Characteristics and Prognosis of Abdominal or Thoracic Aortic Aneurysm Patients Admitted to Intensive Care Units After Surgery Therapy: A Multicenter Retrospective Observational Study

Background: Subsequent intensive care unit (ICU) admissions postoperatively are not rare for patients with abdominal or thoracic aortic aneurysm (AAA or TAA), but a large-scale investigation on these patients is absent. The study aimed to investigate the characteristics and prognosis of AAA or TAA patients admitted to ICU postoperatively. Methods: Patients admitted to ICU postoperatively with a primary diagnosis of AAA or TAA were screened in the eICU Collaborative Research Database, which contained data from multiple ICUs throughout the continental United States in 2014 and 2015. Baseline characteristics and comorbidities and were investigated and factors associated with ICU mortality were explored using univariable logistic regression. Receiver operating characteristic (ROC) curve analysis was performed to evaluate the prognosis predictive performance of the widely used severity scoring system APACHE IVa. Results: 974 patients including 677 AAA and 297 TAA patients admitted to ICU postoperatively were included nally. Compared with TAA, AAA patients had a signicantly higher median age (72 versus 64 years). 10.19% AAA and 2.36% TAA patients suffered from rupture of aortic aneurysm, and 89.07% AAA and 84.51% TAA patients underwent elective surgery. Hypertension requiring treatment was the most common comorbidity (57.31% for AAA and 61.95% for TAA). TAA patients had signicantly higher ICU mortality (9.43% versus 2.36%) than AAA. Several factors were found to be signicantly associated with ICU mortality, including urgent surgery, rupture of aortic aneurysm, TAA, and a higher APACHE IVa score on ICU admission. APACHE IVa showed a good predictive performance for ICU mortality with an area under the ROC curve of 0.9176 (95% CI 0.8789-0.9390). Conclusions: Prognosis of aortic aneurysm patients admitted to ICU postoperatively is yet to improve, and factors associated with prognosis are mainly related to the condition itself. APACHE IVa can be used for prognosis prediction.


Introduction
Aortic aneurysm (AA) is the enlargement of the aorta de ned as a segmental, full-thickness dilation of the blood vessel having at least a 50% increase in diameter compared with the expected normal diameter 1-3 . It usually located in the abdominal aorta, but can also be located in the thoracic aorta. AA usually causes no symptoms, however, it is with an increased risk of aortic rupture due to weakness in the wall of the aorta and could be fatal when ruptured 4 .
As a clinically silent disease, the prevalence and incidence of abdominal aortic aneurysm (AAA) and thoracic aortic aneurysm (TAA) are not fully understood. AAA was reported to have a prevalence of 4-8% [5][6][7][8] and TAA was 0.16-0.34% 9 10 in screening studies. The relatively low prevalence and incidence make it di cult to conduct researches with enough sample sizes to obtain evidence, while obviously it is very important to have a working knowledge of the characteristics and prognosis of AA to aid with clinical decision making.
In the past three decades, the greatest advance in the treatment of AA was the introduction of endovascular aneurysm repair (EVAR), which used stent grafts and caused a major paradigm shift in the eld of AA surgery 11 . For elective AAA repair, the mortality is 3-5% for open surgery, but decreases to 0.5-2% for EVAR [12][13][14] . However, the overall prognosis of AA is still not satisfactory, especially for ruptured AA.
The 30-day and 90-day all-cause mortality rates were reported to be 22.9% and 33.3% respectively for ruptured AAA patients after surgery therapy 15 . To further improve prognosis of AA, no doubt postoperative management should be improved 16 .
Although not all AA patients need intensive care unit (ICU) admissions after surgery therapy 17 , subsequent ICU admissions postoperatively at least are not rare for patients with AAA or TAA. A study from Denmark reported that 62% AAA patients after open AAA repair surgery had ICU stays >24 hours 18 .
In order to further improve the management of AA patients admitted to ICU postoperatively, an overview of their characteristics and risk pro les of prognosis is needed. However, as far as we know, a large-scale investigation on these patients is absent. Therefore, the study aimed to investigate the characteristics and prognosis of AAA or TAA patients admitted to ICU postoperatively.

