| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Guest Access | Sign In via User Name/Password
|
|
|
|||||||
Guest Access | Sign In via User Name/Password |
|||||||||
* From the Hospital Universitario de Canarias, Critical Care Department, Universidad de La Laguna, Tenerife, Spain.
Correspondence to: Nicolás Serrano, MD, PhD, FCCP, Hospital Universitario de Canarias, 38320-La Laguna, Tenerife, Spain; e-mail: nserrano{at}epicure.org
 |
Abstract |
|---|
 TOP Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Design: Prospective observational study.
Setting: A 24-bed ICU in a tertiary referral university hospital.
Patients: Five hundred sixty-nine patients undergoing CABG surgery.
Interventions: Variables of the ICURSS model were recorded at ICU admission. Extubation was performed according to a standard protocol. Patients remaining intubated within 8 h after ICU admission were designated as having early extubation failure (EEF). The next evaluations at 16, 24, 48, 72, and 96 h designated patients as having a prolonged intubation period (PIP) and prolonged mechanical ventilation (PMV) for 24, 48, 72, and 96 h. The ability of the ICURSS model to predict extubation failure at different cutoff values was measured using the Hosmer-Lemeshow goodness-of-fit test and the area under the receiver operating characteristic curve.
Measurements and results: Prolonged intubation rates were as follows: EEF, 40.2%; PIP, 17.2%; PMV for 24 h, 10.4%; PMV for 48 h, 7.6%; PMV for 72 h, 6.5%; and PMV for 96 h, 6.0%. At every cutoff, the ICURSS showed poor discrimination to predict the failure to be extubated. Calibration was also poor, although some ability to predict both EEF and PMV at 
48 h was shown.
Conclusions: Prolonged intubation rates after undergoing CABG surgery in our setting were comparable with those of other reports from institutions where fast-track cardiac anesthesia is currently in practice. In our cohort, the ICURSS was not useful for the prediction of length of intubation.
Key Words: calibration • cardiac surgery • coronary artery bypass grafting • discrimination • extubation failure • goodness-of-fit test • Hosmer-Lemeshow 
2 statistic • ICU • ICU risk stratification score • mechanical ventilation • receiver operating characteristic curve • weaning
|
Introduction |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Although the exact definitions for either early extubation failure (EEF) or prolonged mechanical ventilation (PMV) can be controversial, both situations occur frequently after cardiac surgery. The importance of specifically defining EEF and PMV after CABG surgery may be of interest, because the time intervals following surgery may be associated with different reasons, related in turn with different causes, risk factors, and outcomes.7
In addition, different scoring systems to predict weaning failure after cardiac surgery have been proposed.89 The main finding from these studies has been that postoperatively collected data using established scoring systems, either general101112131415 or specific,1617 as well as documented events of high clinical impact for risk assessment and quality control are reliable predictors of prolonged ventilation.9
Among the more specific models intended to predict complications following cardiac surgery, morbidity prediction after CABG procedures has been widely studied by Higgins and colleagues.18192021 They proposed the ICU Risk Stratification Score (ICURSS), also currently known as the Cleveland Score, which was intended to predict both ICU mortality and morbidity after CABG surgery at the time of ICU admission.19 The ICURSS is a well-known model that includes the following 13 variables: age; body surface area; number of previous cardiac operations; history of previous operation or angioplasty for peripheral vascular disease; preoperative serum creatinine and albumin levels; number of minutes using cardiopulmonary bypass; the use of an intraaortic balloon pump after the cardiopulmonary bypass; and, at ICU admission, heart rate, cardiac index, central venous pressure, arterial bicarbonate level, and the alveolar-arterial oxygen pressure gradient (in millimeters of mercury) [Fig 1 ].
