The incidence and factors associated with graft infection after aortic aneurysm repair
Article Outline
Objectives
The reported rate of abdominal aortic graft infections (AGIs) is low, but its incidence and associated factors have not been evaluated on a population level. We hypothesized that AGI occurs more often in patients with periprocedural nosocomial infections and less often after endovascular aneurysm repair (EVAR).
Methods
A retrospective cohort study was done of all patients undergoing abdominal aortic aneurysm (AAA) repair (1987-2005) in Washington State by using the Comprehensive Hospital Abstract Reporting System (CHARS) data. Nosocomial infection was defined as one or more of pneumonia, urinary tract infections, blood stream septicemia, or surgical site infection at the index admission. Readmissions and reintervention for graft infections defined AGIs excluding the diagnostic code of renal failure or those who appeared to have dialysis grafts.
Results
Between 1987 and 2005, 13,902 patients (mean age, 71.3 ± 8.8 years; 90.8% men) underwent AAA repair (12,626 open, 1276 EVAR). The cumulative rate of AGIs in the cohort was 0.44%. The 2-year rate of AGI was 0.19% among open vs 0.16% in EVAR (P = .75) and 0.2% in both elective and nonelective patients. Open procedures had greater rates of perioperative pneumonia (11.1% vs 2.4%, P < .001), blood stream septicemia (1.6% vs 0.7%, P < .01), and surgical site infection (.5% vs 0%, P < .012) compared with EVAR. When individually analyzed, blood stream septicemia (.93% vs 18%, P = .014) and surgical site infection (1.61% vs 0.19%, P = .01) were significantly associated with AGIs. The median time to AGI was 3.0 years, and AGI presented sooner (≤1.4 years) if nosocomial infection occurred at the index admission. This risk of developing AGI after open repair was highest in the first postoperative year (32% of all AGI occurred in year 1). In an adjusted model, blood stream septicemia was significantly associated with AGI (odds ratio, 4.2; 95% confidence interval, 1.5-11.8)
Conclusions
The incidence of AGI was low, presented most commonly in the first postoperative year, and was similar among patients undergoing open AAA repair and EAVR. Patients with nosocomial infection had an earlier onset of AGI. The 2-year rate of AGI was significantly higher in patients who had blood stream septicemia and surgical site infection in the periprocedural hospitalization. These data may be helpful in directing surveillance programs for AIG.
The incidence of prosthetic aortic graft infections (AGI) has been reported to range between 0.6% and 3%.1, 2, 3 Aortic graft infection is a serious and life-threatening condition with a mortality rate of 20% to 40%, an amputation rate as high as 11%, and a reinfection rate of approximately 18%.4, 5, 6 Factors associated with AGI are not clearly understood and require further delineation. Understanding predictive factors may be important in developing surveillance programs for AGI.
Most studies of AGI are single-center evaluations with small numbers of patients and incomplete data on patient characteristics.4, 7, 8, 9 The use of a population-level database may help define the incidence of AGI, evaluate trends over time, and identify the factors contributing to AGI. Aortic graft infection may be caused by contamination during surgery or infectious processes after surgery, or both, but this has yet to be determined. We hypothesized that AGI is associated with nosocomial infections that occur during the initial hospitalization. We also hypothesized that AGI occurs less frequently with endovascular repair (EVAR) and more commonly with emergency repair.
Methods
Study design
A retrospective cohort study was conducted using a statewide hospital administrative discharge database. Records of inpatient hospitalizations between 1987 and 2005 were evaluated to assess the incidence of aortic graft infections.
Data source
Data were obtained from the Washington State Comprehensive Hospital Abstract Reporting System (CHARS). This data set is derived from all public and private hospitals in the state, excluding Veterans Affairs and United States military hospitals excluded. It contains demographic variables, admission and discharge administrative details, payer status, International Classification of Diseases, Ninth Revision (ICD-9) procedure and diagnosis codes, and hospital identifiers. CHARS also allows for tracking of subsequent hospitalizations and linkage to vital statistics records. The data are maintained by the Washington State Department of Health, Office of Hospital and Patient Data Systems. The reporting is mandatory. Data are currently collected using the Uniform Billing (UB) 92 format (previously, UB82). The CHARS system is designed to accommodate data elements from the Medicare provider-billing file. The reporting is done by the hospitals on the basis of information in the patient charts. The CHARS data are entered by hospital coders according to information in the charts. This study was granted an exception by agreement of the University of Washington Human Subject Review Committee and the Washington State Department of Health. The data set includes only anonymous data and is considered within the public domain.