Data source
The study used data from the eICU Collaborative Research Database (version 2.0), which was made available largely by Philips Healthcare and collaborators at Massachusetts Institute of Technology (MIT) Laboratory for Computational Physiology 19 . It consisted of data of patients admitted to one of 335 critical care units at 208 hospitals that participated in the Philips eICU program throughout the continental United States in 2014 and 2015. The database is released under the Health Insurance Portability and Accountability Act (HIPAA) safe harbor provision. The re-identi cation risk was certi ed as meeting safe harbor standards by Privacert (Cambridge, MA) (HIPAA Certi cation no. 1031219-2). Access to the database was approved after completing the CITI "Data or Specimens Only Research" course. This study was exempt from institutional review board approval due to the retrospective design, lack of direct patient intervention, and the security schema for the re-identi cation risk.

Study population
Patients admitted to ICU postoperatively with a primary diagnosis of AAA or TAA were screened in the database. Data on primary diagnosis was extracted from the table "admissiondx" which contained the primary diagnosis for admission to the ICU per the Acute Physiology And Chronic Health Evaluation (APACHE) scoring criteria. AAA or TAA were identi ed based on structured text including "Aneurysm, abdominal aortic", "Aneurysm, abdominal aortic; with dissection", "Aneurysm, abdominal aortic; with rupture", "Aneurysm, thoracic aortic", "Aneurysm, thoracic aortic; with dissection", and "Aneurysm, thoracic aortic; with rupture". Detailed inclusion criteria were: 1) a primary diagnosis of AAA or TAA; 2) rst ICU admission only (for patients who had more than 1 ICU admission during the same hospitalization); 3) admitted from operating room, recovery room, or post anesthesia care unit (PACU); 4) age ≥18 years. Detailed exclude criteria were: 1) instead of "admit", patients with an ICU stay type of "stepdown/other", "readmit for undo", "pre-admit", "transfer", or "readmit"; 2) patients who had more than 1 hospitalization records in the database; 3) length of ICU stay is missing. Flow chart of the study population was presented in Figure 1.
Baseline characteristics, comorbidities, and study outcomes The following baseline characteristics were collected from the database, including age, sex, ethnicity, body mass index (BMI, calculated by height and weight), year of the hospital discharge date, type of surgery (elective or not), type of AA (ruptured or not), type of ICU, admitted location, Acute physiology IVa score, APACHE IVa score, several lab parameters of the APACHE IVa scoring system, intubation, ventilation, and dialysis on the rst day after ICU admission. Various comorbidities were also collected.
ICU mortality was the primary outcome of the study. Length of ICU stay, hospital mortality, and length of hospital stay were also studied but only for description. All the variables above were extracted from the tables in the database, including "patient", "apachePatientResult", "apacheApsVar", and "pastHistory".

Statistical analysis
Continuous variables were presented as median (25% percentile -75% percentile) and categorical variables were presented as number (percentage). Comparisons between groups were tested by Kruskal-Wallis H test for continuous variables or Chi-squared test (or Fisher's exact test) for categorical variables. The study population was also categorized based on length of ICU stay (<2 days, 2-7 days, and ≥7 days) and study outcomes of each category was described. Factors associated with ICU mortality were explored using univariable logistic regression. Kaplan Meier curves for ICU mortality together with logrank test were employed to compare the survival distribution of AAA and TAA. Receiver operating characteristic (ROC) curve analysis was performed to evaluate the prognosis predictive performance of APACHE IVa. A P value less than 0.05 was considered to indicate statistical signi cance. Empower(R) (www.empowerstats.com; X&Y solutions, Inc., Boston, MA, USA) and R software, version 3.4.3 (http://www.r-project.org; R Foundation for Statistical Computing, Vienna, Austria) were used for statistical analyses.

