Background

There is no widely accepted consensus on the weaning and extubating protocols for neurosurgical patients, leading to heterogeneity in clinical practices and high rates of delayed extubation and extubation failure−related health complications.

Methods

In this single-center prospective observational diagnostic study, mechanically ventilated neurosurgical patients with extubation attempts were consecutively enrolled for 1 yr. Responsive physicians were surveyed for the reasons for delayed extubation and developed the Swallowing, Tongue protrusion, Airway protection reflected by spontaneous and suctioning cough, and Glasgow Coma Scale Evaluation (STAGE) score to predict the extubation success for neurosurgical patients already meeting other general extubation criteria.

Results

A total of 3,171 patients were screened consecutively, and 226 patients were enrolled in this study. The rates of delayed extubation and extubation failure were 25% (57 of 226) and 19% (43 of 226), respectively. The most common reasons for the extubation delay were weak airway-protecting function and poor consciousness. The area under the receiver operating characteristics curve of the total STAGE score associated with extubation success was 0.72 (95% CI, 0.64 to 0.79). Guided by the highest Youden index, the cutoff point for the STAGE score was set at 6 with 59% (95% CI, 51 to 66%) sensitivity, 74% (95% CI, 59 to 86%) specificity, 90% (95% CI, 84 to 95%) positive predictive value, and 30% (95% CI, 21 to 39%) negative predictive value. At STAGE scores of 9 or higher, the model exhibited a 100% (95% CI, 90 to 100%) specificity and 100% (95% CI, 72 to 100%) positive predictive value for predicting extubation success.

Conclusions

After a survey of the reasons for delayed extubation, the STAGE scoring system was developed to better predict the extubation success rate. This scoring system has promising potential in predicting extubation readiness and may help clinicians avoid delayed extubation and failed extubation–related health complications in neurosurgical patients.

Editor’s Perspective
What We Already Know about This Topic
  • There is no widely accepted consensus on the weaning and extubating protocols for neurosurgical patients, leading to heterogeneity in clinical practices and high rates of delayed extubation and extubation failure

What This Article Tells Us That Is New
  • Based on the causes for delayed and/or failed extubation, the authors developed a scoring system to better predict the extubation success that may facilitate decisions regarding extubation management in neurosurgical patients

Neurosurgical patients admitted to the intensive care unit (ICU) often require artificial airway and/or ventilation support due to the inability of airway protection or other spontaneous or iatrogenic reasons affecting spontaneous breathing.1–3  The patient’s response to early initiation of weaning, followed by successful extubation, indicates effective mechanical ventilation and better outcomes. Prolonged mechanical ventilation, repeated failed extubation, and subsequent reintubation attempts in the ICU are associated with adverse clinical outcomes, higher morbidity, and longer hospital stays.4–7  Therefore, an efficient clinical strategy is urgently warranted to predict the right timing for effective weaning and subsequent extubation steps to avoid unnecessary use of mechanical ventilation, as well as minimizing the health risks associated with failed weaning and repeated extubation or intubation processes.

Clinical practice guidelines in general critically ill patients recommend a well defined, albeit imperfect, protocolized weaning and extubation procedure, including evaluation of weaning readiness, spontaneous breathing trial assessment, extubation, and consideration of prophylactic noninvasive ventilation or high-flow nasal oxygen.8,9  However, the lack of strong clinical evidence has been a roadblock in establishing a globally accepted weaning procedure, including the extubation step,10  which often leads to heterogeneity in clinical outcomes of mechanical ventilation, with an increasing rate of extubation failure in neurosurgical patients.11  Notably, in some neuro-ICUs, patients are extubated while passing spontaneous breathing trials, regardless of their neurologic status and airway-protecting function.12–14  However, the decision to extubate usually relies on the clinical judgment of the responsible attending physician.2,15,16  Consequently, the rate of delayed extubation is relatively higher in neurosurgical patients than in general critically ill patients.15,17  Although several grading models with solid methods and reasonable results have been proposed,13,18,19  a predictive scoring system focusing particularly on the level of consciousness and airway-protecting function with simplicity to perform at the bedside is still lacking. Therefore, further research on developing strategies for effective extubation in this subset of the patient population is strongly required.

In this study, we surveyed physicians’ opinions and reasons for their decision on delayed extubation in a cohort of prospectively enrolled mechanically ventilated neurosurgical patients. In addition to routine extubation practice at the physician’s discretion, we developed a diagnostic scoring system for the assessment of extubation readiness, termed the Swallowing, Tongue protrusion, Airway protection reflected by spontaneous and suctioning cough, and Glasgow Coma Scale Evaluation (STAGE) score, and we evaluated the usefulness of this scoring process in predicting extubation success in neurosurgical patients.

Study Population and Ethics

Every admission was consecutively screened at two ICUs having 70-bed capacity at Beijing Tiantan Hospital, Capital Medical University (Beijing, China) from November 1, 2020, to October 31, 2021. The study protocol (KY2022-063-02) was approved by the Institutional Ethics Committee. Written informed consent was obtained from either the patients or their legal representatives. Neurosurgical patients who were subjected to invasive mechanical ventilation longer than 24 h were eligible. Neurosurgical patients were defined a priori as patients with cerebral tumor, head trauma, or subarachnoid or intracerebral hemorrhage undergoing craniotomy. Patients with extubation attempts were finally included. Excluded subjects had: (1) age less than 18 years, (2) pregnancy, (3) existing spinal cord injury, (4) extubation in association with the withdrawal of life-sustaining therapy, or (5) no extubation attempt during their ICU stay (reasons documented). There was no repetition in the patient inclusion process. This study followed the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) Statement for observational studies.20 

