Development of Postoperative Pneumonia After Endovascular Aortic Aneurysm Repair is Associated with an Increased Length of Intensive Care Unit Stay

Objective Endovascular aortic aneurysm repair (EVAR) has increasingly replaced open aortic surgery for treatment of abdominal aortic aneurysms (AAA). One of the key advantages of EVAR is the reduced length of intensive care unit (ICU) stay and hospital stay. This study aimed to identify the risk factors associated with increased ICU length of stay (LOS) after EVAR. Methods The American College of Surgeons (ACS-NSQIP) database for the year 2013 was used. All patients who underwent EVAR were divided into two groups: ICU LOS <1 day vs. ≥1 day. Preoperative, intraoperative, and postoperative factors were compared between these two groups utilizing bivariate logistic regression analysis. Multivariable logistic regression analysis was then used to identify factors that were independently associated with ICU LOS ≥1 day after EVAR. Results A total of 2,468 patients (18.7% females, 81.3% males) were identified. Group 1 (ICU LOS <1 day) = 1,535 patients and Group 2 (ICU LOS ≥1 day) = 933 patients. Multivariable analysis identified the following factors to be associated with ICU LOS ≥1 day: ruptured AAA (OR 3.88, CI 1.97-7.65), the American Society of Anesthesiology (ASA) score of 4-5 (OR 2.82, CI 1.50-5.31), operative time ≥180 minutes (OR 2.10, CI 1.51-2.93), bilateral groin cut down (OR 1.37, CI 1.10-1.71), juxta-renal AAA (OR 1.65, CI 1.16-2.35), renal artery stent (OR 2.13, CI 1.42-3.21), aortic stent (OR 2.39, CI 1.60-3.55), emergency surgery (OR 2.56, CI 1.94-3.38), need for blood transfusion (OR 3.11, CI 2.08-4.65) and postoperative pneumonia (OR 7.04, CI 1.95-25.45). Conclusion Variables identified above can be used to predict the cohort of EVAR patients which will likely require ICU for ≥1 day. Development of postoperative pneumonia is associated with a 7.04 times increase in ICU LOS ≥1 day.


Introduction
The past two decades have seen a revolution in the surgical treatment of abdominal aortic aneurysms (AAA). Since EVAR was first described by Parodi et al., subsequent randomized controlled trials have supported its safety and efficacy [1][2][3]. EVAR has now become the most common modality for the treatment of abdominal aortic aneurysms. One of the main advantages of EVAR over traditional open AAA repair is the reduced intensive care unit (ICU) and hospital length of stay (LOS). There is recent evidence to suggest that EVAR can be performed safely even in the outpatient settings [4]. In the era of increasing focus on reducing the postoperative complications, the length of stay in the intensive care unit and hospital, and healthcare costs, hospitals are developing strategies to improve these metrics.
Most patients do not need ICU care after elective EVAR. However, patients with numerous physiologic risk factors and those who undergo emergent EVAR are in need of ICU care postoperatively. Increased length of ICU stay after EVAR has been shown to be associated with significant complications, including death [5]. In this study, we aimed to determine the associating factors and develop a risk model for predicting the increased ICU LOS after EVAR. For this purpose, we chose to use the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP), since this is the largest surgical database in the US with a record of ICU LOS for a variety of general and vascular surgery operations.

Dataset
For the purpose of this study, we retrospectively analyzed the database provided by the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP). This database provides the users with Participant Use Files (ACS-NSQIP PUFs) [6]. These files contain patient de-identified data from multiple hospitals that participate in the ACS-NSQIP. ACS ensures that the information in the database remains compliant with the requirements of the Health Insurance Portability and Accountability Act (HIPAA). The database contains data from over 250 hospitals across the United States. It contains over 150 patient variables, including preoperative, intraoperative and postoperative factors. In order to ensure the integrity of the data, ACS requires each site to hire a reviewer who is specifically trained to collect and enter the data into the database. Data are collected in a systematic sampling process using an eight-day cycle schedule to avoid bias in selecting cases. To maintain diversity and high-quality data, high volume procedures are limited to a certain maximum number of cases per cycle, and cases with incomplete 30-day follow up are excluded from the database. The database is reliable with the inter-rater reliability (IRR) audit of the overall disagreement rate of approximately 2% among the participating sites [7]. Methods used for this analysis have been described in the previously published literature [7][8][9][10][11].