Subjects
CHARS records from 1987 to 2005 were searched to identify all cases of abdominal aortic surgery indicated by the ICD-9 procedure. Cases included in the cohort were patients aged ≥18 years who had the ICD-9 code for abdominal aortic surgical procedures. Patients were categorized by ICD-9 diagnosis codes for repair of the abdominal aortic aneurysm from procedure codes as repairs with open and endovascular techniques. Classification as elective vs nonelective (which included “urgent” and “emergency”) was based on the CHARS variable for admission type. For this study, any cases that were defined as urgent were placed into the nonelective category.
Variable definitions
Nosocomial infection (NI) was defined as an infection occurring in a patient in a hospital or other health care facility in whom it was not present or incubating at the time of admission; or the residual of an infection acquired during a previous admission. We evaluated diagnostic codes for NI, which were postoperative pneumonia (997.3, 482.x), urinary tract infections (599.0, V13.02), blood stream septicemia (BSI; 038.x), and surgical site infection (SSI; 998.5, 998.51, 998.59), or a combination of these, at the index admission.
To identify patients with AGI, code 996.62 was used to all identify patients with many types of graft infection. This code includes infection and inflammatory reaction as a result of any internal prosthetic device, implant, or graft due to a vascular implant. This code was then combined with appropriate aortic resections associated with revision of AGI codes, including 38.34, resection of vessel with anastomosis abdominal aorta; 38.44, resection of vessel with replacement abdominal aorta; 38.64, other excision of vessels abdominal aorta.
All other patients were excluded, and our study cohort reflects only patients undergoing aortic resection with an associated code of infected graft at the same readmission. To further assure that we did not overestimate, all patients with the code for dialysis renal failure preoperative and postoperatively were excluded (v56.0) to remove any possibility of dialysis graft infections in the cohort.
A modified version of the Charlson Comorbidity Index10 was calculated for each patient. The Charlson measure, which has been adapted for use with administrative claims, is a weighted index of comorbidity status based on ICD-9 codes from inpatient records. Scores range from 0 to 3, where 0 indicates the absence of comorbid conditions and a score >0 indicates the presence of one or more comorbid conditions.
Classification as elective vs nonelective (which included “urgent” and “emergency”) was based on the CHARS variable for admission type. For this study, any cases that were defined as urgent were placed into the nonelective category.
Outcome metrics included in-hospital death, readmission, reintervention, and 30-day mortality. Readmissions classified by the presence of diagnostic codes for vascular graft infections combined with procedures associated with treatment of an infected aortic graft (such as extra-anatomic bypass or redo aortic surgery) were identified. In-hospital death is directly coded as a discharge status for the index admission. The 30-day mortality was defined as “all-cause” death ≤30 days of discharge as ascertained from Washington State Vital Records.
Statistical analysis
Descriptive statistics were calculated for the entire cohort and subgroups. Categoric variables were compared using Pearson χ2 statistic, and continuous variables were evaluated using analysis of variance. Logistic regression was used to create a 2-year model to define risk of AIG. The adjusted model contains all covariates and includes clustering by hospital. The cumulative hazard of readmission with graft infection was estimated using Nelson-Aalen methods. The Nelson-Aalen estimator is a nonparametric estimator of the cumulative hazard function based on a sample that is subject to right censoring. The cumulative hazard is estimated by summing the hazard of, in this case, graft infection over time. This time-to-event analysis was used to account for the variable follow-up time among subject in the cohort. Statistical analysis was performed using Stata 9 statistical software (Stata Corp, College Station, Tex).
As well, to account for the introduction of the code for EVAR in 2000, the 2-year rates of infection were used when open surgery was compared with EVAR. This was done to insure that there was a fair comparison of the two treatment methods and that the incidence of AGI was not skewed by the large open experience reported in the data.
Results
Demographic information
Between 1987 and 2005, 13,902 patients (mean age, 71.3 ± 8.8 years; 90.8% men) underwent abdominal aortic surgery; of these, 12,626 had open repair and 1276 had EVAR (Table I). Patients undergoing EVAR had higher Charlson scores than those having open aortic surgery (P < .001; Table II). A total of 9835 of the cases in the study period were elective procedures (70.8%).