Routine Practice of Weaning and Extubation

In our center, routine practice proceeds as follows: a four-step daily weaning and extubating process in collaboration with the clinicians and respiratory therapists:

Step I: Checking Readiness for Initiation of the Weaning Process

The following aspects were evaluated by clinicians: (1) improvement of the cause of intubation; (2) no signs of intracranial hypertension or brain swelling; (3) hemodynamic stability; (4) positive end-expiratory pressure of 5 cm H2O or less; (5) partial pressure of alveolar oxygen/fraction of inspiration oxygen (Pao2/Fio2) of 200 mmHg or more; and (6) no planned surgery under general anesthesia within the ensuing 72 h. If all aspects were satisfied, the patient was moved to step II for spontaneous breathing trial assessment.

Step II: Spontaneous Breathing Trial

Spontaneous breathing trial was conducted once daily by respiratory therapists using the low-level pressure support ventilation (pressure support of 8 cm H2O or less, positive end-expiratory pressure of 5 cm H2O or less) or T-tube for 30 to 120 min. Failure of spontaneous breathing trial might include: (1) respiratory rate (RR) of greater than 35 breaths/min for more than 5 min; (2) oxygen saturation measured by pulse oximetry (Spo2) of less than 90%; (3) heart rate (HR) of greater than 140 beats/min or a sustained change in HR of more than 20%; (4) systolic blood pressure greater than 180 mmHg or less than 90 mmHg; and (5) signs of anxiety, agitation, or diaphoresis.8,9  If the spontaneous breathing trial was passed, the patient was moved to step III for extubation. For unqualified patients in step I or II, we re-evaluated them from step I on the next day.

Step III: Extubation Decision

For neurosurgical patients, extubation was performed only after reaching a consensus between ICU physicians and neurosurgeons. Otherwise, the patient was re-evaluated the next day from step I.

Step IV: Postextubation Assessment

After extubation, oxygen was delivered through Venturi masks. The flow of gas was adjusted to maintain Spo2 greater than or equal to 92%. If any sign of respiratory distress occurred, noninvasive respiratory support would be used under the supervision of the responsible physicians. Reintubation was needed when: (1) RR was greater than 35 breaths/min for more than 5 min; (2) Spo2 was less than 90%; (3) HR was greater than 140 beats/min, or there was a sustained change in HR greater than 20%; (4) Pao2 was less than 80 mmHg with a Fio2 of 50% or more; (5) Paco2 was greater than 45 mmHg, or there was a change in Paco2 of 20% or more after extubation, with a pH of less than 7.33; or (6) the patient exhibited signs of respiratory muscle fatigue or increased work of breathing.8 

Data Collection and Outcome Measures

Demographics and baseline data were collected for all enrolled patients. Physiologic parameters were recorded on the first day of mechanical ventilation and from the first successful spontaneous breathing trial to extubation, including vital signs, total and subscores on the Glasgow Coma Scale (verbal component was deemed as 1), ventilator modes and parameters, 24-h fluid input and output, blood gas analysis, and the use of sedatives and analgesics. In addition, recorded clinical outcome parameters were as follows: number of extubation failures, tracheostomy rate, duration of mechanical ventilation, length of stay in ICU and hospital, nosocomial pneumonia,21–23  mortality rate, and costs. Details of data collection are shown in Supplemental Digital Content 1: Supplemental Text 1 (https://links.lww.com/ALN/D259). All patients were routinely followed up until their hospital discharge, death, or 60 days postenrollment, whichever occurred first.

Extubation Decision Survey

For each of the spontaneous breathing trial-qualified patients, a survey on the extubation decision from respective ICU physicians and neurosurgeons was conducted by one of the investigators who was not involved in the decision-making process. The reasons for delayed extubation were documented daily as long as the patient was not extubated after a successful spontaneous breathing trial. The questionnaire for the survey of reasons for the delay can be found in Supplemental Digital Content 2: Supplemental Text 2 (https://links.lww.com/ALN/D260).

Derivation of the Model Items through the Nominal Group Technique

First, we employed the nominal group technique to carefully select model items. Subsequently, utilizing the data gathered, we assigned a value to each of these selected items through the application of the multiple logistic regression method.

Before the study, a nominal group of 17 experts from 13 provinces covering high-, middle-, and low-income regions was organized by the National Center for Healthcare Quality Management in Neurocritical Care. This group has been devoted to neurocritical care quality control since 2018 and has developed key performance indicators for neurocritical care quality control. The nominal group consisted of six neurointensivists, four general intensivists, three neurosurgeons, one neurologist, two respiratory therapists, and one neurocritical care nurse all with working experience exceeding 15 yr.

Relevant items of consciousness and airway-protecting function were listed after the literature review and clinical consultation. After five rounds of online face-to-face nominal group meetings and one round of comments and iterative review,24  four airway-protecting functions assessments consisting of swallowing, tongue protrusion, and spontaneous and suctioning cough, as well as one consciousness assessment (motor response in the Glasgow Coma Scale) reached a consensus. The nominal group methods and consensus results are presented in Supplemental Digital Content 3: Supplemental Text 3 (https://links.lww.com/ALN/D261). The assessment criteria were as follows:

  1. Swallowing

    • Strong: No visible accumulation of saliva for either conscious or comatose patients

    • Poor: Visible accumulation of saliva for either conscious or comatose patients

  2. Tongue protrusion

    • Strong: Can protrude outside the mouth

    • Poor: Cannot protrude outside the mouth

  3. Spontaneous cough:

    • Strong: Vigorous

    • Poor: Weak or none

  4. Suctioning cough

    • Strong: Vigorous

    • Poor: Weak or none

  5. Motor response in Glasgow Coma Scale

    • Strong: Can obey commands or localize the pain

    • Poor: Withdrawal, flexion, extension, or no response to pain

After the first successful spontaneous breathing trial, assessments of the five items were performed daily by two respiratory therapists, according to the criteria shown above, and the results were recorded without informing clinicians to avoid interfering with clinical practice. In clinical practice, the extubation decision was made based on the four-step assessment by the physicians in charge who were blinded to the study. After the data collection phase, the STAGE model was developed by assigning specific scores to five selected items, which were determined through the multiple logistic regression analysis.