Prognosis of the study population
The overall ICU mortality for the study population was 4.52% (44/974), and the hospital mortality was 6.26% (61/974). Compared with AAA, TAA patients had signi cantly higher ICU mortality (9.43% versus 2.36%, P<0.001) and hospital mortality (12.12% versus 3.69%, P<0.001). Consistent result was observed in Kaplan-Meier survival curves for ICU mortality (Figure 2), which showed a poorer survival in TAA patients compared with AAA patients (P=0.0072). When the study population was categorized based on length of ICU stay (<2 days, 2-7 days, and ≥7 days), patients with longer length of ICU stay had worse prognosis (Table 3).

Factors associated with ICU mortality
Several baseline characteristics were found to be signi cantly associated with ICU mortality in univariable logistic regression (Table 4). Elective surgery was associated with reduced risk of ICU mortality (Odds ratio (OR) 0.22, 95% con dence interval (CI) 0.12-0.42), while rupture of AA was associated with increased risk of ICU mortality (OR 3.31, 95% CI 1.53-7.18). Compared with AAA, TAA was associated with increased risk of ICU mortality (OR 4.30, 95% CI 2.29-8.08). Acute physiology score IVa (OR 1.05 per 1 score increase, 95% CI 1.04-1.06) and APACHE IVa score on admission (OR 1.05 per 1 score increase, 95% CI 1.04-1.06) were also associated with poor prognosis. However, no comorbidities were found to be signi cantly associated with ICU mortality (Table 5).

Prognosis predictive performance of APACHE IVa
As presented in Table 6, APACHE IVa on ICU admission showed a good predictive performance for ICU mortality with an area under the ROC curve (AUC) of 0.9176 (95% CI 0.8789-0.9390). Compared with AA patients without rupture, the predictive performance of APACHE IVa decreased in patients with rupture of AA (AUC 0.8795 versus 0.9158). TAA patients also saw a poorer predictive performance of APACHE IVa compared with AAA patients (AUC 0.8660 versus 0.9233).

