disease | 12.7 | 10.6 | 0.03 | 12.4 | 12.0 | 0.79 |
| Intact AAA diagnosis | 25.6 | 16.4 | <0.001 | 25.1 | 22.3 | 0.12 |
Perioperative Outcomes
Perioperative outcomes are summarized in table 2. Perioperative mortality was 33.8% for patients who underwent EVAR versus 47.7% for patients undergoing open repair (p<0.001). Patients treated with EVAR suffered a lower rate of most post-operative complications including post operative pneumonia (28.5% for EVAR vs. 35.9% for open repair, p<0.001), acute renal failure (33.4% vs. 45.4%, p <0.001), respiratory failure requiring tracheostomy (4.6% vs. 9.9%, p < 0.001) and gastrointestinal complications including colon resection (4.4% vs 8.5%, p < 0.001) and mesenteric ischemia (7.6% vs. 14.7% p < 0.001). Several procedural complications also were less common in EVAR than in open repairs including embolectomy (3.6% vs. 6.3%, p = 0.003), and wound dehiscence (2.5% vs. 4.6%, p = 0.008). In contrast, procedure related hematoma was more common in patients with EVAR (8.0% vs 4.5%, p < 0.001) and conversion from EVAR to open repair occurred in 4.9% of patients within the index hospitalization or by 30 days.
Table 2.
Perioperative Outcomes
| EVAR (n=1099) | Open (n=1099) | p | ||||
|---|---|---|---|---|---|---|
|
Perioperative Mortality (%)
| ||||||
| All Ages | 33.8 | 47.7 | <0.001 | |||
| 67–69 years | 18.8 | 38.7 | <0.001 | |||
| 70–74 years | 27.7 | 41.0 | <0.001 | |||
| 75–79 years | 30.1 | 43.0 | <0.001 | |||
| 80–84 years | 39.2 | 53.5 | <0.001 | |||
| 85+ years | 50.0 | 61.5 | 0.02 | |||
|
Medical Complications (%)
| ||||||
| Myocardial Infarction | 16.7 | 19.2 | 0.13 | |||
| Pneumonia | 28.5 | 35.9 | <0.001 | |||
| Acute Renal Failure | 33.4 | 45.4 | <0.001 | |||
| Hemodialysis | 0.7 | 0.6 | 0.80 | |||
| Respiratory Failure/Tracheostomy | 4.6 | 9.9 | <0.001 | |||
| Venous Thromboembolism | 8.0 | 6.8 | 0.29 | |||
|
Surgical Complications (%)
| ||||||
| Conversion to Open Repair | 4.9 | |||||
| Reoperation for bleed | 2.3 | 3.0 | 0.24 | |||
| Embolectomy | 3.6 | 6.4 | 0.003 | |||
| Wound Dehiscence | 2.5 | 4.6 | 0.008 | |||
| Operative Site Hematoma | 8.0 | 4.5 | <0.001 | |||
| Gastrointestinal Bleed | 10.3 | 13.8 | 0.01 | |||
| Mesenteric Ischemia | 7.6 | 14.7 | <0.001 | |||
| Bowel Obstruction/Ileus | 12.7 | 17.0 | 0.005 | |||
| Colon Resection | 4.4 | 8.5 | <0.001 | |||
| Ostomy | 2.64 | 5.28 | 0.002 | |||
|
| ||||||
| Discharge to Home (%) | 62.8 | 40.7 | <0.001 | |||
| Length of stay of survivors to hospital discharge (median, IQR in days) | 7 (4–24) | 14 (9–23) | <0.001 | |||
Among patients who survived to discharge, the median length of stay for patients undergoing EVAR was 7 days (IQR 4–24 days), compared to 14 days (IQR 9–23 days) for patients undergoing open repair. 62.8 % of patients who underwent EVAR were discharged home as compared with 40.7% of patients who underwent open repair (p < 0.001).
Late outcomes
Long term survival for the entire cohort and stratified by age is shown in figure 1. EVAR was associated with a survival benefit that persisted for more than four years after intervention (p<0.001) in all age groups.
Figure 1.
Mortality of propensity matched patients undergoing endovascular and open repair of ruptured aortic aneurysm. Overall and stratified by age.
Open surgical and endovascular AAA-related re-interventions were more common after EVAR than open repair (table 3). Among EVAR patients, 1.9% had open re-intervention by 12 months, and 3.9% had open re-intervention by 36 months, versus 0.5% and 0.9% of open repair patients respectively (p = 0.002, log-rank test). In addition, 4.6% of EVAR patients had endovascular re-intervention by 12 months, and 10.9% had endovascular re-intervention by 36 months, versus 0.6% and 1.5% respectively for patients who had open repair (p < 0.001, log-rank test). Laparotomy related complications were more common in patients who had open aortic repair, including incisional hernia repair (1.8% for EVAR vs. 6.2% for open at 36 months, p<0.001), non-surgical bowel obstruction (18.7% for EVAR vs 35.8% for open at 36 months), and any obstruction-related surgical intervention (4.4% for EVAR vs. 9.1% for open at 36 months).
Table 3.