Definitions of Delayed Extubation and Extubation Failure

Timely extubation was defined as extubating within 24 h of the first successful spontaneous breathing trial; otherwise, the extubation was recognized as delayed.15,25  Extubation failure was defined as reintubation within 72 h of a failed extubation;25  otherwise, it was deemed as an extubation success.2,15 

Statistical Analysis

Parametric comparisons were performed between the extubation success and failure groups. Continuous variables were tested for their normal distribution and presented as mean and SD or median (25th, 75th percentiles), as appropriate. Comparisons of continuous variables were performed by Student’s t test in case of a normal distribution or by the Mann–Whitney U test for non-normal distribution. Categorial variables were presented as numbers and percentages and analyzed by the chi-square or Fisher’s exact test. The missing data in our study were minute ventilation, tidal volume (VT), and rapid shallow breathing index for 141 patients using the T-tube on the extubation day. Each of the three parameters had 141 missing data. The missing data rate was 62% (141 of 226) in all the enrollments. All three parameters had 62% (113 of 183) and 65% (28 of 43) missing data rates in the extubation success and failure groups, respectively. Considering the equal distribution of missing data, we choose not to replace them.

A multiple logistic regression model, including five items derived through the nominal group technique, was constructed to predict the extubation success. Results are presented with an odds ratio and 95% CI. The score for each item was given to the nearest integer based on the weighting of odds ratio values, thus creating the STAGE model.13,18  The ability of the score to predict extubation success was evaluated by the receiver operating characteristic curve analysis by calculating the area under the receiver operating characteristics curve (AUC). The calibration curve was developed to assess calibration ability.26  Youden’s index, sensitivity, specificity, and positive and negative predictive values were calculated at different cutoff points of the receiver operating characteristic curve. We used the bootstrap method to perform internal validation of scores. From the original data set, multiple samples were randomly drawn with a replacement 1,000 times. The AUC of the score was corrected to avoid overoptimism.27 

Furthermore, we performed two sensitivity analyses. First, we sequentially removed one, two, three, and four items from the model and repeated the multiple logistic regression analyses. Then, we compared the AUC for each modification with that of the original model. Second, predictive values were compared between the STAGE score and Glasgow Coma Scale alone for all the patients included. We also divided patients into subgroups to test the prediction potential of our scoring system based on the extubation timing (timely or delayed) and the corresponding motor response in Glasgow Coma Scale (5 or greater or less than 5).

Data management and analyses were conducted using Stata v15.0 software (StataCorp, USA) and R 4.1.2 (R Foundation for Statistical Computing, Austria) with the val.prob package. A two-sided P value of less than 0.05 was considered statistically significant. A data analysis and statistical plan was written and filed with a private entity before the data were accessed.

A total of 3,171 patients were screened consecutively during 1 year from November 2020 to October 2021. Among those, 334 adult neurosurgical patients with mechanical ventilation for more than 24 h were eligible. According to the inclusion criteria, we excluded 2,945 patients, including 108 patients who met the readiness of the weaning processes but never had extubation attempts during the ICU stay (fig. 1). Ultimately, 226 neurosurgical patients undergoing extubation were included in the final analysis. The baseline characteristics of the enrolled patients are summarized in table 1.

Table 1.

Baseline Characteristics of Extubation Success and Failure Groups

Baseline Characteristics of Extubation Success and Failure Groups
Baseline Characteristics of Extubation Success and Failure Groups
Fig. 1.

Flowchart illustrating the study population selection process.

Fig. 1.

Flowchart illustrating the study population selection process.

Close modal

Delayed extubation occurred in 25.2% (57 of 226) of enrollments for a total number of 160 days. The median (25th, 75th percentiles) time of delay was 2 (1 to 4) days, ranging from 1 to 8 days (fig. 2). A total of 20 ICU physicians and 30 neurosurgeons were surveyed. Of 160 days for delayed extubation, 131 (82%) days were delayed solely by the ICU physicians, 13 (8%) by neurosurgeons, and 16 (10%) by both sides. According to the questionnaire of the extubation decision survey, the top two reasons provided by each side were the same: weak airway-protecting function and poor consciousness (fig. 3). No significant difference was found in the rate of extubation failure between the timely and delayed extubation groups (17.8% vs 22.8%, P = 0.437).

Fig. 2.

Delayed extubation days from the first successful spontaneous breathing trial to extubation.

Fig. 2.

Delayed extubation days from the first successful spontaneous breathing trial to extubation.

Close modal
Fig. 3.

Reasons for the delay provided by the intensive care unit (ICU) physician and neurosurgeon groups.

Fig. 3.

Reasons for the delay provided by the intensive care unit (ICU) physician and neurosurgeon groups.

Close modal

The overall rate of extubation failure was 19.0% (43 of 226). The median (25th, 75th percentiles) time of reintubation was 7 (0.5 to 34) h. The top three reasons for extubation failure were upper airway obstruction, excess airway secretions, and decreased consciousness (fig. 4).

