Background
The mortality for ruptured abdominal aortic aneurysm (RAAA) remains high in the face of medical progress. Approximately 1% to 2% of all deaths in the western population are caused by RAAA [
1,
2]. The therapeutic options include endovascular aneurysm repair (EVAR) and open surgical repair (OSR). Systematic reviews based on observational studies suggest survival benefits for endovascular treatment when compared with OSR [
3‐
5]. Of note, no significant difference was found in recently published randomized controlled studies [
6‐
8]. It remains unclear whether the controversial results are due to superior treatment or patient selection.
Interestingly, there are predictors of survival independent of the chosen treatment: Hemodynamic shock, loss of consciousness, sex and the anatomy of the aneurysm [
6,
9‐
14]. Not only do preoperative factors predict mortality in RAAA, but postoperative condition might also have a clinical impact on survival. There are several outcome prediction models in the intensive care unit (ICU) environment. The Sequential Organ Failure Assessment score (SOFA) composed of scores from six organ systems determines the grade of organ dysfunction and failure [
15]. The Simplified Therapeutic Intervention Scoring System-28 (TISS-28) is assessed according to therapeutic activities using 28 items [
16]. The Simplified Acute Physiology Score II (SAPS II) is based on 12 physiological variables, age and type of admission [
17]. These three models are widely used by ICUs to track patients and to predict clinical outcome [
15,
18,
19].
Patients with RAAA who undergo OSR may die quickly within days or after weeks in the ICU. Early postoperative prediction of mortality in these patients is questionable but could lead to an adjustment of treatment according to outcome. Furthermore, these scoring models can be used to compare the performance of different departments or can be used to match cohorts according to their critical illness. To date the importance of the proposed scoring models in patients with RAAA is not fully clear and no competitive day by day analysis has yet been performed. The aim of the present study was to evaluate SOFA, SAPS II, and TISS-28 measurements as predictors of survival in a surgical ICU in patients with RAAA treated by OSR.
Methods
Setting
This retrospective study was conducted at the Division for Vascular Surgery, Department of Surgery, University Hospital Leipzig, Germany. The ICU provides 58 beds exclusively for surgical patients and is guided by the Department of Anesthesiology and Intensive Care Medicine. The study was reviewed and approved by the ethics committee of the University of Leipzig.
Patients
Sixty-eight patients with RAAAs were treated between February 2005 and June 2013. Only those patients who underwent OSR and were treated in the ICU were included in this study. Hence, 11 patients had to be excluded because 6 of them underwent EVAR and 5 died before they could be admitted to the ICU. The medical records were reviewed retrospectively based on clinical characteristics and outcome. The rupture of the abdominal aorta was assured according the operative report. One-year follow up was carried out retrospectively and 9 patients (14.5%; 95% confidence interval, 5.8–23.3) were lost to follow up.
Data collection
The data collection in the ICU was performed by a clinical information system (Copra System GmbH, Sasbachwalden, Germany). Medical work and care duties were captured in the electronic records along with automatically collected data from ventilators, vital signs and infusion systems. The SAPS II, SOFA and TISS-28 scores were assessed prospectively in our study. Every score was calculated on a daily basis at 6 a.m. in the morning. According to the standard documentation process, SAPS II and TISS-28 were also scored on the day of admission. Furthermore, the highest score during the stay in the ICU and the mean score were determined for each model. An ‘increase or no change’ was defined as any assessed score higher than or equal to the initial score within 48 h. A ‘decrease’ was determined as any score below the initial score within 48 h.
Scoring models
The SOFA score (0–24) is based on six different organ systems: PaO
2/FiO
2 for respiratory failure; creatinine level or urine output for renal failure; bilirubin for liver failure; Glasgow Coma Scale (GAS) for neurological status; platelet count for coagulation; and mean arterial pressure or administration of vasopressors for cardiovascular system [
15]. The SAPS II score (0–163) include 17 variables composed by 12 physiological variables, age, type of admission and three different underlying disease variables [
17]. The TISS-28 score (1–78) derive from 28 therapeutic activities performed on the ICU subdivided into 7 groups: basic activities, ventilatory support, cardiovascular support, renal support, neurological support, metabolic support, and specific interventions [
16].
Statistical analyses
Statistical analysis was performed using SPSS (version 20; SPSS, Inc., Chicago, IL, USA). Continuous variables are presented as median and categorical values as percentages. The confidence interval (CI) was determined at 95%. The chi-square-test with Yates’ correction was applied to test univariate differences between dichotomous variables. Continuous variables between survivors and non-survivors were assumed to be non-normally distributed and were compared using a non-parametric Mann–Whitney U-test. To assess the discriminative power of the different scores to predict whether the patient will survive or die, the area under the receiver operating characteristic curve (ROC) was calculated. In addition, the Youden index
J was defined to capture an optimal cut-off point for each score and point in time [
20]:
This threshold represents the point with the highest sensitivity and specificity. Graphically, is the maximum vertical distance between the ROC curve and 45-degree diagonal line. Furthermore a univariate analysis was carried out to evaluate the link between in-hospital mortality and the scoring model. A multiple logistic regression analysis was performed to evaluate a possible independent effect of significant factors detected in the univariate analysis. A selection of predictive variables was done by an automatic stepwise procedure in a forward–backward mode, and those with a significance <0.10 were entered into the multiple analysis. The correct classification rate (CCR) for the best model was reported. A P <0.05 was defined as significant.
Discussion
The present study evaluated scoring models in patients with RAAA after OSR and demonstrated their feasibility in the environment of a surgical ICU. SOFA, SAPS II and TISS-28 scores were able to predict mortality within 48 h after surgery even though there were differences in terms of sensitivity, specificity and trend analysis. To the best of our knowledge, this is the first competitive day by day analysis of the presented scoring systems for RAAA including trend analysis. Moreover, for the first time, TISS-28 was evaluated in patients with RAAA.