Outcomes
The main ACS NSQIP PUF was merged with the EVAR-Procedure Targeted PUF to identify patients who underwent EVAR in the calendar year 2015. Patients were divided into two groups based on LOS in the ICU: Group 1 (ICU LOS < 1 day) and Group 2 (ICU LOS ≥1 day). Several preoperative, intra-operative and post-operative variables were compared between these two groups.
All variables were summarized prior to any analysis using frequencies and percentages or means, medians, and standard deviations to examine data quality. The distribution of continuous variables was checked using histograms, box plots, and normal probability plots. Logistic regression was applied using a bivariate approach to search for associations between ICU LOS ≥1 and potential predictor variables. Because there were so many significant variables, only variables with a significant association with ICU LOS ≥1 having a p-value <0.01 were chosen as a subset for consideration for a multivariable predictor model. Before model selection was initiated, a check for multicollinearity between the potential predictor variables was performed using a variance inflation factor (VIF) statistics. The remaining predictors were then pared down to a final model using several methods of selection: best subsets, stepwise, backward, and forward. Inclusion and exclusion criteria for the selection methods that used them were set at p < 0.01 to be more stringent, given the large number of significant predictors. The results of these selection methods were combined to choose the most significant set of predictors of ICU LOS ≥1 collectively. The fit of the final model was assessed using Pearson, Deviance, and the Hosmer and Lemeshow goodness-of-fit tests. Model-adjusted odds ratios were used to quantify the size and direction of the significant associations. Predicted probabilities based on patient characteristics were calculated using the prediction equation created based on the model parameter estimates. All analyses were carried out using SAS version 9.4 (SAS Institute, Cary, NC).

Demographics
A total of 2,512 patients underwent EVAR in 2015. The mean age was 73.8 ± 9.0 years. ICU LOS could be determined in only 2,468 (18.7% females, 81.3% males) of these patients and were included in the analysis. Of these 2,468 patients, 1535 had ICU LOS<1 day (Group 1) and 933 had ICU LOS ≥1 day (Group 2). The mean ICU LOS was 1.0 (±2.8) days.

Bivariate analysis
The following factors were found to have significant association with ICU LOS ≥1 day: sex (female vs male: odd ratio (OR) 1. 26

Predicted probability
The probability of ICU LOS ≥1 day was calculated for each factor identified from multivariable analysis. The probability of ICU LOS ≥1 day was 11.73% for patients who underwent bilateral groin cut down, 12.72% for patients with aneurysm diameter ≤5cm, 13.82% for patients who had non-infrarenal proximal aneurysm, 16.14% for patients who needed unplanned readmission, 16.95% for patients whose operative time was ≥180 minutes, 17.14% for patients who needed renal stent, 18.81% for patients who needed aortic stent, 19.90% for patients who underwent EVAR as non-elective surgery, 21.48% for patients who had an ASA classification of 4-5, 23.17% for patients who needed transfusion (intra/postoperative), 27.36% for patients who had ruptured aneurysm as indication for surgery, and 40.57% for patients who had postoperative pneumonia.
Patients who had all of these factors have a 99.99% probability of having ICU LOS ≥ 1 day after EVAR. Patients without any of these factors had a probability of 8.84% at baseline ( Table 3).