Fig. 4.

Reasons for extubation failure.

Fig. 4.

Reasons for extubation failure.

Close modal

Compared with the extubation success group, the failure group presented a higher prevalence of cerebrovascular diseases in the past, lower Glasgow Coma Scale, and higher Acute Physiology and Chronic Health Evaluation Score (APACHE) II at ICU admission and a longer duration of mechanical ventilation before the initiation of the weaning process and the first successful spontaneous breathing trial (table 1). On the extubation day, higher RR, rapid shallow breathing index, lower Glasgow Coma Scale and motor response in Glasgow Coma Scale, and poorer airway protection functions were observed in the extubation failure group (table 2). In addition, these patients exhibited a higher rate of tracheostomy and pneumonia, longer duration on mechanical ventilation, extended ICU or hospital length of stay, and increased healthcare costs (table 3).

Table 2.

Comparison of Parameters on Extubation Day between Extubation Success and Failure Groups

Comparison of Parameters on Extubation Day between Extubation Success and Failure Groups
Comparison of Parameters on Extubation Day between Extubation Success and Failure Groups
Table 3.

Comparison of Clinical Outcomes between Extubation Success and Failure Groups

Comparison of Clinical Outcomes between Extubation Success and Failure Groups
Comparison of Clinical Outcomes between Extubation Success and Failure Groups

Dichotomized swallowing, tongue protrusion, spontaneous cough, suctioning cough, and motor response in Glasgow Coma Scale were included in the multiple logistic regression analysis. A scoring system with weighting according to odds ratios is shown in table 4. The AUC for the total STAGE score to predict extubation success was 0.72 (95% CI, 0.64 to 0.79; fig. 5). After internal validation using bootstrap, AUC was 0.71 (95% CI, 0.64 to 0.79). The receiver operating characteristic curve of the STAGE score after internal validation with the cutoff point, sensitivity, specificity, and positive and negative predictive values is available in Supplemental Digital Content 4: Supplemental Figure 1 (https://links.lww.com/ALN/D262). The calibration curve demonstrated high goodness of fit between predicted probability and observed proportion (Supplemental Digital Content 5: Supplemental Figure 2, https://links.lww.com/ALN/D263).

Table 4.

Results of Multiple Logistic Regression Analysis Associated with Extubation Success

Results of Multiple Logistic Regression Analysis Associated with Extubation Success
Results of Multiple Logistic Regression Analysis Associated with Extubation Success
Fig. 5.

The receiver operating characteristic curve of the Swallowing, Tongue protrusion, Airway protection reflected by spontaneous and suctioning cough, and Glasgow Coma Scale Evaluation (STAGE) score to predict extubation success rate. The area under the receiver operating characteristics curve is 0.72 (95% CI, 0.64 to 0.79).

Fig. 5.

The receiver operating characteristic curve of the Swallowing, Tongue protrusion, Airway protection reflected by spontaneous and suctioning cough, and Glasgow Coma Scale Evaluation (STAGE) score to predict extubation success rate. The area under the receiver operating characteristics curve is 0.72 (95% CI, 0.64 to 0.79).

Close modal

Determined by the highest Youden index, the cutoff point of the STAGE score was 6, with 59% (95% CI, 51 to 66%) sensitivity, 74% (95% CI, 59 to 86%) specificity, 90% (95% CI, 84 to 95%) positive predictive value, and 30% (95% CI, 21 to 39%) negative predictive value. When the STAGE score was greater than or equal to 9, the model had a 100% (95% CI, 90 to 100%) specificity and 100% (95% CI, 72 to 100%) positive predictive value for predicting extubation success. Details of different thresholds of the STAGE score are shown in Supplemental Digital Content 6: Supplemental Table 1 (https://links.lww.com/ALN/D264). The extubation success rates were increased from 67 and 72% to 84 and 100%, respectively, in the STAGE score groups of 2 to 3, 4 to 5, 6 to 8, and 9 to 12, respectively (fig. 6).

Fig. 6.

Extubation success rates in different Swallowing, Tongue protrusion, Airway protection reflected by spontaneous and suctioning cough, and Glasgow Coma Scale Evaluation (STAGE) score groups.

Fig. 6.

Extubation success rates in different Swallowing, Tongue protrusion, Airway protection reflected by spontaneous and suctioning cough, and Glasgow Coma Scale Evaluation (STAGE) score groups.

Close modal

In the sensitivity analyses, all AUC values in the reconstructed models with the sequential removal of one, two, three, and four items were lower than the original model including all five items of the STAGE score (Supplemental Digital Content 7: Supplemental Table 2, https://links.lww.com/ALN/D265). In addition, the STAGE score showed better predictive value in subgroups of high motor response in Glasgow Coma Scale (AUC = 0.76 in patients with motor response in Glasgow Coma Scale of 5 or higher; P = 0.020). No significant differences were found between the timely and delayed extubation groups (fig. 7, A and B). Of all the patients included, the STAGE score showed superior predictive value compared with Glasgow Coma Scale alone (AUC = 0.72 for the STAGE score and 0.58 for Glasgow Coma Scale; P = 0.024). The AUC of the STAGE score for patients with Glasgow Coma Scale scores of less than 5 was 0.56 (95% CI, 0.23 to 0.89).

Fig. 7.