Table I. Cohort characteristics
| Total | Nonelective | Elective | P | |
|---|---|---|---|---|
| Subjects, No. | 13,902 | 4067 | 9835 | |
| Age, mean years | 71.3 | 71.7 | 71.1 | .001 |
| Male, % | 79.3 | 77.8 | 80.0 | .003 |
| Charlson Comorbidity Index category, % | <.001 | |||
 0 | 1.9 | 2.2 | 1.4 | |
| 1 | 58.3 | 61.6 | 56.9 | |
| 2 | 32.1 | 29.2 | 33.4 | |
| 3+ | 7.7 | 7.0 | 8.0 | |
| Endovascular repair, % | 9.2 | 2.9 | 11.8 | <.001 |
| Any nosocomial infection, % | 13.1 | 18.6 | 10.8 | <.001 |
| Pneumonia | 10.3 | 13.4 | 9.1 | <.001 |
| Urinary tract infection | 1.6 | 2.8 | 1.2 | <.001 |
| Blood stream septicemia | 1.6 | 3.2 | 0.9 | <.001 |
| Surgical site infection | 0.4 | 0.8 | 0.3 | <.001 |
| Length of stay, mean days | 9.1 | 11.7 | 8.0 | <.001 |
| Discharge status, % | <.001 | |||
| Home | 79.3 | 58.6 | 87.8 | |
| Care facility | 10.9 | 16.8 | 8.5 | |
| In-hospital death | 9.8 | 24.6 | 3.7 | |
| Readmission with graft infection, % | 0.4 | 0.4 | 0.5 | .6 |
| Graft infection ≤2 years | 0.2 | 0.2 | 0.2 | .6 |
| Days to graft infection, median | 937 | 540 | 1330 | .1 |
| Graft infection rate/1000 person-years | 0.7 | 0.7 | 0.7 | .5 |
Table II. Cohort characteristic by procedure type
| Characteristic | Open | EVAR | P |
|---|---|---|---|
| Subjects, No. | 12,626 | 1276 | |
| Age, mean years | 71.0 | 73.5 | <.001 |
| Male, % | 79.9 | 83.4 | <.001 |
| Charlson Comorbidity Index category, % | <.001 | ||
| 0 | 2 | 1.1 | |
| 59.2 | 49.1 | ||
| 2 | 31.5 | 38.4 | |
| 3+ | 7.4 | 11.4 | |
| Elective admission, % | 68.7 | 90.8 | <.001 |
| Any nosocomial infection, % | 14.0 | 4.0 | <.001 |
| Pneumonia | 11.1 | 2.4 | <.001 |
| Urinary tract infection | 1.7 | 1.2 | .2 |
| Blood stream septicemia | 1.6 | 0.7 | .01 |
| Surgical site infection | 0.5 | 0.0 | .01 |
| Length of stay, mean days | 9.6 | 3.6 | |
| Discharge status, % | <.001 | ||
| Home | 78.1 | 90.7 | |
| Care facility | 11.3 | 7.0 | |
| In-hospital death | 10.6 | 2.3 | |
| Readmission with graft infection, % | 0.4 | 0.2 | .3 |
| Graft infection ≤2 years, % | 0.2 | 0.2 | .8 |
| Days to graft infection, median | 1100 | 684 | |
| Graft infection rate/1000 person-years | 0.7 | 1.6 | .07 |
Outcomes
The in-hospital mortality rate for the entire cohort was 9.81%, with 2.35% for EVAR vs 10.6% for open surgery (P < .001). The 30-day mortality was 3.8% for elective aortic surgery vs 25.0% for nonelective aortic surgery. The cumulative rate of AIG in the cohort was 0.44%. The in-hospital mortality rate at the first readmission for those identified with AGI was 18%, and the overall 1-year mortality rate for those identified to have AGI was 28%.
Group comparison
Patients undergoing aortic aneurysm repair were compared evaluating elective or nonelective repair (Table I), open and EVAR procedures (Table II), NI at the index visit (Table III), and by the development of AGI (Table IV). Nonelective patients were more likely male and significantly more likely to develop NIs (18.6 vs 10.8, P < .001). Nonelective patients had longer lengths of stay (11.7 vs 8.0, P < .001) and were less likely to be discharged home (58.6% vs 87.8%, P < .001). The rate of readmission with a graft infection and the 2-year rate of graft infection were similar (0.2% vs 0.2%, P = .6).
