Introduction

In medical surgical intensive care unit (ICU) populations, extubation failure is defined as reintubation within 72 h. The incidence and negative consequences of extubation failure include increased duration of mechanical ventilation, increased ICU length of stay (LOS), increased nosocomial pneumonia, and increased mortality [1–7]. Variables associated with extubation failure in these populations have also been identified, including older age [8], use of continuous sedation [9], severity of illness on ICU admission, prolonged ventilation before extubation [10], hypercapnea [11], positive fluid balance [12], and neurologic impairment [11, 13]. A review of the literature makes it clear that these variables differ depending on the population studied [14]. Studies of extubation failure in neurocritical care patient populations are limited at best and tend to emphasize neurosurgical patients. As the number of neurocritical care units (NCCUs) that care for both neurological and neurosurgical patients grows, it becomes increasingly important to study the NCCU population to address how primary neurological injury affects ventilator management. This issue is also important in the general medical and surgical ICU, where up to 20% of patients are intubated for reasons involving neurological dysfunction [15].

Unlike the medical-surgical ICU, the NCCU specializes in caring for patients with neurological injuries, including stroke, subarachnoid hemorrhage, central nervous system infection, brain trauma, intracerebral tumors, spinal injuries, and peripheral nerve injuries. Many of these patients are subsequently intubated owing to concerns that airway compromise might arise as a result of decreasing levels of consciousness. In such patients, extubation is delayed because of clinical concerns regarding their impaired mental status and the risk of aspiration, pneumonia, and acute respiratory failure. These concerns exist even though their pulmonary system is relatively unaffected and their work of breathing is normal. Approaches to the extubation of these neurologically impaired patients are largely based on expert opinion [16–19]. Large studies emphasizing prospective observation cohorts or clinical trials are lacking. Existing studies suggest that the risk of extubation failure is much lower than that predicted by expert opinion [13]. Coplin et al. [20] found that NCCU patients with a Glasgow Coma Scale (GCS) <4, without high airway care requirements or pneumonia, and with a cough or gag reflex, may be safely extubated. In the medical-surgical ICU, such parameters are the basis of standard daily spontaneous breathing trials that facilitate rapid liberation from the ventilator [21–23]. At present, well-designed prospective studies in the general medical and surgical ICU support the idea that early extubation and spontaneous breathing trials reduce the rates of ventilator-associated pneumonia (VAP), reduce hospital LOS, and improve mortality [23]. Coplin and colleagues [20, 24] have also completed a study that suggests that standard daily spontaneous breathing trials, when adapted for NCCU patients, may facilitate rapid liberation from ventilation. Such strategies appear to reduce rates of pneumonia, ICU and hospital LOS, and mean hospital charges [20, 24].

Because relatively few studies have looked at patients with a primary neurological injury key, clinical questions remain. What is the extubation failure rate in patients with a primary neurological injury who do not have respiratory failure? What is the clinical path leading to reintubation in these patients? We completed a retrospective review of all the admissions to our NCCU over 5 years to estimate the rate of extubation failure and identify the clinical variables associated with reintubation for use as a guide for future studies.

Methods

Study Population

We performed a retrospective study of all patients admitted to the Johns Hopkins Hospital NCCU between January 1, 2002 and March 1, 2007. The Johns Hopkins Hospital NCCU is a 22-bed closed ICU for patients with primary neurological and neurosurgical diagnoses who require critical and intermediate levels of care. Diagnoses include stroke, subarachnoid hemorrhage, intracranial hemorrhage, traumatic brain injury, tumor, central nervous system infection, neurosurgical procedures, neurointerventional procedures, spinal cord injuries, peripheral nerve disease, and neuromuscular disease. All patients were intubated by a fellow, resident, or a qualified physician. All fellows or residents intubated under the supervison of a qualified attending physician. All patients were evaluated prior to extubation by a respiratory therapist, who performed pulmonary mechanics testing for negative inspiratory flow, functional vital capacity, resting spontaneous breathing index, cuff-leak test, and presence of cough and gag reflexes. The respiratory therapist also performed a standard spontaneous breathing trial, including continuous positive airway pressure (CPAP), pressure support, or T-piece trial for at least 1 h prior to extubation with an FiO2 of 40%. This was completed under the supervision of a specialist neuro-intensivist physician. The physician team made the final decision to extubate. The decision criteria differed among attending physicians but always included the following: the ability to demonstrate adequate oxygenation of SaO2 > 92% during a spontaneous trial of breathing, sufficient neurological recovery to allow for control of the patient’s respiratory drive, and the patient’s ability to follow commands.

Electronic Databases

Our NCCU uses several patient information management systems, including Eclypsis, Casemix, Electronic Patient Record, Provider Order Entry System, and an independent respiratory therapy database. Eclypsis is a clinically oriented database that tracks laboratory results as well as nurse and physician notes. Casemix is a financial database used to track patient demographics as well as coding and other financially relevant information. Electronic Patient Record is a clinically oriented database that includes laboratory results, physician notes, and clinical documentation of test results. Provider Order Entry System is an electronic order entry system. For this study, data gathered electronically were supplemented by a review of written charts.