Comparison of areas under the receiver operating characteristics curve (AUCs) for the Swallowing, Tongue protrusion, Airway protection reflected by spontaneous and suctioning cough, and Glasgow Coma Scale Evaluation (STAGE) score in different subgroups. (A) Comparison of AUCs for the STAGE score between the timely and delayed extubation groups. P = 0.720. (B) Comparison of AUCs for the STAGE score between patients with motor responses in the Glasgow Coma Scale less than 5 and those with motor responses in the Glasgow Coma Scale 5 or greater. P = 0.020.

Fig. 7.

Comparison of areas under the receiver operating characteristics curve (AUCs) for the Swallowing, Tongue protrusion, Airway protection reflected by spontaneous and suctioning cough, and Glasgow Coma Scale Evaluation (STAGE) score in different subgroups. (A) Comparison of AUCs for the STAGE score between the timely and delayed extubation groups. P = 0.720. (B) Comparison of AUCs for the STAGE score between patients with motor responses in the Glasgow Coma Scale less than 5 and those with motor responses in the Glasgow Coma Scale 5 or greater. P = 0.020.

Close modal

In this diagnostic study, neurosurgical patients undergoing the weaning process and extubation attempts were investigated. The major findings included the following: (1) decisions of delayed extubation were made in one-fourth of patients mainly by the ICU physicians; (2) the STAGE scoring system combining the airway-protecting function and consciousness assessment results was developed to predict the success rate of extubation; and (3) when the cutoff point of the STAGE score was set at 6, it could predict extubation success and exclude extubation failure with acceptable overall value. However, a STAGE score higher than 9 might have a better probability of predicting extubation success.

Limited data on weaning and extubation of neurosurgical patients lead to high rates of decisions to delay. Here, we defined delayed extubation with a time window of 24 h after a successful spontaneous breathing trial for the reason that extubation assessment and decision-making were performed every 24 h in our routine practice, as well as based on the studies of McCredie et al.15  and Taran et al.25  The rate of delayed extubation in our cohort (25%) was comparable in other brain-injured patients (27 to 30%) but higher than that of patients admitted to the general ICUs.25 

Although the rate of delayed extubation was higher in this cohort, the reasons for nonextubating decisions by the physicians were poorly reported. Our questionnaire survey revealed that consideration of the airway-protecting function and consciousness were the major determinants in extubation decision-making. Compared with general ICU patients, due to severe brain injury or some special lesions in the brainstem, neurosurgical patients present with much more complicated conditions of consciousness and airway clearance capacity. Even with a successful spontaneous breathing trial, there are possibilities of extubation failure that were consistent with the recommendations.10  Interestingly, we found that delayed extubation events were mostly decided by the ICU physicians (82%) but not the neurosurgeons (fig. 3). Neurosurgeons tend to prioritize neurologic function related to the surgical aspect, whereas ICU physicians might also consider other factors such as the high possibility of airway obstruction, potential decreased respiratory function or consciousness in respiratory or intracranial infection, and the workloads during holidays, as shown in figure 3.

Eleven patients with successful spontaneous breathing trials underwent direct tracheostomy for different reasons (fig. 1). Some were due to persistently low consciousness or weak airway function, while some patients and families chose direct tracheostomy instead of waiting for extubation to shorten the length of stay in the ICU or hospital, expediting the transfer of patients to local primary institutions, where patients could receive care from their family members and benefit from the lower cost of living in another city. We are not sure of their actual extubation outcomes if they were extubated and excluded them from our study.15,19  The extubation failure rate (19%) of the final enrollments was in the range of 10 to 38%, as reported by previous studies.3,14,16,19,28  However, whether the delay would affect extubation failure remains to be discussed. Coplin et al.17  and McCredie et al.15  found that extubation failure rates did not vary between the timely and delayed extubation groups, which was consistent with our findings. Since prolonging the duration of mechanical ventilation alone, even after meeting the extubation criteria, might not benefit neurosurgical patients, identifying an objective evaluation strategy is of great importance.

Due to the scarcity of evidence and the variation in extubation assessment practices for neurosurgical patients, extubation decisions have often been subjectively determined by physicians. In our study, we sought to address this issue by employing the nominal group technique to minimize deviations in item selection. This approach allowed us to incorporate the perspectives of a broader community in a democratic and transparent manner.24,29  Moreover, the statistical significance of all five selected items between the extubation success and failure groups, as demonstrated by chi-square tests, enhances the credibility of our selection process. We proceeded to assign scores to each item based on the results of multiple logistic regression analysis, resulting in the formation of our final model. The STAGE score exhibited favorable values for both the AUC and the calibration curve, indicating acceptable prediction and calibration capabilities.

According to the highest Youden index, the cutoff point of the STAGE score was set at 6 to predict rates of extubation success and failure. For patients with STAGE scores greater than or equal to 6 and already meeting other extubation readiness criteria, clinicians could extubate in a timely manner to reduce the chance of any unnecessary delay. Notably, a STAGE score greater than or equal to 9 revealed its promising potential in predicting extubation success (100% specificity and positive predictive value), indicating immediate readiness for extubation. However, an increase in the wait time for the score to 9 might be associated with more delayed extubation. Therefore, the extubation decision should be carefully weighed against the advantage of a higher extubation success rate and the disadvantage of prolonged mechanical ventilation time.