Late Outcomes
| 12 months | 36 months | p§ | |||
|---|---|---|---|---|---|
| EVAR | Open | EVAR | Open | ||
| Open surgical re-intervention+ | 1.9% | 0.5% | 3.9% | 0.9% | 0.002 |
| Endovascular re-interventionx | 4.6% | 0.6% | 10.9% | 1.5% | <0.001 |
| Laparotomy Related Complications | |||||
| Bowel Obstruction | |||||
| Non-surgical admission* | 7.3% | 15.4% | 18.7% | 35.8% | <0.001 |
| Surgical** | 0.7% | 1.8% | 2.5% | 3.0% | 0.157 |
| Incisional Hernia | 0.3% | 1.2% | 1.8% | 6.2% | <0.001 |
| Any surgical intervention*** | 1.0% | 3.0% | 4.4% | 9.1% | <0.001 |
- Conversion to open repair, open aneurysm repair, aortobifemoral bypass, axillofemoral or axillobifemoral bypass, repair of infected graft or graft–enteric fistula, thrombectomy, femoral–femoral bypass
- Repeat endovascular aneurysm repair, embolization, angioplasty(aortic or iliac), extension cuff
Hospital admission with diagnosis of bowel obstruction without lysis of adhesions or bowel resection
Lysis of adhesions or bowel resection performed for diagnosis of obstruction
Surgical intervention for bowel obstruction or incisional hernia
- late outcomes were analyzed using survival analysis, and p-values are those generated by the log rank test
Trends in rAAA mortality
Trends of EVAR usage and 30-day mortality by treatment type are presented in figure 2. Use of EVAR for rAAA increased from 6% of repair cases in 2001 (220 cases, 4.2% of total rAAA admissions) to 31% of repairs in 2008 (828 cases, 20% of total rAAA admissions). The number of untreated rAAA remained nearly constant, but the proportion of untreated rAAA increased from 27% to 38.7% from 2000 to 2004, and then decreased to 34.5% in 2008. Perioperative mortality for rAAA treated by EVAR decreased from 46% in 2001, to 27% in 2008, and mortality for rAAA treated by open repair decreased from 44.7% to 40%. Thirty-day mortality of rAAA without treatment remained approximately 80%. Overall mortality for rAAA admissions, whether treated or not, decreased from 55.8% to 50.9% over the time period 2001 to 2008.
Figure 2.
Trends in rAAA treatment, and 30 day mortality by treatment type 2001–2008
Sensitivity Analysis
As described in the statistical appendix, our analyses of unmeasured selection suggest that patients whose unobserved factors make them more likely to be treated with EVAR have a higher predicted mortality than other patients with the same observed predictors. This suggests that any unmeasured confounders were associated with an under-estimation rather than overestimation of treatment effect and would bias our findings against (rather than for) EVAR. (Statistical Appendix)
DISCUSSION
In this comprehensive analysis of Medicare patients, we found that endovascular repair of ruptured AAA was associated with lower perioperative and long-term mortality, fewer inhospital complications, and shorter length of stay. The long-term survival benefit of EVAR persisted for more than four years. Consistent with our prior work on intact AAA repair, patients undergoing endovascular repair had a higher likelihood of reintervention for AAA, but this was balanced by lower rates of laparotomy-related interventions. Over this time period, overall mortality of patients admitted to a hospital with a rAAA regardless of intervention has declined nearly 5%, while the proportion of repairs performed using EVAR has increased from 6% to 31%.
Because our study is observational, unmeasured selection remains of significant concern. Most important, it is possible that hemodynamically unstable patients are preferentially offered open repair, which would result in higher mortality for patients treated with open repair. We used three approaches to mitigate this concern. First, we created matched cohorts using propensity score models that used all available patient factors including age, sex, race, ethnicity, coexisting conditions, and hospital level factors such as AAA repair volume. Because patients transferred from outside hospitals as opposed to the emergency department are more likely to be hemodynamically stable and have a contained rupture, we also included an indicator variable for outside hospital transfer, as well as admission through the emergency department in our propensity model. We did observe that more patients who received EVAR had a prior diagnosis of intact AAA, were less commonly admitted through the emergency department, and more commonly transferred between hospitals before treatment. In our final propensity matched analysis, these variables were well balanced between groups (Table 1) which should eliminate bias associated with these conditions. Second, we performed a sensitivity analysis that simultaneously modeled the selection effect and outcomes.29 The results of this sensitivity analysis suggest that unobserved features that make EVAR more likely are modestly associated with higher mortality. Hence, this suggests that sicker patients may be selected for EVAR, thus biasing our results in favor of open repair.
Finally, we examined overall trends in rupture repair and mortality. In the decades preceding the introduction of EVAR, the rate of repair and mortality associated with rAAA repair was stable.6 Since the introduction of EVAR, however, overall mortality for rAAA hospitalizations with or without treatment has decreased.1 Over the same time period, the proportion of rAAA repairs performed by EVAR has increased, while the mortality for EVAR and open repair for rAAA has decreased. If improved mortality in EVAR patients was due to selection of healthier patients, then we would expect to see an increase in mortality for patients undergoing open repair as EVAR was adopted. Instead we see a decrease in mortality with open rAAA repair, consistent with sicker patients being offered EVAR. This finding is in contrast to previous work where we have shown that mortality after open intact AAA repair has remained stable over this same time period.10 Thus, it is unlikely that our results could be explained by more unstable patients being preferentially treated with open repair rather than EVAR.
