Vascular surgery patients in intensive care: a bi‐national cohort study over 15 years

To describe the characteristics and outcomes of patients admitted to Intensive Care following vascular surgery, and their subsequent usage of Intensive Care resources, over a 15‐year period in Australia and New Zealand.


Introduction
Arterial disease is a significant contributor to morbidity, mortality, and healthcare costs in Australia and New Zealand, and globally. [1][2][3][4] Compared to other surgical specialties, vascular surgery has been forecast to have the largest growth in demand between 2013 and 2025, with the increase being primarily due to the ageing population and the rise of diabetes. 5 A percentage of vascular surgery patients require admission to an Intensive Care Unit (ICU) early in the post-operative period, particularly relevant following major open arterial procedures. With the evolution of endovascular procedures this percentage has likely decreased over time.
Significant variability is also seen in the post-operative outcomes of patients undergoing vascular surgery. [6][7][8][9][10][11][12][13] This variance depends on the nature of the pathology, the modality of intervention and patient comorbidities. To date, data looking at vascular surgery patients admitted to ICU has mostly been limited singlecentre studies. [14][15][16][17][18] Despite significant changes in vascular surgery over recent years, multi-centre data looking at ICU utilization and post-operative mortality is lacking. This study aims to describe the number, demographics, and outcomes for patients admitted to ICUs in Australia and New Zealand following vascular surgery, and their subsequent use of ICU resources over a 15-year period.

Study design
This was a retrospective cohort study that used the Australian and New Zealand Intensive Care Society (ANZICS) Adult Patient Database (APD). The project was approved by the Alfred Health Human Research Ethics Committee (ref: 417/20), and the governance committee of the ANZICS Centre for Outcome and Resource Evaluation. The project was designed in accordance with the STROBE statement (Table S1).

Data source
The ANZICS APD is a bi-national registry that contains deidentified information on all admissions to participating adult Intensive Care Units (ICUs). The registry's primary purpose is benchmarking of units across the two countries, with research being a secondary purpose. Data is collected by trained staff working in each unit. Monitoring of data quality is undertaken by the custodian, ANZICS. Data captured includes patient demographic information, reason for admission to ICU, biochemical, physiological, and chronic health parameters from the first 24 h of ICU admission required for calculation of severity of illness scores and information on patients' outcomes. 19

Study setting
All Australian and New Zealand sites that reported to the APD during the study period were included. In 2019 it was estimated 90% of ICU admissions in the two countries occurred in units that reported to the APD. 20 In both countries, patients are primarily cared for by an Intensive Care Specialist during their ICU admission with the vascular surgeon consulting on their care. Following discharge from ICU the vascular surgeon takes over care.

Study population
The study included all adult (≥18 years of age) patients admitted to a participating ICU between 2005 and 2019 (inclusive) with a diagnosis code relating to a vascular surgery (Table S2). In the instance of the same patient having multiple ICU admissions only the first was included. Patients admitted for palliation or organ donation were excluded. Patients discharged to another ICU were excluded from the univariable and multivariable analysis.

Variables
Patient demographic variables included age at the time of admission to ICU, gender, the presence of chronic comorbidities were extracted from the ANZICS APD. Chronic diseases collected are modified APACHE-IIIJ and are defined by the ANZICS APD Data Dictionary, and are described in Table S3. 21 Admission variables included the source of admission to hospital and ICU, the nature of the admission (elective or non-elective), whether the patient received invasive ventilation at any point in the admission, the length of stay in ICU and hospital, severity of illness (quantified by the APACHE-IIIJ and ANZROD scores), the type of surgery performed and the discharge location from ICU and hospital. 21 The APACHE III-J score is an internationally used measure to quantify the severity of illness for patients admitted to ICU. 22 ANZROD is a highly discriminatory mortality prediction model specifically calibrated to the Australian and New Zealand setting. Both the APACHE-III-J and ANZROD are calculated using physiological and biochemical parameters measured in the first 24 hours of the ICU stay as well as patient demographic data and diagnosis information. Neither score can be calculated with only information known pre-operatively. 23 Hospital case volume was also recorded, defined as the average annual number of vascular surgery admissions. ICU resource usage was defined by the number of days a patient spent in ICU.