To date, several evaluation batteries have been investigated to successfully predict extubation readiness. In the Extubation strategies in Neuro-Intensive care unit patients and associations with Outcomes (ENIO) study,19  the AUCs for predicting extubation success were both 0.79 in two models including 20 and 7 variables, respectively. It was the largest multicenter study with external validation, and the results were highly reliable for extrapolation.19  However, this model was somewhat difficult to perform at the bedside. In addition, some simplified models have been developed. The model of Godet et al.13  including cough, deglutition, gag reflex, and Coma Recovery Scale–Revised visual had an AUC of 0.82. They considered cough as one parameter either spontaneously and/or during suctioning.13  In our clinical practice, we have observed that extubation success rates differed between the patient with either one and/or both. Since these two reflexes had different brain circuits and motor commands in the central nervous system,30  the two items were separated in our study. The Visual pursuit, Swallowing, Age, Glasgow for Extubation (VISAGE) score by Asehnoune et al.18 , had an AUC of 0.75, consisted of age, visual pursuit, swallowing, and the Glasgow Coma Scale. Different from these models, considering that some parameters with clinical significance might be missing without reaching “statistical significance,” the five items of the STAGE score were selected based on experts’ opinions rather than the solo statistical results.

Whether the Glasgow Coma Scale alone could affect the extubation outcome remains controversial. Although several studies suggest that a low Glasgow Coma Scale may be an independent risk factor for extubation failure,15,18,28,31,32  some studies report different opinions. Coplin et al.17  report that 80% (39 of 49) of enrolled subjects with Glasgow Coma Scale scores of 8 or lower and 91% (10 of 11) with Glasgow Coma Scale scores of 4 or lower underwent extubation successfully. The study of Manno et al.33  demonstrates that for comatose patients with adequate airway function and successful spontaneous breathing trial, it might be safe to extubate early, which is in agreement with our results. In this study, 38 comatose patients had STAGE scores greater than or equal to 6. Although lacking spontaneous cough and tongue protrusion, these patients had strong swallowing and suctioning cough to clear secretions with an acceptable rate (33 of 38, 87%) of extubation success.

Types of brain injury might affect weaning and extubation outcomes due to impaired airway protection or consciousness level. In our study, heterogeneity of extubation failure rates could be seen among patients with head trauma (29%), cerebral tumor (20%), and subarachnoid or intracerebral hemorrhage (15%). Further research is required to clarify the relationship between different types of brain injury and extubation outcomes.

Our study also suffers from several limitations. First, this is a single-center study with internal validation; further studies with external validation are needed in the future. Second, the items of the STAGE score were obtained after the nominal group technique. Five items all met the statistical differences after group comparisons enhancing the credibility of our selection process. Our model demonstrated acceptable prediction and calibration values. While acknowledging the potential for selection bias in our study, we believe these findings might offer a new clue for developing models using the nominal group technique. We are looking forward to future research that evaluates the feasibility of this method in diagnostic predictive studies. Third, delayed extubation was a subjective decision. Different clinicians might determine timely or delayed extubation for different reasons based on their personal clinical experiences. Since rates of delayed extubation and extubation failure were consistent with previous studies, we assumed that our results might be convincing. Fourth, sample size estimation was not conducted during the study design, and patients were enrolled consecutively for 1 year. After finishing our study, we read the methods proposed by Riley et al.34  and calculated the sample size, which is shown in Supplemental Digital Content 8: Supplemental Text 4 (https://links.lww.com/ALN/D266). Fifth, in our daily practice, we perform cuff leak tests only for patients at higher risk of upper airway obstruction according to the guideline.10  Therefore, we did not include these results in our study.

Conclusions

The rates of delayed extubation and extubation failure are high in critical neurosurgical patients. According to the most common reasons for extubation delay, we developed the STAGE scoring system by combining the assessment points of airway-protecting function and level of consciousness. Therefore, the STAGE score could guide further in deciding on the patient’s extubation readiness and may help clinicians minimize extubation failure-associated long-term adverse consequences.

Research Support

Support was provided solely from institutional and/or departmental sources.

Competing Interests

The authors declare no competing interests.

Supplemental Digital Content 1: Supplemental Text 1. Details of data collection, https://links.lww.com/ALN/D259

Supplemental Digital Content 2: Supplemental Text 2. Questionnaire for extubation decisions survey, https://links.lww.com/ALN/D260

Supplemental Digital Content 3: Supplemental Text 3. The nominal group methods and consensus results, https://links.lww.com/ALN/D261

Supplemental Digital Content 4: Supplemental Figure 1. AUC of the STAGE score after internal validation, https://links.lww.com/ALN/D262

Supplemental Digital Content 5: Supplemental Figure 2. The calibration curve of the model, https://links.lww.com/ALN/D263

Supplemental Digital Content 6: Supplemental Table 1. Details of different thresholds of the STAGE score, https://links.lww.com/ALN/D264

Supplemental Digital Content 7: Supplemental Table 2. Reconstructed AUCs after factor removal, https://links.lww.com/ALN/D265

Supplemental Digital Content 8: Supplemental Text 4. Sample size calculation, https://links.lww.com/ALN/D266