Table III. Cohort characteristic by index nosocomial infection
| Characteristic | Nosocomial infection | P | |
|---|---|---|---|
| No | Yes | ||
| Subjects, No. | 12,080 | 1822 | |
| Age, mean years | 71.1 | 72.4 | <.001 |
| Male, % | 80 | 75.1 | <.001 |
| Charlson Comorbidity Index category, % | .005 | ||
| 0 | 1.9 | 1.7 | |
| 1 | 58.1 | 59.5 | |
| 2 | 32 | 33.2 | |
| 3+ | 8 | 5.7 | |
| Elective admission, % | 72.6 | 58.4 | <.001 |
| Length of stay, mean days | 8 | 16 | <.001 |
| Discharge status, % | <.001 | ||
| Home | 81.8 | 62.7 | |
| Care facility | 9.2 | 22.8 | |
| In-hospital Death | 9.1 | 14.5 | |
| Readmission with graft infection, % | 0.4 | 0.6 | .4 |
| Graft infection ≤2 years | 0.2 | 0.3 | .2 |
| Days to graft infection, median | 1100 | 511.5 | .3 |
| Graft infection rate/1000 person-years | 0.7 | 0.9 | .2 |
Table IV. Cohort characteristic by graft infection (readmission)
| Characteristic | Aortic graft infection | P | |
|---|---|---|---|
| No | Yes | ||
| Subjects, No. | 13841 | 61 | |
| Male, % | 79.3 | 82.0 | .6 |
| CCI category, % | .008 | ||
| 0 | 1.9 | 6.6 | |
| 1 | 58.3 | 59.0 | |
| 2 | 32.1 | 34.4 | |
| 3+ | 7.8 | 0.0 | |
| Elective admission, % | 70.7 | 73.8 | .6 |
| Endovascular repair, % | 9.2 | 6.7 | .5 |
| Length of stay, mean days | 9.1 | 11.5 | .03 |
| Discharge status, % | .04 | ||
| Home | 79.2 | 88.5 | |
| Care facility | 10.9 | 11.5 | |
| In-hospital death | 9.9 | 0.0 | |
Patients undergoing open procedures (Table II) were more often younger, healthier, more likely to develop NIs, and less likely to be discharged to home. Patients having open repair and EVAR had similar 2-year rates of graft infection and infection rates (0.2% vs 0.2%, P = .8). When compared using 1000 person-years, however, the rate of infection for EVAR was nearly half (0.7% vs 1.6%, P = .07), although this was not a significant difference. Patients who had open aortic procedures had greater rates of pneumonia (11.1% vs 2.4%, P < .001), BSI (1.6% vs 0.7%, P < .01), and SSI (.5% vs 0%, P < .012) compared with EVAR.
Patients who developed NIs (Table III) at their index visit were more likely to have longer index hospitalizations (8 days vs 16 days, P < .001) and less likely to have had elective surgery (72.6 vs 58.4, P < .001). Patients who developed a NI at the index admission had shorter times to developing AIG (511 vs 1100 days). Most (72.6%) of the cases without development of NI were elective cases compared with patients developing a NI, of which 58.4% were elective (P < .001).
The 2-year rate of AGI was 0.19% among open vs 0.16% in EVAR (P = .75) and 0.2% in both elective and nonelective patients. The 2-year rate of AGI was higher, although not significant, after any index NI (0.33% vs 0.17%, P = .16) or pneumonia (0.35% vs 0.18%, P = .16). Blood stream septicemia (0.93% vs 0.18%, P = .014) and SSI (1.61% vs 0.19%, P = .01) were significantly associated with the development of AGI. The median time to AGI presented earlier if a perioperative NI occurred (1100 vs 511, P = .3). This risk of developing AGI after open repair was highest in the first postoperative year, with 32% of all graft infections in the open cohort presenting in year 1. Of interest was that no 1-year AGI was seen in the EVAR group. Odds ratios (ORs) for AGI within 2 years showed that after adjustment, BSI was significantly associated with AGI (OR, 4.2; 95% confidence interval (CI), 1.5-11.8; Table V). The Fig demonstrates the difference in the cumulative incidence of AGI in patients undergoing open repair and EVAR (log-rank P = .34).