Discussion
This analysis of a large, multi-institutional database of all EVARs performed across the US aims to determine the factors associated with increased length of ICU stay during the postoperative period. It identifies the following risk factors: ruptured AAA, emergent surgery, longer operative time, higher ASA score, bilateral groin cut-downs, extension of aneurysm into the supra-renal aortic segment, need for aortic or renal stents, postoperative pneumonia, bleeding and unplanned readmission to hospital. The predicted probability model shows that in the absence of these risk factors, the probability of prolonged ICU admission is 8.8%, and in the presence of all of these risk factors, the predicted probability of prolonged ICU stay is 99.99%. Presence of post-operative pneumonia itself is associated with a 41% probability of prolonged ICU admission. These findings have potentially important implications for the management of EVAR patients.
Our study shows several important factors that contribute to prolonged ICU stay after EVAR ( Table 2). While some of these factors can be considered modifiable, several of them cannot be modified. For example, variables such as ruptured AAA, high ASA scores and juxta-renal AAAs cannot be changed. On the other hand, factors such as bilateral femoral cut-downs, prolonged operative time and postoperative pneumonia can be considered potentially preventable. For factors that cannot be modified, this study highlights the importance of recognizing patients who are at high risk for requiring prolonged ICU stay. It also highlights risk factors that can be prevented to reduce ICU LOS. Postoperative pneumonia is associated with an extremely high risk for prolonged ICU stay after EVAR.
Endovascular aneurysm repair vs open repair (EVAR trial 1) and the Dutch randomized trial comparing conventional and endovascular repair of AAA (DREAM Trial) have established EVAR as a safe and effective alternative surgical treatment for AAA [2][3]. Among many advantages of EVAR over open repair are the reduced operative time, length of ICU stay, length of hospital stay, and early ambulation. In addition, EVAR is more cost-effective as compared to open AAA repair. While the cost of endografts is the major contributor toward the overall cost of EVAR, major reduction in the cost comes from reduced ICU LOS and hospital LOS. During the past decade, the overall cost associated with EVAR has improved, and most of this cost reduction comes from reduced LOS [12]. Our intent for this study was to examine all factors associated with ICU LOS and to focus on those factors, which are deemed preventable. Our study shows that the development of postoperative pneumonia is a significant preventable factor. Strategies to improve preventable factors may help us reduce the ICU LOS, associated morbidities, and overall cost of EVAR. Prolonged ICU LOS after EVAR itself has been associated with the development of serious complications, including death [5]. With an increasing focus on reducing postoperative complications and stress on using cost-effective strategies, our analysis identifies an area of improvement for patients undergoing EVAR. Postoperative complications impose a tremendous financial burden on both the patient and the healthcare system [13][14]. A retrospective review of the veteran affairs (VA) surgical patients showed that the estimated excess cost from postoperative complications ranged from $8,338 for superficial surgical site infection to $29,595 for failure to wean from ventilation within 48 hours. For pneumonia itself, the estimated excess cost was $12,798 [14]. Postoperative complications are highly associated with readmission, and readmission comes with additional cost. Lawson et al. show that by 5% reduction in postoperative complications, a potential 2,092 readmissions can be prevented, saving Medicare 31 million dollars per year [15].
Patients undergoing major vascular surgery operations are at significantly high risk for developing postoperative complications compared to those who undergo other major surgical operations [16]. Postoperative morbidity is associated with a significantly high risk of mortality [17]. Association of postoperative morbidity is, in fact, higher than that of baseline comorbid disease [18]. While the other complications after EVAR have been described in literature before, there is sparse literature on the impact of postoperative pneumonia on the overall prognosis. A retrospective review of all vascular surgery operations performed across the US has shown that postoperative pneumonia is associated with a significantly high risk of developing end-organ dysfunction, 30-day mortality, prolonged length of hospital stay and hospital readmission [19]. Our study shows that the development of postoperative pneumonia is associated with increased LOS in the ICU in the postoperative period. These findings are important, as postoperative pneumonia can be considered preventable for elective operations. Katsura [23]. Such strategies can be employed to prevent postoperative pneumonia and hence the complications associated with it. There is evidence to support that development of pneumonia prevention bundles by hospitals reduces the incidence of pneumonia [24][25]. The findings from our study, in combination with the above-mentioned literature, make the case for employing all pneumonia prevention strategies for EVAR patients to reduce the incidence of this complication and the associated increased length of ICU stay. As shown by Greenblatt et al., reducing the incidence of pneumonia after aortic surgery can also help reduce the incidence of unplanned readmission to the hospital [26].
This study has several limitations. It is a retrospective analysis where findings are limited to those hospitals that were enrolled in ACS-NSQIP in 2015. In addition, the data is not allinclusive, and it is a self-reporting registry. We lack information about the documented reason for ICU admission. The strengths of this study are that it is based on the largest surgical database in the US. The findings from this database in previously published literature have been shown to be reliable and reproducible in subsequent studies. The large volume of patients in the dataset provides a powerful tool for meaningful statistical analysis. It includes large, tertiary care academic medical centers and small community hospitals, hence providing data from a broad variety of sources. Correlation of pneumonia with increased ICU LOS and other morbidities correlates with previously published literature. Our data shows that the occurrence of more than one complication is associated with longer length of ICU stay.

Conclusions
With EVAR becoming the predominant modality for surgical treatment of AAA, attention should be paid to factors associated with increased morbidity and mortality after EVAR. Prolonged ICU LOS is associated with poor outcomes and increased cost. This study identifies the factors associated with prolonged ICU LOS after EVAR and identifies that postoperative pneumonia is associated with a significant risk of increased ICU LOS.

Additional Information Disclosures
Human subjects: All authors have confirmed that this study did not involve human participants or tissue. Animal subjects: All authors have confirmed that this study did not involve animal subjects or tissue. Conflicts of interest: In compliance with the ICMJE uniform disclosure form, all authors declare the following: Payment/services info: All authors have declared that no financial support was received from any organization for the submitted work. Financial relationships: All authors have declared that they have no financial relationships at present or within the previous three years with any organizations that might have an interest in the submitted work. Other relationships: All authors have declared that there are no other relationships or activities that could appear to have influenced the submitted work.