The tested scores were developed with assistance of statistical modeling technique [
15‐
17]. In contrast, older scores like APACHE or APACHE II were built by a subjective method as experts selected and weight variables from a panel to compose these scores [
21,
22]. Nevertheless, especially APACHE II and Multiple organ dysfunction score showed its applicability in RAAA, but with variance in discrimination to predict outcome [
23‐
26]. The comparison between these studies is difficult due different designs and clinical settings. Thus, local customization and validation of scores is even more important and can improve discrimination power [
27,
28]. Interestingly, same parameters are considered in different scores. For instance, parameters of APACHE II like gas exchange and acid–base balance (pH, HCO3, PaO2, FiO2, creatinine, potassium, age, systolic blood pressure) are assessed by SAPS II. The presented scores were validated mainly in surgical patients [
18,
27,
29], but their applicability to the vascular field are unclear. Therefore, the evaluation in RAAA is essential because of individual differences from score to score.
The proposed scoring models are based on postoperative parameters. They can not only be used to predict outcome, even critical illness can be tracked day by day. Hence, they are in sharp contrast to GAS, Edinburgh Rupture Aneurysm Score (ERAS) or Hardman Index [
13,
30,
31]. These scores are composed by preoperative parameters and can be calculated in the emergency department to determine the likelihood whether a patient will survive or die. Interestingly, these scores were developed exclusively in patients with RAAA.
The discriminative power to distinguish between survivors and non-survivors according the ROC calculation after surgery was highest for mean SOFA followed by mean SAPS II and mean TISS-28. The highest values were found in means, what can be explained by a higher numbers of values, which were considered in the statistical analysis. Previous studies confirm the results for SOFA scores in critically ill patients (AUC range: 0.69 to 0.92) [
29,
32]. The scoring systems appear to have a trend towards outcome within the first 48 h. In consideration of the serial measurements of Figures
1,
3 and
4, the values increased for survivors and decreased for non-survivors. A trend analysis was carried out to improve the discriminative power of cut-off values in scoring models and SAPS II showed the best performance. SAPS II (cut-off value >45) had a significant increase of almost 30% mortality when the score increased or did not change and a significantly lower mortality of 25% when the score decreased. In the trend analysis of SOFA, only a slightly higher mortality (8%) was found. This might be affected by the assessment at 6 a.m. according to our documentation policy. Potentially dynamic changes of SOFA score may arise directly after surgery, which would be consistent with recently published studies [
32,
33]. In contrast, the threshold of TISS-28 was able to predict mortality (29% vs. 55%) but changes in that value could not improve sensitivity and specificity. Overall, TISS-28 was inferior to SOFA and SAPS in terms of ROC calculation and predicting outcome.
Since 2005, all three investigated models were assessed semi-automatically in our clinical information system. The widespread availability of electronic devices in the ICU environment led to the recording of vital signs, medication, care duties and medical work. Patients with RAAA represent a high-risk group for complications and mortality. Although the implementation of scoring systems need additional effort, it may allow inter-individual decision-making and may provide information for relatives.
New generations of ICU scoring systems are promising, e.g. APACHE III or SAPS III, and of high interest for future studies [
27,
28]. However, the present study assessed the evaluated scores prospectively during the clinical routine and a consideration of additional new scores would include a major confounder in terms of a retrospective calculation prone to missing values.
There are limitations in the present analysis. The most important are the retrospective review of the medical records and the underpowered number of patients. Due to the mortality rate over time, the numbers of patients and measurement drops in relation to time. Thus, an analysis of a certain time point after one week might not be substantial. The reported in-hospital mortality of 41.7% is consistent with previous reports. Reimerink et al. reports a mortality rate of 49% (45% to 55%) in a meta-analysis [
34]. Patient co-morbidities of both groups did not differ significantly. The median age of the present cohort was 75.9 (IQR; 64.6 – 80.7), what is consistent to large epidemiological studies [
34]. The survivor group was younger than the non-survivor group (72.1 years vs. 80.9 years). An advanced age is a negative predictor for survival [
35], but age is not considered in the SOFA and TISS-28. Strikingly, these scores were able to predict patients, who will die and will be older. In a recent study, SAPS II was able to predict mortality in patients >90 years [
36]. Of note, the tested scores are applicable in the elderly.
Patients treated by EVAR were excluded from this study to test the scores exclusively in patients’ who underwent open surgery, as a single treatment option. Clearly, patients who underwent surgery or endovascular repair have major differences in postoperative morbidities and physiological changes [
7,
37,
38]. Moreover, patients who undergo EVAR are selected and biased. EVAR suitability is determined by the anatomic configuration of the aortic neck and iliac arteries, while OSR is not limited by the aneurysm morphology. Thus, patients, who are not suitable for EVAR, usually undergo surgery, as reflected in observational studies [
39,
40]. Therefore, these patients were excluded from our analysis.
Serial measurements of scoring models have a big impact on the assessment of critically ill patients. Most ICU’s are directed by anesthesiologist and primary sections like vascular surgery are losing influence in terms of the handling and assessment of these scores. Therefore, the evaluation of widely used scores in vascular surgical patients is indispensable to ensure an objective assessment of vascular patients in the ICU and to avoid interpreting scores on the basis of a heterogeneous patient cohort.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
FK and PTF conducted the study design. GW and HMH reviewed the medical history of all patients and revised critically the manuscript for important content. MS and FK performed the statistical analysis. FK wrote the manuscript. PTF and UXK supervised the study, interpreted data and have given final approval for publication. SJ and IM helped writing the manuscript and revised critically the manuscript. All authors discussed the results and approved the final manuscript.