1.
Pelosi
P
,
Ferguson
ND
,
Frutos-Vivar
F
,
Anzueto
A
,
Putensen
C
,
Raymondos
K
,
Apezteguia
C
,
Desmery
P
,
Hurtado
J
,
Abroug
F
,
Elizalde
J
,
Tomicic
V
,
Cakar
N
,
Gonzalez
M
,
Arabi
Y
,
Moreno
R
,
Esteban
A
;
Ventila Study Group
:
Management and outcome of mechanically ventilated neurologic patients.
Crit Care Med
2011
;
39
:
1482
92
2.
Karanjia
N
,
Nordquist
D
,
Stevens
R
,
Nyquist
P
:
A clinical description of extubation failure in patients with primary brain injury.
Neurocrit Care
2011
;
15
:
4
12
3.
Ko
R
,
Ramos
L
,
Chalela
JA
:
Conventional weaning parameters do not predict extubation failure in neurocritical care patients.
Neurocrit Care
2009
;
10
:
269
73
4.
MacIntyre
N
:
Discontinuing mechanical ventilatory support.
Chest
2007
;
132
:
1049
56
5.
Thille
AW
,
Harrois
A
,
Schortgen
F
,
Brun-Buisson
C
,
Brochard
L
:
Outcomes of extubation failure in medical intensive care unit patients.
Crit Care Med
2011
;
39
:
2612
8
6.
Frutos-Vivar
F
,
Esteban
A
,
Apezteguia
C
,
González
M
,
Arabi
Y
,
Restrepo
MI
,
Gordo
F
,
Santos
C
,
Alhashemi
JA
,
Pérez
F
,
Peñuelas
O
,
Anzueto
A
:
Outcome of reintubated patients after scheduled extubation.
J Crit Care
2011
;
26
:
502
9
7.
Rothaar
RC
,
Epstein
SK
:
Extubation failure: Magnitude of the problem, impact on outcomes, and prevention.
Curr Opin Crit Care
2003
;
9
:
59
66
8.
Boles
JM
,
Bion
J
,
Connors
A
,
Herridge
M
,
Marsh
B
,
Melot
C
,
Pearl
R
,
Silverman
H
,
Stanchina
M
,
Vieillard-Baron
A
,
Welte
T
:
Weaning from mechanical ventilation.
Eur Respir J
2007
;
29
:
1033
56
9.
MacIntyre
NR
,
Cook
DJ
,
Ely
EW
, Jr
,
Epstein
SK
,
Fink
JB
,
Heffner
JE
,
Hess
D
,
Hubmayer
RD
,
Scheinhorn
DJ
;
American College of Chest Physicians; American Association for Respiratory Care; American College of Critical Care Medicine
:
Evidence-based guidelines for weaning and discontinuing ventilatory support: A collective task force facilitated by the American College of Chest Physicians; the American Association for Respiratory Care; and the American College of Critical Care Medicine.
Chest
2001
;
120
:
375S
95S
10.
Robba
C
,
Poole
D
,
McNett
M
,
Asehnoune
K
,
Bösel
J
,
Bruder
N
,
Chieregato
A
,
Cinotti
R
,
Duranteau
J
,
Einav
S
,
Ercole
A
,
Ferguson
N
,
Guerin
C
,
Siempos
II
,
Kurtz
P
,
Juffermans
NP
,
Mancebo
J
,
Mascia
L
,
McCredie
V
,
Nin
N
,
Oddo
M
,
Pelosi
P
,
Rabinstein
AA
,
Neto
AS
,
Seder
DB
,
Skrifvars
MB
,
Suarez
JI
,
Taccone
FS
,
van der Jagt
M
,
Citerio
G
,
Stevens
RD
:
Mechanical ventilation in patients with acute brain injury: Recommendations of the European Society of Intensive Care Medicine consensus.
Intensive Care Med
2020
;
46
:
2397
410
11.
Bosel
J
:
Who is safe to extubate in the neuroscience intensive care unit?
Semin Respir Crit Care Med
2017
;
38
:
830
9
12.
Kutchak
FM
,
Rieder
MM
,
Victorino
JA
,
Meneguzzi
C
,
Poersch
K
,
Forgiarini
LA
, Jr
,
Bianchin
MM
:
Simple motor tasks independently predict extubation failure in critically ill neurological patients.
J Bras Pneumol
2017
;
43
:
183
9
13.
Godet
T
,
Chabanne
R
,
Marin
J
,
Kauffmann
S
,
Futier
E
,
Pereira
B
,
Constantin
JM
:
Extubation failure in brain-injured patients: Risk factors and development of a prediction score in a preliminary prospective cohort study.
Anesthesiology
2017
;
126
:
104
14
14.
Vidotto
MC
,
Sogame
LC
,
Calciolari
CC
,
Nascimento
OA
,
Jardim
JR
:
The prediction of extubation success of postoperative neurosurgical patients using frequency-tidal volume ratios.
Neurocrit Care
2008
;
9
:
83
9
15.
McCredie
VA
,
Ferguson
ND
,
Pinto
RL
,
Adhikari
NK
,
Fowler
RA
,
Chapman
MG
,
Burrell
A
,
Baker
AJ
,
Cook
DJ
,
Meade
MO
,
Scales
DC
;
Canadian Critical Care Trials Group
:
Airway management strategies for brain-injured patients meeting standard criteria to consider extubation: A prospective cohort study.
Ann Am Thorac Soc
2017
;
14
:
85
93
16.
Anderson
CD
,
Bartscher
JF
,
Scripko
PD
,
Biffi
A
,
Chase
D
,
Guanci
M
,
Greer
DM
:
Neurologic examination and extubation outcome in the neurocritical care unit.
Neurocrit Care
2011
;
15
:
490
7
17.
Coplin
WM
,
Pierson
DJ
,
Cooley
KD
,
Newell
DW
,
Rubenfeld
GD
:
Implications of extubation delay in brain-injured patients meeting standard weaning criteria.
Am J Respir Crit Care Med
2000
;
161
:
1530
6
18.
Asehnoune
K
,
Seguin
P
,
Lasocki
S
,
Roquilly
A
,
Delater
A
,
Gros
A
,
Denou
F
,
Mahé
PJ