Outcomes
The primary outcome was mortality, defined as not surviving to hospital discharge. The secondary outcome was the proportion of total ICU bed days that were for vascular surgery patients.

Statistical analysis
Categorical data were presented as frequencies with percentages, all continuous variables were non-parametric, as such medians with inter-quartile ranges were reported. Comparisons between survivors and non-survivors were made using the Chi-square and Wilcoxon tests as appropriate. No sample size calculation was performed, all patients meeting inclusion criteria during the maximal study period were included.
Multivariable logistic regression was used to investigate the relationship between study variables and mortality. All variables described in Table 1 that were known prior to the ICU admission were included in the univariable analysis. As ICU severity of illness scores included post-operative data they were not included. The reference group for each variable was the largest, except for surgery type for which the second largest was used as the largest group (carotid endarterectomy) had few outcomes (deaths) occurring in it. All cases in which data were missing were excluded from this analysis. All variables were significantly associated with the outcome and deemed clinically significant in the univariable analysis, and we therefore included them in the multivariable analysis. Odds ratios (ORs) are reported with 95% confidence intervals (95%CI).
Differences in variables over time were assessed using negative binomial regression. A two-sided P-value of <0.05 was used as the level of significance. All analysis was conducted using the R Project for Statistical Computing Software Version 1.3.959.

Cohort derivation
There were 1 968 378 unique adult ICU admissions recorded in the ANZICS APD from January 2005 to December 2019. After applying exclusion criteria, a cohort of 69 718 patients admitted to 179 ICUs was derived, whose primary ICU admission diagnosis was related to vascular surgery. Figure S1 describes this process.
The number of ICUs contributing to the APD increased during the study period from 134 in 2005 to 206 in 2019.

Cohort characteristics
The cohort had a median age of 73 years and were predominantly male. Carotid endarterectomy was the most common procedure patients were admitted after, followed by an elective open aneurysm repair. Admissions were mostly elective and from the patient's home. Only a small proportion of admissions were from the ward (i.e., unplanned admissions following a Medical Emergency Team call). A quarter of patients were invasively ventilated during their admission. The median length of stay in ICU and hospital was just over a day and a week respectively. The  proportion of patients discharged to aged care was significantly higher than that admitted from aged care (5.4% versus 0.9%, P < 0.001). The characteristics of the cohort are described in detail in Table 1.

Mortality and associated factors
Overall, 5.2% of patients did not survive to hospital discharge. The factors associated with mortality in the multivariable analysis are described in Table 2. An increasing hospital case volume was associated with lower mortality, as were admissions to private or rural/ regional ICUs and ICUs in New Zealand. Women had higher mortality than men, as did patients that had any of the four comorbidities (renal, liver, cardiovascular or respiratory disease). Relative to elective open aneurysm repairs, ruptured aneurysm repairs had the highest mortality and carotid endarterectomies had the lowest.

Changes over time
There was an increase over time in the proportion of vascular surgery admissions that were non-elective in nature, with an average annual increase of 1.9% (P < 0.001). There was a concomitant decrease in elective admissions (Fig. S2). Large changes in the case-mix of vascular admissions was seen over time (Fig. 1). The largest decrease was seen in the elective open aneurysm repair group, in 2005 37% of vascular admissions were for these patients, whilst in 2019 they only represented 16% of the cohort. Conversely, the Endovascular Aortic Repair group grew from 2% of Note: All admissions with missing data were excluded from this analysis. An odds ratio > 1.0 means the variable was associated with mortality. For the comorbidity variables the reference group was the absence of the comorbidity. Admission year was omitted from the table but included in the analysis to increase readability, a full version of the analysis is included in Table S4. the cohort in 2005 to 12% in 2019. There was no change in the median illness severity during the study period (Fig. S2). Using the multivariable model described above adjusted mortality rates per year for the cohort were derived. Over the study period mortality decreased significantly (P < 0.001) (Fig. 2).