|
|
Materials and Methods |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Airway and ventilatory management was carried out according to a standard extubation protocol based on the clinical judgment of the ICU medical team. Basically, patients met standard extubation criteria if they were hemodynamically stable (ie, normotension, heart rate of < 120 beats/min, and no signs of low cardiac output or myocardial ischemia) and showed no evidence of early surgical complications, such as significant bleeding. Once major pulmonary complications were ruled out by means of auscultation and chest roentgenogram, mechanical ventilatory support was decreased from the assist-control volume-cycled mode to pressure support ventilation mode. Lung function was considered to be satisfactory if a tidal volume of > 5 mL/kg at a positive end-expiratory pressure level of 5 cm H2O rendered a PaO2 of > 80 mm Hg and a PaCO2 of < 45 mm Hg at fraction of inspired oxygen of 0.4, with a spontaneous respiratory rate < 30 breaths/min. The criteria used for deciding whether to extubate were mainly supported by clinical judgment regarding the ability of the patient to tolerate the change from intermittent positive-pressure ventilation to pressure support ventilation modes, and further decreases of pressure support levels along with hemodynamic stability and lack of major complications. Finally a T-piece was placed 15 to 30 min before extubation, and muscle strength, spontaneous ventilation, and level of consciousness were assessed again in each patient before deciding whether to extubate. When patients had a T-piece placed, the decision to extubate had been already made. The T-trial was mainly used to confirm this decision, and this maneuver was aborted only in those cases where it was unsuccessful, with the patient again receiving mechanical ventilation as a result.
Attending physicians shared similar criteria and aimed to extubate the patient as early as was clinically possible, but they were kept blind to the goals of the study at that time. Pain control was continued after leaving the OR by means of the titrated continuous IV infusion of fentanyl or, more recently, remifentanil. Once in the ICU, IV analgesia was usually continued until after extubation.
Occasionally, patients were managed by means of epidural analgesia. In these cases, the potential effects on either the central or the muscular control of respiratory function were carefully monitored, as well as the hemodynamic effects (hypotension). If they were present in any case, the infusion would be stopped as a result.
In rare instances, patients arrived in the ICU after they had been extubated in the OR. In such cases, the time of intubation was computed as equal to 0 h, and the patient was considered to have been successfully extubated if reintubation was not needed at another time. Those patients who remained intubated within the first 8 h after ICU admission and were judged as being unable to be extubated at that time were designated as having EEF. The next evaluations were carried out at 16, 24, 48, 72, and 96 h after ICU admission. Thus, patients who remained intubated after 16 h were designated as having a prolonged intubation period (PIP). After the first 24 h, the situation was defined as PMV, and different time cutoffs that were used to define this condition at 24, 48, 72, and 96 h.
We defined patients who died during the evaluation period (ie, 0 to 96 h) would be considered as having extubation failure for every cut point, independently of the moment of death and the number of hours of intubation received. Additionally, in every patient needing reintubation, the time of intubation was computed as a continuum from ICU admission to the last successful extubation. Although these subgroups may add up to only a few cases, this approach might skew the outcomes, and the results obtained including and not including data for these patients could differ. As a result, the data were analyzed and compared both including and not including the data for these patients.
The dependent variable was the dichotomous condition of successful extubation (ie, yes or no) present at each one of the six different time cutoffs listed above. Each outcome variable was matched against the ICURSS values that were calculated in every patient after ICU admission. The predictive ability for extubation failure of the ICURSS model at the six different cutoff points was calibrated using the Hosmer-Lemeshow (HL) goodness-of-fit test.22 To use the HL statistic, the expected number of events in most groups must exceed five, and none of the groups have expected values of < 1. A lower HL 2 statistic means better performance of the test. On the other hand, if statistically significant differences between expected and observed outcomes were found, it would mean poor calibration of the model as a result. Consequently, patients were ranked according their ICURSS values and were distributed across n groups, with n –2 degrees of freedom (df).
Discrimination was measured using the area under the curve (AUC) of the receiver operating characteristic (ROC).23 According to the AUC ROC values, discrimination was categorized as "perfect" (AUC, 1), "good" (AUC, > 0.8), "moderate" (AUC, 0.6 to 0.8), or "poor" (AUC, < 0.6).24
Statistical analyses were performed using a statistical software package (SPSS, version 12.0.1 for Windows; SPSS Inc; Chicago, IL). The data are presented as the median and the mean ± SD. To compare continuous variables between two groups, median (interquartile range), Student t test, and Mann-Whitney test were used. A p value of < 0.05 was considered to be statistically significant.
|
Results |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Figure 2 shows the rate of patients remaining intubated after ICU admission. EEF at 8 h occurred in 229 patients (40.2%). In these patients (Table 2 ), ICURSS points were significantly higher than in those 340 patients who had been successfully extubated before 8 h in their ICU stay, and whose mean duration of intubation was 5.58 ± 1.46 h. The results of the same procedure using the cutoffs defined by PIP at 16 h and PMV at 24, 48, 72, and 96 h are also shown in Table 2.