We also examined trends in unoperated rAAAs. The improved outcomes with EVAR and open repair over time could be explained by an increasing proportion of patients being turned down for surgery. While we see an increase in the proportion of unrepaired cases of all rAAA admissions with a concomitant decrease in the proportion of open repairs from 2001 to 2003 (Figure 2),, after 2003 this trend reverses, and there has since been a steady decline in the percentage of rAAAs that were unrepaired. Also over this time period, the proportion of EVAR has increased, and the proportion of open repair has continued to decline. Hence the observed improved mortality for all ruptures, regardless of treatment, may be driven by improved outcomes in EVAR vs. open repair, and by the fact that adoption of EVAR is associated with an increasing proportion of rAAA receiving any treatment.
Our results also are consistent with reports in the clinical literature of use of EVAR for rupture patients. Some centers perform EVAR on patients in shock without preoperative imaging,17 and have optimized protocols to expedite the use of EVAR in critically ill patients18,19 It has also been argued that EVAR may be of most benefit to the most critically ill, as they are least likely to survive an open procedure.30 These authors describe a protocol of rapid transport to the operating room with percutaneous femoral access with an awake patient through which an aortic occlusion balloon may be inserted for rapid supraceliac aortic occlusion without opening the abdomen; thus avoiding the commonly observed loss of hemodynamic stability associated with release of hemoperitoneum. At this point the surgeon may proceed with EVAR if appropriate or open repair if EVAR is not possible based on anatomy. The adoption of such protocols and the multispecialty coordination and training may also help improve mortality with open repair as well.
One surprising finding is that in our cohort, patients with unoperated rAAA had a 30 day mortality of approximately 80%, in contrast to previous findings that over 98% of patients with unoperated rAAA die within 30 days.31 There are several possible explanations for this. One possibility is that some patients are coded as rAAA as a “rule out” diagnosis, and they in fact do not have rAAA. It is also possible that some ‘unoperated’ cases identified using hospital claims for our trends analysis received repair but were not correctly coded. Importantly for our analysis, we have no reason to suspect that coding practices changed over time, and hence our analyses of trends in repair type and mortality should be unaffected. Given the concern of diagnostic accuracy of hospital claims alone, for our primary analysis of mortality, early and late complications of EVAR and open repair, we used both physician and hospital claims to ensure a more homogenous cohort of rAAA patients (Appendix).
While our findings are consistent with prior studies of rAAA in Medicare patients20 our results extend earlier work in several ways. First, we include more recent data with procedures performed up to 2008. As the period 2005–2008 demonstrated the fastest growth of EVAR for rAAA, these more recent data better reflect current approaches. Second, we report rates of perioperative and postoperative complications, rates of reintervention, discharge disposition and length of stay, which have not been reported in the Medicare population. Third, we incorporate longer-term data on outcomes and show that the benefits of endovascular repair are durable for more than 4 years. This is longer than has been seen in trials for EVAR for intact AAA, which may indicate that there is a larger survival advantage for rAAA than for intact AAA using endovascular approaches. Fourth, our approach incorporates physician diagnoses in addition to hospital discharge data. When we compare hospital coding to physician coding, we find significant discrepancies (see Appendix Figure 1), suggesting substantial potential for misclassification of intact AAAs as ruptures. Although our approach likely excludes some true rupture cases from our analyses, by requiring consistent coding of rAAA across physician and hospital claims, we generated a cohort that is likely less contaminated by these incorrectly coded cases. As 80% of intact AAAs are now repaired using EVAR and given the substantially lower perioperative mortality seen with EVAR for intact AAA, inappropriately including intact AAAs in our cohort likely would bias our results in favor of EVAR. Our approach attempts to minimize the likelihood of this occurrence by using the strictest criteria possible to identify cases of ruptured AAA from administrative data.
Our results are in contrast to two randomized controlled trials that showed no mortality benefit for EVAR versus open repair for rAAA.23,24 In general, these trials were small (< 150 patients), and hence had limited power to detect the mortality differences seen in our work, and less power to detect the early and late complications we report. These trials also required pre-operative imaging to assess eligibility for endovascular repair, excluding patients too unstable to undergo CT scan, limiting generalizability of their results. Larger, more inclusive clinical trials are ongoing.25
There are several limitations to our work. Administrative data are subject to coding errors and variability and the lack of clinical data makes it impossible to validate the coding of rAAA using other sources such as imaging results. We did make efforts to minimize errors in initial diagnosis by requiring consistent physician and hospital claims, though this did exclude many potential cases (Appendix Figure 1). We also do not have data regarding aneurysm anatomy which is important in determining whether a patient can received EVAR, but we attempted to account for this by eliminating patients with renal or visceral bypass or involvement of the thoracic aorta.
In summary, EVAR is associated with improved perioperative and long-term mortality for ruptured AAA, shorter length of stay, and decreased laparotomy related complications. The use of EVAR for rAAA has increased from 2001 to 2008, while the overall mortality for patients hospitalized for rAAA has decreased from 55.8% to 50.9%. As centers continue to adopt EVAR, and current centers refine protocols for its emergent use for rAAA, there will be further opportunities to study the effects of EVAR on rAAA outcomes.
Acknowledgments
The opinions expressed do not necessarily represent the views or policy positions of the Centers for Medicare and Medicaid Services.