Proportion of ICU admissions for vascular surgery
During the study period the absolute number of vascular ICU admissions increased, but the proportion of ICU bed days for vascular surgery decreased by over a fifth from 3.6% in 2005 to 2.9% in 2019 (P < 0.001). All hospital types had decreases in vascular admissions, with the largest decrease seen in the Rural / Regional group. Figure S4 visualizes these changes.

Discussion
Overview To our knowledge, this is the first multi-national study to examine patients admitted to ICU following vascular surgery. The adjusted mortality for these patients decreased over the study period. In addition, the proportion of ICU bed days used for vascular surgery patients decreased despite the raw number of admissions increasing.

Cohort characteristics
The proportion of admissions in our study from Australia and New Zealand were similar to those reported for vascular surgery procedures in the two countries, reflected in an existing bi-national audit. 24 This suggests that Australia and New Zealand have similar rates of ICU admission following vascular surgery. This supports that the cohorts from the two countries were likely homogeneous, adding validity to the combination of their data in our analysis. The most common procedure in our cohort was a carotid endarterectomy (22%). Due to representing such a large proportion of the cohort future work should closely examine if any of these patients could be managed in a more acute ward environment with invasive haemodynamic monitoring capabilities versus an ICU bed. Large changes in case-mix over time were seen. Of note the proportion of admissions for open aneurysm repairs more than halved, whilst for Endovascular Aortic Repair (EVAR) it increased six-fold. With endovascular procedures becoming more common it is likely they will represent an increasingly large proportion of total vascular surgery procedures. [25][26][27] Mortality and associated factors Multiple patient and admission characteristics were associated with mortality. Female patients had a higher mortality than males. This has been previously attributed to females having more advanced disease, being older and having more comorbidities at time of surgery. [28][29][30][31][32] In this study, the finding was still present after controlling for age and comorbidities. Additional research is required to look into the reasons for this difference. 31 As expected, the type of surgery had a large impact on survival. Carotid endarterectomies had the lowest mortality, further prompting the need to carefully select which patients in this cohort require ICU care versus a highacuity ward bed. As expected, patients transferred from an aged care facility had an increased mortality. The decision in this cohort to operate and subsequently admit to ICU should be made with this in mind. Additional factors such as frailty have been shown to influence mortality and should be considered by surgeons in these instances, particularly in the outpatient setting whilst planning for an intervention. 33 During the study period a significant decrease was seen in adjusted mortality. There are likely multiple factors that have contributed to this. Improving operative and anaesthetic techniques have likely decreased the morbidity surgery confers. In addition to this the overall outcome was survival to discharge, not survival to ICU discharge, therefore improved ward care could be a contributor. Improved patient selection may also have contributed, there may be less instances of patients being admitted to ICU after surgery where their condition was too advanced and instead required medical or palliative management. Primary, and secondary prevention along with improved screening and diagnostic tests likely led to vascular pathologies being identified earlier and thus the patients were not as acutely unwell when receiving an intervention. Finally, there may have been changes to the nature of hospitals that provided data, the number of contributing sites to the ANZICS APD increased from 134 to 206 during the study period. In view of the above its should be noted there was no significant change in the median APACHE-IIIJ score over the study period, suggesting the acuity of patients admitted to ICU did not change.

