Sex Disparity in Outcomes of Ruptured Abdominal Aortic Aneurysm Repair Driven by In-Hospital Treatment Delays

Linda J. Wang, MD, MBA; Satinderjit Locham, MD; Hanaa Dakour-Aridi, MD; Keith D. Lillemoe, MD; Bryan Clary, MD; Mahmoud B. Malas, MD, MHS


Annals of Surgery. 2019;270(4):630-638. 

In This Article


Study Sample

This analysis was performed using prospectively collected data in the Vascular Quality Initiative (VQI) of the Society for Vascular Surgery (SVS). The VQI has been well described and is an international database with contributions from >500 centers across the United States and Canada.[14] Deidentified demographic, clinical, and procedural variables are collected at each participating institution and centralized. Additional data collected include perioperative complications and all-cause mortality, as obtained via linkage to the Social Security Death Index. Data collection and input is performed at the institutional level utilizing a multidisciplinary group of physicians, data managers, and quality professionals. Regular auditing is performed by the VQI to ensure data fidelity and minimize entry errors. Participation in this quality registry as a patient safety organization (PSO) and quality assurance tool was approved, and direct informed consent waived, at each center's Institutional Review Board. Specific approval for this data collection and assessment was permitted by the VQI PSO Research Advisory Committee. More information about the VQI can be found at

Patients, Data Collection, Definitions, and Endpoints

All AAA repairs in the VQI from participating centers in the United States from January 2003 through May 2017 were identified and reviewed. All nonruptured cases were excluded from analysis; evidence of rupture must have been confirmed via imaging, angiography, or operative evidence. After exclusion, 165 centers contributed cases to the open rAAA module, whereas 176 contributed to the endovascular rAAA module over the study period.

Demographics, medical comorbidities, and operative details were reviewed. In the VQI, the sex variable is binary (male and female) and providers are instructed to input sex assigned at birth. We calculated body mass index (BMI) using documented heights and weights and obesity was defined as having a BMI ≥ 30 kg/m2. End-stage renal disease (ESRD) was defined as dialysis dependence or having a history of a kidney transplant. A patient was considered to have iliac aneurysmal disease if one or both common iliac arteries had a diameter >1.5 cm. Transfer status was defined as transfer from another hospital or acute care facility. The VQI defines door-to-intervention time as "time from arrival at hospital performing treatment until initial incision or sheath placement for aortic repair. If patient was transferred from another facility, calculate time from admission into the facility where the procedure was performed." Additional information regarding the definitions of medical comorbidities and procedure details within VQI, as modeled after the Vascular Study Group of New England, have been previously described.[15,16]

Our primary endpoint was 30-day death, whereas secondary endpoints included 1-year all-cause mortality and major perioperative complications. The latter included the following complications occurring before hospital discharge: stroke, myocardial infarction (MI), respiratory complication (pneumonia or reintubation), acute kidney injury (creatinine increase >0.5 mg/dL and/or dialysis requirement), mesenteric ischemia, lower extremity ischemia requiring operative intervention, and/or unplanned return to the operating room. Patients who died before discharge were excluded from length of stay calculations.

Time-delay Subgroup Analysis

Time-delay cohorts were created using the 90-minute benchmark.[9] Patients were stratified by those who were treated within 90 minutes of admission versus those treated outside of this time frame. Patients who ruptured after admission or who were missing one or more temporal data point were excluded from this subanalysis.

Statistical Analysis

Statistical analysis was performed with Stata/SE 15 software (StataCorp LLC, College Station, TX). Dichotomous variables are described as a percentage of the cohort. Continuous variables are expressed as median [interquartile range (IQR)]. Univariate analysis was performed to compare baseline clinical, operative, and demographic features and postoperative complications between sex cohorts. For each comparison, the chi-square test was used for discrete variables and the nonparametric equality of median test for continuous variables. Significant factors in univariate analysis were included in a logistic regression model to determine independent predictors of 30-day mortality. Subsequent goodness-of-fit (Hosmer–Lemeshow lack of fit) tests were performed and area under receiver-operating characteristic (ROC) curve calculated. The overall cohort models adjusted for open surgical repair, age, race, obesity, current smoking, coronary artery disease (CAD), congestive heart failure (CHF), prior aortic surgery, facility transfer, and maximal aortic diameter. The OSR-specific models adjusted for age, obesity, current smoking, CAD, CHF, and iliac aneurysmal disease. The EVAR-specific models adjusted for age, ESRD, obesity, current smoking, chronic obstructive pulmonary disease (COPD), and maximal aortic diameter. For all models, age was treated as a categorical variable and patients were stratified into 3 age groups: <60, 60 to 79, and ≥ 80. Survival was calculated using Kaplan–Meier life tables. Cox proportional hazards models were created to identify multivariate predictors of survival. A P value of ≤.05 was considered statistically significant.