Funding:
MLS, AJO, and BEL were supported by NIH grants 5R01HL105453-02 and 1RC4MH092717-01 for comparative effectiveness research, REB was supported the NIH T32 Harvard-Longwood Research Training in Vascular Surgery grant HL007734
Statistical Appendix
Our primary sensitivity analysis was designed to test for the extent to which unmeasured differences in health status might be biasing our results. Although propensity score methods account for confounding by observed predictors they offer no reassurance against unobserved confounders. One approach to account for unmeasured confounding is to jointly model perioperative mortality conditional on EVAR and use of EVAR versus open repair. Because perioperative mortality and EVAR are dichotomous variables, the bivariate probit model1–5 may be used to account for unmeasured confounding while modeling the effects of treatment (EVAR versus open) and other observed predictors of perioperative mortality. This model is defined by a simultaneous equation system wherein an equation for treatment selection is linked to an equation for the outcome from treatment through their shared dependence on unmeasured variables. The direction and magnitude of the net effect of the unmeasured confounders is quantified by the extent to which the relationship between perioperative mortality and EVAR deviates from that predicted based on the observed predictors. The size and magnitude of this deviation is quantified in terms of a correlation coefficient. The further the correlation is from 0 the more that unmeasured selection effects are expected to yield biased estimates under a propensity score or other method reliant on the non-existence of unmeasured confounders. For a complete description of the bivariate probit model in the context of a comparison between EVAR and open repair, please refer to O’Malley et al. (2011)6.
Table A.
Parameter estimates of regression coefficients for bivariate probit model adjusting for clinical and non-clinical adjusters, reason and source of admission, and institutional volume for the same procedure that was performed on the patient
| Term | Perioperative equation | Endovascular Equation | ||||
|---|---|---|---|---|---|---|
|
| ||||||
| Estimate | t-value | p-value | Estimate | t-value | p-value | |
| Key predictors | ||||||
| Endovascular repair | −0.630 | −2.18 | 0.0292 | |||
| Proportion Endo | 2.064 | 21.040 | <.0001 | |||
| BC(Total volume)# | 0.058 | 1.890 | 0.059 | |||
| λ (Total Volume) | 0.352 | 2.560 | 0.010 | |||
| BC(Endo volume, Endo pats)# | −0.170 | −1.210 | 0.227 | |||
| BC(Open volume, Open pats)# | −0.042 | −2.540 | 0.011 | |||
| λ (Endo Volume, Endo pats) | −0.294 | −0.680 | 0.493 | |||
| λ (Open volume, Open pats) | 0.274 | 1.870 | 0.061 | |||
|
| ||||||
| Non-Clinical Adjusters | ||||||
| Procedure date | 0.008 | 10.040 | <.0001 | |||
| Procedure date (Endo pats) | 0.002 | 1.210 | 0.227 | |||
| Procedure date (Open pats) | −0.001 | −1.470 | 0.142 | |||
| Urgent Admission | 0.088 | 2.910 | 0.004 | −0.331 | −8.020 | <.0001 |
| Transfer | −0.062 | −0.870 | 0.384 | 0.259 | 2.860 | 0.004 |
|
| ||||||
| Patient characteristics | ||||||
| Intercept | −0.066 | −1.220 | 0.224 | −3.132 | −24.860 | <.0001 |
| Age 70–74 (baseline 67–69) | 0.158 | 4.340 | <.0001 | 0.029 | 0.490 | 0.625 |
| Age 75–79 (baseline 67–69) | 0.509 | 12.730 | <.0001 | 0.185 | 2.900 | 0.004 |
| Age 80 & over (baseline 67–69) | 0.766 | 10.300 | <.0001 | 0.369 | 3.530 | 0.000 |
| Male | −0.124 | −4.660 | <.0001 | 0.100 | 2.350 | 0.019 |
| Black | −0.299 | −4.990 | <.0001 | 0.291 | 3.510 | 0.000 |
| End stage renal disease (ESRD) | 0.342 | 2.680 | 0.007 | 0.254 | 1.540 | 0.123 |
| Chronic renal insufficiency | 0.203 | 3.740 | 0.000 | −0.059 | −0.760 | 0.446 |
| Coronary bypass surgery (CABG) | −0.378 | −3.360 | 0.001 | −0.352 | −1.500 | 0.132 |
| PTCA | −0.254 | −2.540 | 0.011 | 0.127 | 0.980 | 0.328 |
| CAD without procedure | 0.038 | 1.050 | 0.296 | 0.153 | 3.000 | 0.003 |
| Chronic heart failure (CHF) | 0.155 | 4.020 | <.0001 | 0.173 | 3.090 | 0.002 |
| COPD | 0.076 | 2.760 | 0.006 | 0.044 | 1.030 | 0.302 |
| Vascular disease | 0.102 | 3.250 | 0.001 | −0.028 | −0.580 | 0.564 |
| Prior AAA diagnosis | −0.002 | −0.070 | 0.942 | 0.117 | 2.400 | 0.016 |
The estimated selection effect (correlation between the unmeasured latent variables affecting perioperative mortality and likelihood of receiving EVAR) is 0.093 (t = 1.15, p = 0.250). #The Box-Cox transformation7 with parameter λ of x is given by x(λ) = (xλ −1)/λ if λ ≠ 0 or otherwise x(λ) = log(x).
The results of the bivariate probit analysis are presented in Table A; estimates (together with t-values and p-values) for the equation predicting the likelihood of perioperative mortality are presented in the first three columns to the right of the list of predictors while those for the equation predicting the likelihood of undergoing endovascular surgery are presented in the rightmost three columns. Those predictors that are particular to a given model (e.g., the effect of EVAR on mortality) only have estimates for that model.