Proportion of ICU admissions for vascular surgery
This study indicates that the proportion of post-operative ICU bed days for vascular surgery patients decreased from 3.6% in the year 2005 to 2.9% in 2019, a relative decrease of 21%. Multiple factors may have contributed to this change such as the rate of ICU admission after vascular surgery decreasing, or admissions for other reasons increasing at a greater rate. Of note the largest decrease in bed day use was seen in rural and regional hospital. This may reflect a further centralisation of vascular surgery in metropolitan centres. We identified that sites with lower case volumes had worse outcomes, low-case volumes for a site likely coincides with low volumes for individual surgeons and lower exposure to patient management for the entire treating team. Overall, alongside the purported increases in vascular surgery procedures, ICUs in Australia, and New Zealand are also admitting more patients after vascular surgery, however, as a proportion of total ICU bed days there has been a decrease.

Implications
There are several implications of the results presented. First, the results may be used for future studies to compare progress of various interventions over time, particularly with respect to survival. Secondly, our results demonstrate that broadly, strategies undertaken by Vascular Surgeons, Intensivists and other clinical staff in Australia and New Zealand in the care of vascular surgery patients over time have improved patient outcomes. Finally, many associations between patient and admission characteristics have been identified that are associated with mortality. Future studies could investigate these characteristics further to identify if any areas of patient care could be altered to mitigate against these differences.

Strengths/weaknesses
This study had several strengths. First, the study was bi-national, describing patients from both Australia and New Zealand. Secondly, the study included data on all admissions after vascular surgery in most ICUs in the two countries. Thirdly, the study examined data over a 15-year period facilitating an examination into any long-term trends. Finally, the cohort was not limited to those undergoing a single type of vascular procedure, thus providing a broad overview of all vascular surgery patients requiring admission to ICU in Australia and New Zealand.
Whilst the strengths of the study increased its generalisability to a global audience there were several limitations that must be considered when reviewing our results. First, the study was retrospective in nature. This limited our ability to select and report specific variables and determine the causality of associations. Second, the database used only described the type of procedure a patient underwent. There were no further data on operative factors which have been shown to be strong predictors of recovery, nor was data available regarding the speciality of the primary surgeon (e.g., Vascular or General). 34 Also, there were no data on the care or patient course except for the final discharge location after the patient left ICU. Third, it is likely there is variation between units regarding the threshold for ICU admission post-operatively, data for patients admitted to a general ward post-operatively were not available, precluding a comparison of these cohorts that may provide insight into such variation. Fourth, for instances where a patient had a vascular surgery but was admitted to ICU later in their admission for another reason (e.g., Sepsis) the diagnosis code would be assigned to the admission reason, not the surgery. Further in cases of trauma where there was not an isolated vascular injury the diagnosis code would be for trauma. This resulted in these patients not being included. Finally, this study did not examine the long-term outcomes of patient following discharge from hospital or report any subjective outcomes such as quality-of-life at discharge. Previously patients admitted to ICU post-operatively have reported a reduced qualityof-life in the years following their admission, bringing into question how appropriate it is to report survival outcomes in isolation. 35

Conclusion
In this study of patients admitted to ICU after vascular surgery in Australia and New Zealand we have identified multiple patient and admission characteristics associated with mortality. Additionally, over a 15-year period we found decreasing mortality, an increase in raw admission numbers, but a decrease in vascular surgery admissions as a proportion of all ICU bed days. These findings provide a benchmark for vascular surgery units worldwide and demonstrate progress has been made in the improvement of patient outcomes in recent years.

Supporting information
Additional Supporting Information may be found in the online version of this article at the publisher's web-site: Table S1 STROBE Statement. Table S2: Diagnosis codes included. Table S3: Chronic health condition definitions. Table S4: Multivariate logistic regression for mortality. (N = 65 433). Figure S1: Inclusion and exclusion flow chart for cohort. Figure S2: Change in admission type for vascular surgery over time. (N = 69 446) Figure S3: Change in median APACHE-IIIJ score over time.
(N = 69 446) Figure S4: Change in admissions for vascular surgery as a proportion of total ICU bed days grouped by hospital type over time.