|
|
2 statistic value means better performance of the test. A p value of < 0.05 means that there will be statistically significant differences between the expected and observed outcomes and, as a result, poor calibration of the model. The values of the HL 2 statistics are shown in Table 2. A better performance for ICURSS calibration is found at extreme cutoff values (ie, EEF at 8 h and PMV at 48 h). By contrast, the ability of ICURSS to predict intermediate situations, as defined by PIP and PMV at 24 h, did not reach statistical significance. The results of the discriminations obtained from the calculation of the AUC ROC also were analyzed. At every cutoff value, the AUC ROC ranged from 0.602 to 0.676. Since the 95% confidence interval of the AUCs includes values of < 0.6, the AUC is consistently poor in the current study, giving the ICURSS a lesser ability to discriminate in predicting weaning failure after CABG surgery (Table 2).
|
Discussion |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In previous years, several scoring systems based on clinical criteria have been proposed to predict failure to wean from mechanical ventilation in postoperative cardiac surgical patients.91617 Simultaneously, these previous studies described their rates of extubation failure, and our study has been successfully compared with those rates. Similarly, we have also utilized an ICU scoring system that is designed to evaluate morbidity and mortality after CABG surgery to determine its value as a guide for predicting early tracheal extubation.
In 1996, Spivack et al16 developed a specific scoring system that was intended to predict the risk of PMV based on several preoperative risk factors in CABG patients. However, only reduced left ventricular ejection fraction was identified as a predictive risk factor for PMV.16
In addition, Wong et al17 in 1999 proposed a cardiac risk score that was based on preoperative as well as postoperative risk factors to study the risk factors for delayed extubation (ie, > 10 h) in patients undergoing FTCA. The risk factors identified were increased age, female gender, postoperative use of intraaortic balloon pump, inotropes, bleeding, and atrial arrhythmia.17
Kern and coworkers,9 in 2001, used the general severity of illness scoring systems the Acute Physiology and Chronic Health Evaluation (APACHE) II,10 the Simplified Acute Physiology Score II,12 Therapeutic Intervention Scoring System,1525 and the Organ Failure Score26 to predict the risk for PMV in patients after cardiac surgery. They concluded that postoperatively collected data using common severity-of-illness scoring systems resulted in a more sensitive model than preoperative and intraoperative data.9
In all of the above-mentioned studies, postoperative data have been shown to be more predictive for this purpose than preoperative or intraoperative data.91617 In accordance with this information, a postoperative ICU admission score was chosen for use in the present study. The ICURSS had been already proposed in 1997 by Higgins and colleagues19 for predicting morbidity and mortality risk after CABG surgery. It is currently a well-known model for predicting morbidity at the time of ICU admission. It uses the most easily acquired patient data, and, in clinical use, has proven to be the most practicable.27 Although the developers of the ICURSS included PMV as an outcome variable that was pooled with other serious ICU morbidity conditions such as stroke, low cardiac output, myocardial infarction, serious infection, or renal failure, to our knowledge, the ICURSS system has not been previously tested to specifically predict failure to be extubated after CABG surgery.
Interestingly, the rates we have found at our institution for weaning failure and mortality after CABG surgery are comparable to those in previous reports. Thus, 9% of the patients studied by Kern and colleagues9 received mechanical ventilation for > 48 h. In the series by Spivack et al,16 PMV and death were considered to be rare events (8.3% and 2.0%, respectively). In our study, the death rate was < 1.5% (ie, 8 of 569 patients), and the rate of PMV for > 48 h was 7.6%, which is similar to the previously reported data. In addition, of the 885 patients studied by Wong and colleagues,17 25% had delayed extubation. This condition, designated as PIP in our study, was present in 17.2% of our patients (Fig 2). The same authors17 also found female gender as a significant risk factor. However, the discordance between the number of male and female patients (male patient/female patient ratio, 2.74) in our population was similar to that reported by Wong et al,17 in which 76% of patients were male and 24% were female, and the male patient/female patient ratio was 3.17. Thus, regarding gender, our incidence of delayed extubation may not have been underestimated.