The effect of EVAR (the effect of primary interest) is significant and negative (− 0.630, p = 0.0292) implying that the risk of perioperative mortality is lower for EVAR than for open, all else equal. Secondly, the selection effect correlation parameter is estimated to be 0.093 (p = 0.250) suggesting that individuals that are more likely to receive EVAR in unmeasured ways are more likely to suffer perioperative mortality. Therefore, the propensity score analysis that does not adjust for unmeasured confounding is more likely biased against finding a significant effect (as EVAR patients essentially have unmeasured risk factors that place them at greater risk for perioperative mortality) than finding a significant effect when none exists.
Due to the relative scarcity of EVAR observations effects estimated for it are less precise than those for open. For example, the effect of institutional EVAR volume has a larger negative coefficient than that for institutional open volume (−0.170 versus −0.042) but is less significant (p-value 0.227 versus 0.011). Both volume effects are consistent with past findings that the risk of perioperative mortality is lower at institutions with higher annual volumes of the same procedures. Further face validity in the bivariate probit model is seen from the fact that perioperative mortality is much higher for urgent cases (0.088, p = 0.004) and that the relative utilization of EVAR has increased rapidly over time (effect of 0.008 per day, p < 0.0001). The coherency of these findings with past observations by others offers further evidence that the estimated effects under this model are reliable.
Appendix Table 1.
Physician and Hospital Coding of ruptured and intact AAA, 2005–2008.
| Hospital Diagnosis | Physician Diagnosis | n | 30 day mortality | % open repairs | % of open repairs coded as ruptures |
|---|---|---|---|---|---|
| Rupture | Intact | 190 | 20% | 62% | 19% |
| Intact | Rupture | 2519 | 7% | 47% | 15% |
| Rupture | Rupture | 5351 | 45% | 82% | 87% |
Patients with consistent coding for rAAA in hospital and physician codes have a substantially higher 30 day mortality of 45%, which is consistent with prior reports regarding mortality of rAAA. Also a substantially higher proportion of repairs were performed as open repairs (82%), which is also consistent with prior reports regarding trends in rAAA repair. As separate codes exist for open repair of rAAA versus open repair of intact AAA, we examined the percentage of open rAAA repairs that had the procedure code for rupture repair. Among patient with consistent rupture diagnosis, 87% of open repairs were coded as rupture repairs, while 15–19% of patients with inconsistent codes were coded as rupture repairs. Taken together, it appears that patients with inconsistent coding often have intact AAA, as EVAR is the dominant treatment for intact AAA, and the mortality is much lower than for rAAA, and their inclusion into the cohort would bias the results to suggest a survival benefit for EVAR.
Appendix Table 2.
Outcome definitions
| Perioperative Outcomes | |
|---|---|
| Medical Complications
| |
| 410.00 | Acute Myocardial Infarction* |
| 481.00, 482.00 | Pneumonia* |
| 584.5, 584.9, 584.0 | Acute Renal Failure* |
| 90935, 90937, 90940, 36800, 36810, 36815, 36818, 36819, 36820, 36821, 36825, 36830, 36831, 36832, 36833, 36834 | Hemodialysis** |
| 415.1, 453.4 | Venous Thromboembolism* |
| Surgical Complications
| |
| 34830, 34831, 34832 | Conversion to Open** |
| 35860, 35840 | Reoperation for Bleed** |
| 31603, 31605, 31600 | Tracheostomy** |
| 34201, 34203 | Embolectomy** |
| 97602, 97597, 97598 | Wound Dehiscence** |
| 99812 | Hematoma** |
| 578 | GI Bleed* |
| 55700, 55790 | Mesenteric Ischemia** |
| 560 | Bowel Obstruction (Intestinal obstruction without mention of hernia)* |
| 44140, 44141, 44143, 44144, 44145, 44146, 44147, 44150, 44151, 44155, 44160, 44204, 44205, 44206, 44207, 44208, 44210, 44211, 44212, 44213, 44110 | Colon Resection** |
| Late Outcomes | |
|---|---|
| Open Surgical Reintervention**
| |
| 34830, 34831, 34832 | Conversion to Open Repair |
| 35082, 35082, 35092, 35103 | Open AAA Repair |
| 35907, 35870 | Graft Infection |