,
Nesseler
N
,
Demeure-Dit-Latte
D
,
Launey
Y
,
Lakhal
K
,
Rozec
B
,
Mallédant
Y
,
Sébille
V
,
Jaber
S
,
Le Thuaut
A
,
Feuillet
F
,
Cinotti
R
;
ATLANREA group
:
Extubation success prediction in a multicentric cohort of patients with severe brain injury.
Anesthesiology
2017
;
127
:
338
46
19.
Cinotti
R
,
Mijangos
JC
,
Pelosi
P
,
Haenggi
M
,
Gurjar
M
,
Schultz
MJ
,
Kaye
C
,
Godoy
DA
,
Alvarez
P
,
Ioakeimidou
A
,
Ueno
Y
,
Badenes
R
,
Suei Elbuzidi
AA
,
Piagnerelli
M
,
Elhadi
M
,
Reza
ST
,
Azab
MA
,
McCredie
V
,
Stevens
RD
,
Digitale
JC
,
Fong
N
,
Asehnoune
K
;
ENIO Study Group, PROtective VENTilation network, European Society of Intensive Care Medicine, Colegio Mexicano de Medicina Critica, Atlanréa Group, and Société Française d’Anesthésie-Réanimation–SFAR Research Network
:
Extubation in neurocritical care patients: The ENIO international prospective study.
Intensive Care Med
2022
;
48
:
1539
50
20.
Collins
GS
,
Reitsma
JB
,
Altman
DG
,
Moons
KG
:
Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): The TRIPOD statement.
BMJ
2015
;
350
:
g7594
21.
Guidelines for prevention of nosocomial pneumonia
:
Centers for Disease Control and Prevention.
MMWR Recomm Rep
1997
;
46
:
1
79
22.
American Thoracic Society; Infectious Diseases Society of America
:
Guidelines for the management of adults with hospital-acquired, ventilator-associated, and healthcare-associated pneumonia.
Am J Respir Crit Care Med
2005
;
171
:
388
416
23.
Papazian
L
,
Klompas
M
,
Luyt
CE
:
Ventilator-associated pneumonia in adults: A narrative review.
Intensive Care Med
2020
;
46
:
888
906
24.
Jones
J
,
Hunter
D
:
Consensus methods for medical and health services research.
BMJ
1995
;
311
:
376
80
25.
Taran
S
,
Angriman
F
,
Pinto
R
,
Ferreyro
BL
,
Amaral
ACK
;
Toronto Intensive Care Observational Registry
:
Discordances between factors associated with withholding extubation and extubation failure after a successful spontaneous breathing trial.
Crit Care Med
2021
;
49
:
2080
9
26.
Van Calster
B
,
McLernon
DJ
,
van Smeden
M
,
Wynants
L
,
Steyerberg
EW
;
Topic Group ‘Evaluating Diagnostic Tests and Prediction Models’ of the STRATOS Initiative
:
Calibration: The Achilles heel of predictive analytics.
BMC Med
2019
;
17
:
230
27.
Steyerberg
EW
,
Harrell
FE
,
Borsboom
GJ
,
Eijkemans
MJ
,
Vergouwe
Y
,
Habbema
JD
:
Internal validation of predictive models: Efficiency of some procedures for logistic regression analysis.
J Clin Epidemiol
2001
;
54
:
774
81
28.
Namen
AM
,
Ely
EW
,
Tatter
SB
,
Case
LD
,
Lucia
MA
,
Smith
A
,
Landry
S
,
Wilson
JA
,
Glazier
SS
,
Branch
CL
,
Kelly
DL
,
Bowton
DL
,
Haponik
EF
:
Predictors of successful extubation in neurosurgical patients.
Am J Respir Crit Care Med
2001
;
163
:
658
64
29.
Vella
K
,
Goldfrad
C
,
Rowan
K
,
Bion
J
,
Black
N
:
Use of consensus development to establish national research priorities in critical care.
BMJ
2000
;
320
:
976
80
30.
Ando
A
,
Farrell
MJ
,
Mazzone
SB
:
Cough-related neural processing in the brain: A roadmap for cough dysfunction?
Neurosci Biobehav Rev
2014
;
47
:
457
68
31.
Qureshi
AI
,
Suarez
JI
,
Parekh
PD
,
Bhardwaj
A
:
Prediction and timing of tracheostomy in patients with infratentorial lesions requiring mechanical ventilatory support.
Crit Care Med
2000
;
28
:
1383
7
32.
Wang
S
,
Zhang
L
,
Huang
K
,
Lin
Z
,
Qiao
W
,
Pan
S
:
Predictors of extubation failure in neurocritical patients identified by a systematic review and meta-analysis.
PLoS One
2014
;
9
:
e112198
33.
Manno
EM
,
Rabinstein
AA
,
Wijdicks
EF
,
Brown
AW
,
Freeman
WD
,
Lee
VH
,
Weigand
SD
,
Keegan
MT
,
Brown
DR
,
Whalen
FX
,
Roy
TK
,
Hubmayr
RD
:
A prospective trial of elective extubation in brain injured patients meeting extubation criteria for ventilatory support: A feasibility study.
Crit Care
2008
;
12
:
R138
34.
Riley
RD
,
Ensor
J
,
Snell
KIE
,
Harrell
FE
, Jr
,
Martin
GP
,
Reitsma
JB
,
Moons
KGM
,
Collins
G
,
van Smeden
M
:
Calculating the sample size required for developing a clinical prediction model.
BMJ
2020
;
368
:
m441
This is an open-access article distributed under the terms of the Creative Commons Attribution-Non Commercial-No Derivatives License 4.0 (CCBY-NC-ND), where it is permissible to download and share the work provided it is properly cited. The work cannot be changed in any way or used commercially without permission from the journal.