Table V. Odds ratios for graft infection ≤2 years
| Covariates | Unadjusted OR | 95% CI | Adjusted OR⁎ | 95% CI |
|---|---|---|---|---|
| EVAR index procedure | 0.8 | 0.2-3.3 | 0.9 | 0.2-3.3 |
| Age 65+ | 1.1 | 0.4-2.9 | 1.1 | 0.4-3.3 |
| CCI category (1+) | 0.5 | 0.1-3.7 | 0.5 | 0.1-4.8 |
| Income above median | 1.7 | 0.8-3.8 | 1.7 | 0.9-3.3 |
| Elective admission | 0.8 | 0.4-1.8 | 0.9 | 0.5-1.8 |
| High-volume hospital | 1.1 | 0.3-4.6 | 1.0 | 0.3-3.8 |
| Any nosocomial infection | 1.9 | 0.7-4.7 | ||
| Pneumonia | 2.0 | 0.7-5.2 | 1.7 | 0.8-3.9 |
| Bloodstream septicemia | 5.1 | 1.2-21.7 | 4.2 | 1.5-11.8 |
| Surgical site infection | 8.7 | 1.2-65.2 | 6.8 | 0.8-61.1 |
⁎The adjusted model contains all covariates and includes clustering by hospital. |
Fig.
Nelson-Aalen cumulative hazard estimates. Graft infection readmission by index nosocomial infection (NI).
Discussion
To our knowledge, this study is the first population-level analysis of AGI. Most of the current published reports are of case series with limited numbers whereas this study is a population level analysis of AGI.2, 11, 12 Management of AGI ranges from complete graft excision, to extra-anatomic bypass, to in-line reconstruction using a variety of conduits. The overall mortality rate from AGI is high, from 20% to 40%, and AGI carries many other complications, including reinfection of graft material after surgery, chronic septicemia, and major amputation.4, 7, 8, 13, 14, 15
Little is known of the true incidence of AGI or contributing factors that may increase risk. This study demonstrates that the rate of AGI is low, is significantly associated with BSI and SSI, and has a 1-year mortality rate of 28%. This low rate of infection and high mortality has been similarly reported in previous small series.13, 15, 16 We found no difference in the 2-year rate of AGI between patients having open and EVAR repair. Aortic graft infection was also similar between elective and nonelective procedures. Most of the infections presented in the first postoperative year in the open repair cohort, which is contrary to conventional wisdom suggesting that AGI occurs at 3 to 5 years after aortic surgery.17 This study also demonstrates that AGI after EVAR is extremely rare18 and that the 2-year rates of AGI were similar among patients undergoing open repair and EVAR.
Improvement in the diagnosis of AGI as well as management schemes preventing AGI, such as use of rifampin grafts,3, 19, 20, 21 have been suggested. With the knowledge of possible causes of AGI as well as the onset and time of an aortic infection, it may be possible to prevent these graft infections or intervene earlier in their course. Patients after open aortic repair with known BSI or SSI infections may benefit from surveillance in the first postoperative year because this appears to represent the period of greatest risk of infection. Possibilities include the more liberal use of computed tomography in patients at high risk or longer treatment of BSI infections with antibiotic therapy.
The finding that EVAR and open procedures had similar rates of infection suggests that intraoperative contamination may not be the only source of graft infection. It would not be expected that EVAR, where the graft material is not manipulated, would carry a similar rate of AGI. This study supports the theory that periprocedural infection has a significant role in the development of AGI. The role of NIs, specifically BSI, suggests that there may be an intermediate step of hematogenous seeding of aortic grafts in the genesis of AGI.
This study has several limitations. The CHARS database did not include patients in military hospitals, Veterans Affairs medical centers, or Washington residents who underwent aortic surgery in the state and immediately moved to a different state. Because of the way the CHARS database is coded, there is the concern for the code containing grafts related to access for dialysis. All patients with the diagnosis of dialysis and dialysis grafts were removed from the study. This may have led to an under-representation of graft infection because there is the possibility that patients with AGIs may also incur renal failure as a postoperative complication.