| 35654, 35621 | Aortobifemoral, axillofemoral or axillobifemoral bypass |
| 35875, 35876 | Thrombectomy |
| 35131, 35132 | Iliac Aneurysm Repair |
| 35661 | Femoral-femoral bypass |
| Endovascular Reintervention**
| |
| 34800, 34802, 34803, 34804, 34805, 0078T, 0080T, 0001T, 0002T | Endovascular AAA repair (redo) |
| 35472, 35473, | Angioplasty |
| 34825, 34826 | Cuff Extension |
| 37204 | Embolization |
| 34900, 75954 | Endovascular Iliac Aneurysm Repair |
| Laparotomy Related Complications
| |
| Bowel obstruction (non-surgical admission) * | |
| 560.1 | Paralytic ileus |
| 560.8 | Intestinal obstruction without mention of hernia |
| 560.81 | Intestinal or peritoneal adhesions with obstruction |
| 560.89 | Pseudo-obstruction or mural thickening causing obstruction |
| 560.9 | Unspecified intestinal obstruction |
| 552.21 | Incisional hernia with obstruction |
| Bowel Obstruction (Surgical) ** | |
| 44005 | Enterolysis (freeing of intestinal adhesion) |
| 44180 | Laparoscopy, surgical, enterolysis (freeing of intestinal adhesion) |
| 44202 | Laparoscopy, surgical; enterectomy; resection of small intestine |
| 44203 | … each additional small intestine resection |
| 44120 | Enterectomy, resection of small intestine; single resection |
| 44130 | Enteroenterostomy, anastomosis of intestine |
| 44186 | Laparoscopy, surgical; jejunostomy (eg, for decompression or feeding) |
| 44187 | … ileostomy or jejunostomy, non-tube |
| 44140 | Colectomy, partial; with anastomosis |
| 44141 | … with skin level cecostomy or colostomy |
| 44143 | … with end colostomy and closure of distal segment |
| 44144 | … with resection, with colostomy or ileostomy and creation of mucofistula |
| 44160 | Colectomy, partial, with removal of terminal ileum with ileocolostomy |
| 44204 | Laparoscopy, surgical; colectomy, partial, with anastomosis |
| 44205 | … with removal of terminal ileum with ileocolostomy |
| 44206 | … with end colostomy and closure of distal segment |
| 44213 | Laparoscopy, surgical; mobilization of splenic flexure with partial colectomy |
| 44188 | Laparoscopy, surgical, colostomy or skin level cecostomy |
| Incisional Hernia ** | |
| 49560 | Repair initial incisional or ventral hernia; reducible |
| 49561 | … incarcerated or strangulated |
| 49565 | Repair recurrent incisional or ventral hernia; reducible |
| 49566 | … incarcerated or strangulated |
| 49568 | Implantation of mesh for incisional or ventral hernia |
| 49654 | Laparoscopy, surgical, repair, incisional hernia; reducible |
| 49655 | … incarcerated or strangulated |
| 49656 | Laparoscopy, surgical, repair, recurrent incisional hernia; reducible |
| 49657 | … incarcerated or strangulated |
ICD-9 codes
CPT codes
Appendix Table 3.
Perioperative Outcomes in unmatched sample, by treatment
| EVAR (N=1126) | Open (N=9872) | p | |
|---|---|---|---|
|
Perioperative Mortality (%)
| |||
| All Ages | 33.8 | 50.2 | <.0001 |
| 67–69 years | 18.8 | 39.6 | <.0001 |
| 70–74 years | 27.6 | 42.2 | <.0001 |
| 75–79 years | 30.2 | 49.2 | <.0001 |
| 80–84 years | 38.8 | 58.0 | <.0001 |
| 85 + years | 50.5 | 67.0 | <.0001 |
|
Medical Complications (%)
| |||
| Myocardial Infarction | 16.9 | 15.7 | 0.3199 |
| Pneumonia | 28.7 | 31.7 | 0.0405 |
| Acute Renal Failure | 33.7 | 39.4 | 0.0002 |
| Hemodialysis | 0.7 | 0.6 | 0.7956 |
| Respiratory Failure/Tracheostomy | 4.5 | 8.5 | <.0001 |
| Venous Thromboembolism | 8.1 | 4.4 | <.0001 |
|
Surgical Complications (%)
| |||
| Conversion to Open Repair | 4.8 | ||
| Reoperation for bleeding | 2.3 | 2.5 | 0.7082 |
| Embolectomy | 3.6 | 6.6 | <.0001 |
| Wound Dehiscence | 2.5 | 3.8 | 0.0285 |
| Operative Site Hematoma | 8.2 | 3.9 | <.0001 |
| Gastrointestinal Bleed | 10.4 | 11.2 | 0.4169 |
| Mesenteric Ischemia | 6.4 | 7.5 | 0.192 |
| Bowel Obstruction/Ileus | 13.1 | 15.7 | 0.0182 |
| Colon Resection | 4.4 | 6.7 | <0.0001 |
| Ostomy | 2.8 | 4.2 | 0.018 |
|
| |||
| Discharge to Home | 62.3% | 45.3% | <.0001 |
| Length of Stay (Median, IQR) | 7 (4–13) | 13 (9–21) | <.0001 |
Appendix Figure 1.
Cohort formation and determination of treatment type
Appendix Figure 2.
Mortality of unmatched patients undergoing endovascular and open repair of ruptured aortic aneurysm. Overall and stratified by age.