Discussion
The study focused on AA patients who were admitted to ICU after surgery therapy and investigated characteristics and prognosis of these patients. Predictors associated with prognosis and prognosis predictive performance of the widely used severity scoring system APACHE IVa were also explored. The main ndings of the study were: 1) there were some great difference in baseline characteristics and comorbidities between AAA and TAA patients who were admitted to ICU postoperatively; 2) prognosis of these patients is yet to improve, especially for TAA patients; 3) instead of comorbidities, factors associated with prognosis were mainly related to the condition itself; 4) APACHE IVa showed good prognosis predictive performance. Based on a relatively large sample size and the recent data (year 2014 and 2015), these ndings provided an updated overview of AA patients (either AAA or TAA) admitted to ICU after receiving surgery therapy, which might help to improve postoperative management of AA patients and therefore to further improve their prognosis.
Compared with most other studies, the baseline characteristics and comorbidities of AA patients in our study were broadly similar, including an advanced age, predominantly males, and a higher prevalence of hypertension comorbidity. For example, in a cohort from Japan who underwent open or endovascular repair for AAA 20 , only 1.4% (14/999) patients aged 50 years or younger, 85.3% (852/999) were male, and 79.9% (798/999) had hypertension. In a retrospective cross-sectional study from Iran 21 , the mean age of AAA patients in emergency department was 68.11±11.98 years, 84% were male, and 51.6% had hypertension. In a cohort of 107 untreated TAA patients 22 , the average age was 59.3 years, 73.8% (79/107) patients were male, and 47% had diastolic hypertension. Similarities between these studies might be related to the nature of the diseases, but it should be noticed that in our study the study population were only AA patients who survived after the surgery therapy and needed an admission to ICU. This might explain the slight difference in these characteristics between studies. A recent study found that female AAA patients might have higher risk of rupture due to gender differences in morphological and hemodynamic characteristics of AAA 23 , and considering that patients with rupture were more likely to be admitted to ICU after surgery 17 , our study thus included more female AAA patients. However, given the limited sample size of all these studies, variations in statistics of these characteristics could also explain the difference. In terms of ethnicity, most patients were Caucasian in our study, which was consistent with several other studies 24-26 that supported Caucasian race might be a risk factor for development of AAA. Researches that compared AAA with TAA were rare, making it impossible to compare our results with others. In our study, the TAA patients were younger, with a much lower proportion of rupture, and different patterns of comorbidities when compared with AAA patients. These differences might be related to the "selection bias" of our study population. As mentioned above, only patients who underwent a surgery, survived, and at the same time needed to be admitted to ICU would be included into our study.
Considering TAA had worse prognosis 27 , it is not strange to nd that patients who survived and therefore were enrolled were younger and presented a different pattern of comorbidities compared with the AAA patients who were much elder.
Reports about prognosis of AA patients varied. A study included 138 ruptured AAA patients who received open surgical repair and survived at least 48 hours in ICU between 1999 and 2003 reported a 30-day mortality of 22% 28 . This mortality rate was much higher than that of our study population, which might be related to the different study population since ruptured AA had signi cantly higher risk of mortality. In our study we also found that compared with patients without ruptured AA, the risk of ICU mortality of patients with ruptured AA increased (OR 3.31, 95% CI 1.53-7.18). Another reason could be the different surgery therapies used, since EVAR was found to have lower perioperative mortality than open surgery 29 .
Although in our study detailed information about surgery therapy was unavailable, given the study period was between 2014 and 2015, and most cases received elective surgery, they might be less likely to receive open surgery. We also found that patients with longer length of ICU stay had worse prognosis, which was consistent with the study conducted by Gavali et al 30 .
Among the many variables, only a few factors that related to the condition itself were signi cantly associated with prognosis, including elective surgery, rupture of AA, and type of AA, while none of the comorbidities show a signi cant association. These results suggested at least the short-term prognosis was mainly determined by the condition itself. APACHE IVa was also found to be associated with ICU mortality of the study population, and ROC curve analysis showed it had good discriminatory capability. This is consistent with Kabbani et al 31 study where they found APACHE III score on ICU admission was an excellent discriminator of hospital mortality (AUC 0.92, 95% CI 0.83-1.00) for patients after open thoracoabdominal and open AAA repair. As far as know, there is no study evaluate the predictive performance of APACHE IVa in AAA and TAA patients after surgery therapy, and thus our nding could be seen as a validation of this scoring system in AA patients.

Strengths and Limitations
Strengths of the study included a multiple centers design and the updated data. The relatively large sample size also increased the power of the study. In addition, bene ting from the great abundance of variables in the database, various variables about baseline characteristics and comorbidities were investigated in the study. However, the study inevitably had some limitations. First, as a retrospective study, all the data we used were directly extracted from the available database without validation, and therefore measurement error should be considered as the main source of bias in the study. For example, it is possible that a diagnosis of AAA, TAA, or a TAA that also involves the aortic valve could be entered into the database incorrectly, since it was usually entered by nurses instead of clinicians. Second, since the database only contained data during the ICU hospitalization, information about surgery therapy before ICU was unavailable. We therefore applied strict inclusion criteria on the study population, and only included patients who were with a primary diagnosis of AAA or TAA and were directly admitted to ICU from operating room, recovery room, or PACU. Another concern was potential selection bias of the study population. Since not every ICU or hospital participated in the eICU program, AA patients admitted to those unanticipated ICU/hospitals would always be missed in our study. Nevertheless, further studies are needed to validate the ndings in our study.

Conclusion
Prognosis of aortic aneurysm patients admitted to ICU postoperatively is yet to improve, and factors associated with prognosis are mainly related to the condition itself, while APACHE IVa is a reliable tool for prognosis prediction.   Abbreviations: ICU, intensive care unit.  Area under the Receiver operating characteristic curve; CI, con dence interval. Figure 1 1) instead of "admit", patients with an ICU stay type of "stepdown/other", "readmit for undo", "pre-admit", "transfer", or "readmit"; 2) patients who had more than 1 hospitalization records in the database; 3) length of ICU stay is missing. Flow chart of the study population was presented