While in our work the term extubation failure was intended to mean the failure to be extubated, other authors have defined extubation failure as the requirement for reintubation and mechanical ventilation after prior successful weaning from ventilatory support and extubation. Thus, Rady and Ryan,28 in a later article using the same database as that used by Higgins et al,19 defined extubation failure in this latter way. The frequency of extubation failure thus considered was found to be 6.6%, which is quite superior to the 1.5% of patients who were weaned from mechanical ventilation after cardiac surgery and required reintubation and ventilatory support in the present study.
Because the rate of extubation failure has been analyzed by a number of previous studies, now including ours, the main interest of the current study was in describing the performance of the ICURSS in predicting extubation failure. The mean values of the ICURSS were nearly identical at all cutoff times for the two groups considered (ie, extubated and nonextubated patients) [Table 2]. We used the ROC analysis and HL statistic to evaluate the performance of the ICURSS in predicting extubation failure. The AUC ROC clearly showed that the ICURSS performs poorly in discriminating extubation failure from nonfailure at every cutoff point. As a result, our study was clearly a negative study.
The calibration of the model was also poor. The ICURSS model was only able to predict those patients who were quickly extubated and those who were likely to fail extubation for > 48 h. In this sense, better calibration was found at lower and higher cutoff times (ie, to predict either EEF or PMV at 48 h). This may explain the fact that the ICURSS can accurately predict those patients who will be successfully extubated within the first 8 h from among those who are in good condition at ICU admission, and also can accurately predict those patients who will remain intubated for > 48 h from among those who are in very bad condition at ICU admission. However, the calibration may also yield false-positive results for these cutoff times because of the small number of patients in the failure group.
Conversely, the worst prediction ability was found in intermediate situations in which extubation might occur between 8 and 48 h after ICU arrival. As a result, we can consider situations that include cutoff times at 16 h (PIP) and 24 h (PMV at 24 h) to be in a "gray area," in which the ICURSS has a poorer calibration ability. Accordingly with Yende and Wunderink,7 the time intervals may be associated with different reasons, which are related in turn to different causes, risk factors, and outcomes,7 and accurate predictions of weaning failure at these cutoffs have been found to be more difficult to estimate using the model.
Our study has important limitations. First of all, although characteristics of our population and the results may be comparable to those in other previously cited reports, the sample size of our study is limited. In addition, patients were recruited at a single institution with a defined extubation policy. As a result, ability of the ICURSS to predict weaning failure should be tested and validated at each institution, and a multiinstitutional study would be also desirable.
Second, analyses for the six outcomes are not independent, and we assumed no special problems in making statistical inferences or in evaluating logistic regression procedures when multiple tests were performed. We evaluated six sets of calibration and discrimination analyses in which each one of the six outcomes was expressed as a dichotomous variable in an independent way. Although the model was unable to explain why it had failed to predict extubation failure in the most critical group of patients, a post hoc case-by-case analysis that included a new set of retrospective data might appear to have been desirable, but it may have been inappropriate since this was a prospective study. Undoubtedly, this approach would provide a better scoring system or would yield important modifications to the current model.
Third, we voluntarily introduced a methodological bias to consider dying patients as experiencing extubation failure for every fixed cutpoint, but probably the low death rate in our series minimized its impact. And last, but not least, we are also aware of another bias in which the cases of the patients who needed reintubation were considered using the whole time of intubation that elapsed from ICU admission to the last successful extubation, independently of the time interval for which the patient was extubated prior to reintubation. Similarly, the low rate of reintubation in our sample precluded the impact of this bias as well.
Thus, we conclude that the rates of failure to wean from mechanical ventilation at different time intervals after CABG surgery at our institution are comparable with those reported from institutions where FTCA is currently in practice. The performance of the ICURSS system at the time of ICU admission to predict weaning failure as a form of morbidity after CABG surgery has been evaluated. Our study has shown that the ICURSS performs poorly in predicting extubation outcome. Although extreme conditions such as EEF and PMV after 48 h may be explained by the ICURSS, its utility in predicting these situations cannot be considered.
|
Footnotes |
|---|
Received for publication December 3, 2004. Accepted for publication February 1, 2005.
|
References |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  F. Filsoufi, P. B. Rahmanian, J. G. Castillo, J. Chikwe, and D. H. Adams Predictors and Early and Late Outcomes of Respiratory Failure in Contemporary Cardiac Surgery Chest, March 1, 2008; 133(3): 713 - 721. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. K. Jiricka Ask the Experts. Crit. Care Nurse, June 1, 2006; 26(3): 70 - 72. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