- 1.Amemiya T. Bivariate Probit Analysis: Minimum Chi-Square Methods. Journal of the American Statistical Association. 1974 [Google Scholar]
- 2.Gitto L. Dialysis modality selection according to the medical care provided at dialysis units: An econometric analysis. Health Services and Outcomes Research Methodology. 2004;5:227–41. [Google Scholar]
- 3.Zeng F, O’Leary JF, Sloss EM, Lopez MS, Dhanani N, Melnick G. The effect of medicare health maintenance organizations on hospitalization rates for ambulatory care-sensitive conditions. Med Care. 2006;44(10):900–7. doi: 10.1097/01.mlr.0000220699.58684.68. [DOI] [PubMed] [Google Scholar]
- 4.Bhattacharya J, Goldman D, McCaffrey D. Estimating probit models with self-selected treatments. Stat Med. 2006;25(3):389–413. doi: 10.1002/sim.2226. [DOI] [PubMed] [Google Scholar]
- 5.Freedman DA, Sekhon JS. Endogeneity in Probit Response Models. Political Analysis. 2010;18(2):138. [Google Scholar]
- 6.O’Malley AJ, Cotterill P, Schermerhorn ML, Landon BE. Improving observational study estimates of treatment effects using joint modeling of selection effects and outcomes: the case of AAA repair. Med Care. 49(12):1126–32. doi: 10.1097/MLR.0b013e3182363d64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Box G, Cox D. An Analysis of Transformations. Journal of the Royal Statistical Society Series B (Methodological) 1964;26(2):211–52. [Google Scholar]
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
References
- 1.Schermerhorn ML, Bensley RP, Giles KA, Hurks R, O’malley AJ, Cotterill P, et al. Changes in abdominal aortic aneurysm rupture and short-term mortality, 1995–2008: a retrospective observational study. Ann Surg. 2012;256(4):651–8. doi: 10.1097/SLA.0b013e31826b4f91. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Giles KA, Pomposelli F, Hamdan A, Wyers M, Jhaveri A, Schermerhorn ML. Decrease in total aneurysm-related deaths in the era of endovascular aneurysm repair. J Vasc Surg. 2009;49(3):543–50. doi: 10.1016/j.jvs.2008.09.067. discussion 550–1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Kochanek KD, Xu J, Murphy SL, Miniño AM, Kung HC. Deaths: Final Data for 2009. Centers for Disease Control: National Vital Statistics Reports. 2011;59(4):1. [PubMed] [Google Scholar]
- 4.Bengtsson H, Bergqvist D. Ruptured abdominal aortic aneurysm: a population-based study. J Vasc Surg. 1993;18(1):74–80. doi: 10.1067/mva.1993.42107. [DOI] [PubMed] [Google Scholar]
- 5.Bown MJ, Sutton AJ, Bell PRF, Sayers RD. A meta-analysis of 50 years of ruptured abdominal aortic aneurysm repair. Br J Surg England. 2002;89(6):714–30. doi: 10.1046/j.1365-2168.2002.02122.x. [DOI] [PubMed] [Google Scholar]
- 6.Heller JA, Weinberg A, Arons R, Krishnasastry KV, Lyon RT, Deitch JS, et al. Two decades of abdominal aortic aneurysm repair: have we made any progress? J Vasc Surg. (2000) 2000;32(6):1091–100. doi: 10.1067/mva.2000.111691. [DOI] [PubMed] [Google Scholar]
- 7.Visser P, Akkersdijk GJM, Blankensteijn JD. In-hospital operative mortality of ruptured abdominal aortic aneurysm: a population-based analysis of 5593 patients in The Netherlands over a 10-year period. Eur J Vasc Endovas. 2005;30(4):359–64. doi: 10.1016/j.ejvs.2005.05.005. [DOI] [PubMed] [Google Scholar]
- 8.Prinssen M, Verhoeven ELG, Buth J, Cuypers PWM, van Sambeek MRHM, Balm R, et al. A randomized trial comparing conventional and endovascular repair of abdominal aortic aneurysms. N Engl J Med. 2004;351(16):1607–18. doi: 10.1056/NEJMoa042002. [DOI] [PubMed] [Google Scholar]
- 9.Greenhalgh RM, Brown LC, Kwong GP, Powell JT, Thompson SG. Comparison of endovascular aneurysm repair with open repair in patients with abdominal aortic aneurysm (EVAR trial 1), 30-day operative mortality results: randomised controlled trial. Lancet. 2004;364:843–8. doi: 10.1016/S0140-6736(04)16979-1. [DOI] [PubMed] [Google Scholar]
- 10.Schermerhorn ML, O’Malley AJ, Jhaveri A, Cotterill P, Pomposelli F, Landon BE. Endovascular vs. open repair of abdominal aortic aneurysms in the Medicare population. N Engl J Med. 2008;358(5):464–74. doi: 10.1056/NEJMoa0707348. [DOI] [PubMed] [Google Scholar]
- 11.Landon BE, O’Malley AJ, Giles K, Cotterill P, Schermerhorn ML. Volume-Outcome Relationships and Abdominal Aortic Aneurysm Repair. Circulation. 2010;122(13):1290–7. doi: 10.1161/CIRCULATIONAHA.110.949172. [DOI] [PubMed] [Google Scholar]
- 12.Mcphee J, Eslami MH, Arous EJ, Messina LM, Schanzer A. Endovascular treatment of ruptured abdominal aortic aneurysms in the United States (2001–2006): a significant survival benefit over open repair is independently associated with increased institutional volume. J Vasc Surg. 2010;49(4):817–26. doi: 10.1016/j.jvs.2008.11.002. [DOI] [PubMed] [Google Scholar]