The potential for inclusion bias as a result of the limited coding schemes for the many clinical entities that make up vascular graft infection cannot be entirely excluded. Classification as elective vs nonelective, which included “urgent” and “emergency,” was based on the CHARS variable for admission type. For this study, any cases that were defined as urgent were placed into the nonelective category. This may be an over-representation and is a coding bias based on the institutions submitting data. As well, the Charlson Comorbidity Index is intended to adjust for comorbid illnesses, it does not incorporate the degree of illness, nor has it been specifically applied to the population of patients undergoing aneurysm or infected graft repair. There is also the possibility of a type II error because the occurrence of AGI is so low in the population at large. Another limitation of discharge data is that the ICD-9 codes are relatively nonspecific for organisms and inconsistent in entry; therefore, no one organism was identified from discharge data. Finally, we used patient status as a Medicaid beneficiary or uninsured as a proxy of socioeconomic status22, 23 owing to lack of more direct assessments.
Conclusion
We are suggesting a paradigm shift in the way AGI is perceived. This study supports that AGI is significantly associated with periprocedural infections and occurs most commonly in the first year after surgery. The identification of patients at highest risk may be lead to the creation of preventive programs for high-risk patients to prevent future aortic infection. Further studies that look at possible prevention plans of AGI are warranted, although this will be difficult due to the low incidence.
Author contributions
References
- . Prosthetic graft infection after descending thoracic/thoracoabdominal aortic aneurysmectomy: management with in situ arterial allografts. J Vasc Surg. 2001;33:671–678
- . Management of infected aortic prosthetic grafts. Vasc Endovascular Surg. 2004;38:75–82
- . A systematic review and meta-analysis of treatments for aortic graft infection. J Vasc Surg. 2006;44:38–45
- . Long-term outcome after treatment of aortic graft infection with staged extra-anatomic bypass grafting and aortic graft removal. J Vasc Surg. 2000;32:451–459discussion 459-60
- Prevention and treatment of aortic graft infection. Expert Rev Anti Infect Ther. 2007;5:305–315
- . Long-term results following surgical management of aortic graft infection. Arch Surg. 1991;126:507–511
- The results of in situ prosthetic replacement for infected aortic grafts. Am J Surg. 1999;178:136–140
- Total excision and extra-anatomic bypass for aortic graft infection. Am J Surg. 1991;162:145–149
- . 11-year experience with anatomical and extra-anatomical repair of mycotic aortic aneurysms. Eur J Vasc Endovasc Surg. 2004;27:585–589
- . Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. 1992;45:613–619
- Vascular prosthetic graft infection: epidemiology, bacteriology, pathogenesis and treatment. Acta Chir Belg. 2002;102:238–247
- Evolution from axillofemoral to in situ prosthetic reconstruction for the treatment of aortic graft infections at a single center. J Vasc Surg. 2006;43:1166–1174
- . Prosthetic graft infection. Surg Clin North Am. 1994;74:571–590
- . Axillobifemoral bypass and aortic exclusion for vascular septic lesions: a multicenter retrospective study of 98 cases (French University Association for Research in Surgery). Ann Vasc Surg. 1992;6:119–126
- . Surgical management of infected abdominal aortic grafts: review of a 25-year experience. J Vasc Surg. 1986;3:725–731
- . Aortoiliac stent graft infection: current problems and management. Ann Vasc Surg. 2004;18:521–526
- Long-term follow-up of positive cultures in 500 abdominal aortic aneurysms. Arch Surg. 1993;128:284–288
- . Fungal infection of aortoiliac endograft: a case report and review of the literature. Ann Vasc Surg. 2007;21:228–231
- . Prevention of vascular graft infection by rifampin bonding to a gelatin-sealed Dacron graft. Ann Vasc Surg. 1991;5:408–412
- The prophylactic efficacy of rifampicin-soaked graft in combination with systemic vancomycin in the prevention of prosthetic vascular graft infection: an experimental study. J Surg Res. 2005;129:329–334
- Graft infectivity of rifampin and silver-bonded polyester grafts to MRSA contamination. Vasc Endovasc Surg. 2005;39:411–420
- . Patient demographic and socioeconomic characteristics in the SEER-Medicare database applications and limitations. Med Care. 2002;40(8 suppl):IV-19-25
- . Overview of the SEER-Medicare data: content, research applications, and generalizability to the United States elderly population. Med Care. 2002;40(8 suppl):IV-3-18
Competition of interest: none.CME article
PII: S0741-5214(07)01691-6
doi:10.1016/j.jvs.2007.10.030
© 2008 The Society for Vascular Surgery. Published by Elsevier Inc. All rights reserved.