- 13.Marin ML, Veith FJ, Cynamon J, Sanchez LA, Lyon RT, Levine BA, et al. Initial experience with transluminally placed endovascular grafts for the treatment of complex vascular lesions. Ann Surg. 1995;222(4):449–65. doi: 10.1097/00000658-199522240-00004. discussion 465–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Yusuf SW, Whitaker SC, Chuter TA, Wenham PW, Hopkinson BR. Emergency endovascular repair of leaking aortic aneurysm. Lancet. 1994;344(8937):1645. doi: 10.1016/s0140-6736(94)90443-x. [DOI] [PubMed] [Google Scholar]
- 15.Alsac JM, Desgranges P, Kobeiter H, Becquemin JP. Emergency endovascular repair for ruptured abdominal aortic aneurysms: feasibility and comparison of early results with conventional open repair. Eur J Vasc Endovasc. 2005;30(6):632–9. doi: 10.1016/j.ejvs.2005.06.010. [DOI] [PubMed] [Google Scholar]
- 16.Najjar SF, Mueller KH, Ujiki MB, Morasch MD, Matsumura JS, Eskandari MK. Percutaneous endovascular repair of ruptured abdominal aortic aneurysms. Arch Surg. 2007;142(11):1049–52. doi: 10.1001/archsurg.142.11.1049. [DOI] [PubMed] [Google Scholar]
- 17.Veith FJ, Lachat M, Mayer D, Malina M, Holst J, Mehta M, et al. Collected world and single center experience with endovascular treatment of ruptured abdominal aortic aneurysms. Ann Surg. 2009;250(5):818–24. doi: 10.1097/SLA.0b013e3181bdd7f5. [DOI] [PubMed] [Google Scholar]
- 18.Mehta M, Taggert J, Darling RC, III, Chang BB, Kreienberg PB, Paty PSK, et al. Establishing a protocol for endovascular treatment of ruptured abdominal aortic aneurysms: Outcomes of a prospective analysis. J Vasc Surg. 2006;44(1):1–8. doi: 10.1016/j.jvs.2006.02.057. discussion 8. [DOI] [PubMed] [Google Scholar]
- 19.Nedeau AE, Pomposelli FB, Hamdan AD, Wyers MC, Hsu R, Sachs T, et al. Endovascular vs open repair for ruptured abdominal aortic aneurysm. J Vasc Surg. 2012 May 23;55(4):1001–7. doi: 10.1016/j.jvs.2011.12.067. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Egorova N, Giacovelli J, Greco G, Gelijns A, Kent CK, McKinsey JF. National outcomes for the treatment of ruptured abdominal aortic aneurysm: comparison of open versus endovascular repairs. J Vasc Surg. 2008;48(5):1092–100. 1100.e1–2. doi: 10.1016/j.jvs.2008.06.036. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Holt PJE, Karthikesalingam A, Poloniecki JD, Hinchliffe RJ, Loftus IM, Thompson MM. Propensity scored analysis of outcomes after ruptured abdominal aortic aneurysm. Br J Surg. 2010;97(4):496–503. doi: 10.1002/bjs.6911. [DOI] [PubMed] [Google Scholar]
- 22.Giles KA, Hamdan AD, Pomposelli FB, Wyers MC, Dahlberg SE, Schermerhorn ML. Population-based outcomes following endovascular and open repair of ruptured abdominal aortic aneurysms. J Endovasc Ther. 2009 Oct;16(5):554–64. doi: 10.1583/09-2743.1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Hinchliffe RJ, Bruijstens L, MacSweeney STR, Braithwaite BD. A randomised trial of endovascular and open surgery for ruptured abdominal aortic aneurysm - results of a pilot study and lessons learned for future studies. Eur J Vasc Endovasc. 2006;32(5):506–13. doi: 10.1016/j.ejvs.2006.05.016. discussion514–5. [DOI] [PubMed] [Google Scholar]
- 24.Reimerink JJ, Hoornweg LL, Vahl AC, Wisselink W, van den Broek TAA, Legemate DA, et al. Endovascular Repair Versus Open Repair of Ruptured Abdominal Aortic Aneurysms: A Multicenter Randomized Controlled Trial. Ann Surg. 2013;258(2):248–56. doi: 10.1097/SLA.0b013e31828d4b76. [DOI] [PubMed] [Google Scholar]
- 25.Powell JT. Time to IMPROVE the management of ruptured abdominal aortic aneurysm: IMPROVE trialists. Eur J Vasc Endovasc. 2009;38(2):237–8. doi: 10.1016/j.ejvs.2009.04.002. [DOI] [PubMed] [Google Scholar]
- 26.Desgranges P, Kobeiter H, Castier Y, Sénéchal M, Majewski M, Krimi A. The Endovasculaire vs Chirurgie dans les Anévrysmes Rompus PROTOCOL trial update. J Vasc Surg. 2010;51(1):267–70. doi: 10.1016/j.jvs.2009.10.128. [DOI] [PubMed] [Google Scholar]
- 27.Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care. 1998;36(1):8–27. doi: 10.1097/00005650-199801000-00004. [DOI] [PubMed] [Google Scholar]
- 28.Cochran W. Controlling Bias In Observational Studies: A Review. Sankhyā: The Indian Journal of Statistics. 1973;35(4):427–429. Series A. [Google Scholar]
- 29.O’Malley AJ, Cotterill P, Schermerhorn ML, Landon BE. Improving observational study estimates of treatment effects using joint modeling of selection effects and outcomes: the case of AAA repair. Med Care. 2011;49(12):1126–32. doi: 10.1097/MLR.0b013e3182363d64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Veith FJ, Powell JT, Hinchliffe RJ. Is a randomized trial necessary to determine whether endovascular repair is the preferred management strategy in patients with ruptured abdominal aortic aneurysms? J Vasc Surg. 2010;52(4):1087–93. doi: 10.1016/j.jvs.2010.05.142. [DOI] [PubMed] [Google Scholar]
- 31.Lindholt JS, Søgaard R, Laustsen J. Prognosis of ruptured abdominal aortic aneurysms in Denmark from 1994–2008. Clin Epidemiol. 2012;4:111–3. doi: 10.2147/CLEP.S31098. [DOI] [PMC free article] [PubMed] [Google Scholar]