The Eclypsis and Casemix databases were used initially to identify patients with ICD-9 codes of interest and location of interest while hospitalized. They were also used to ascertain census, admission dates, extubation dates, reintubation dates, and other clinical data, such as ICD-9 codes and procedures performed. The Casemix database was used to check and verify information obtained from Eclypsis and vice versa. A systematic chart review was completed for all patients who where reintubated to verify critical information and obtain the clinical impression of the providers who cared for the patient. Databases that operate independently of these were also used. These included databases maintained by our respiratory therapist group and our ICU administrators.

Data Collection

Eclypsis and Casemix were queried with a range of ICD-9 codes relevant to brain injury (codes available on request) to identify all admissions to the NCCU, their primary and secondary diagnoses, and whether the patient was intubated during their NCCU stay. We selected those patients who had a primary neurological diagnosis and were intubated non-electively or had remained intubated for >48 h (to exclude routine postoperative patients). The following clinical characteristics were obtained: age, gender, extubation success or failure, reintubation, percutaneous tracheostomy, time from extubation to reintubation, and reason for initial intubation. These data were initially coded in the Eclypsis database and were verified against other databases such as the respiratory therapy database and written clinical records. We then reviewed the written charts of the patients with failed extubation and collected the following immediate pre-intubation variables: clinical reason for reintubation as described by the treating physician (which we verified by reviewing the clinical history), GCS, arterial blood gas (ABG), respiratory rate, minute ventilation, pulse oximetry, presence or absence of atelectasis, pulmonary edema, focal infiltrate on chest radiograph, presence of pneumonia, and presence of systemic infection.

Definitions

Reintubation was defined as reinstitution of mechanical ventilation any time during the patient’s hospital stay. Extubation failure was defined as reinstitution of mechanical ventilation within 72 h of extubation for reasons other than a planned procedure. Successful extubation was defined as remaining free of mechanical ventilation until discharge. Nosocomial pneumonia was defined according to the National Nosocomial Infections Surveillance System (NNIS) definition [25]. Systemic infection was defined as presence of fever, white blood cell count >12,000, and positive blood, urine, or sputum cultures without other identified cause of fever and elevated white blood cell count. Respiratory distress was defined as any clinical signs of ventilatory difficulty that resulted in desaturation or increasing FiO2 requirement and was suggestive to the physician team of impending respiratory failure. Different classifications of respiratory distress included in the study are listed in Table 1. Neurological respiratory insufficiency was defined as any evidence of respiratory distress not attributed to a primary pulmonary disorder (a primary pulmonary disorder was defined as any known chronic medical disorder that affected the work of breathing or oxygenation, any acute pulmonary process such as aspiration-induced pneumonia or prolonged ventilation, pulmonary embolus, or a primary cardiac disorder affecting ventilation or oxygenation). In retrospect these factors were determined through examination of the ICU medical database, clinician records, nursing records, and respiratory therapy records. If these primary pulmonary factors were excluded and a primary brain injury was identified and known to be the primary reason for intubation, the patient was classified as having neurological respiratory insufficiency. The general categories and definitions of events that led to reintubation are listed in Table 1.

Table 1 Definitions of clinical scenarios that lead to initial intubation or reintubation
Full size table

Statistical Analysis

All statistics were calculated with IBM SPSS Statistics or Microsoft Excel. Standard Chi-square analysis was used to compare incidence and prevalence between groups. Microsoft Excel was also used to generate descriptive statistics.

Results

During the 5-year period studied, 257,444 admissions were made to the neurosciences center, and 9,673 admissions representing 7,576 patients were made to our NCCU. Of those admitted to the NCCU, 1,901 patients (2,135 admissions) were mechanically ventilated. Of these patients, 636 were intubated for non-neurological reasons; i.e., they had a primary pulmonary disorder resulting in respiratory distress (135), or they were ventilated in the NCCU during emergence from anesthesia (501). The remaining 1,265 patients were intubated due to a primary neurological issue. These patients either suffered no pulmonary injury or had decreased ventilation resulting from spinal cord or peripheral nerve injury. The average age of these patients was 56 years (range 12–92 years), and 628 (50%) were male. The average length of NCCU stay was 7 days (range 2–92 days). The average GCS at the time of extubation was 11 ± 3 (±SD; range 3–15). Of the 1,265 patients, 713 (56%) carried a primary diagnosis involving the brain. Their injuries did not involve the peripheral nervous system or spinal cord. The most common neurological diagnoses causing primary brain injury were: trauma (153 patients), subarachnoid hemorrhage (126 patients), ischemic stroke (70 patients), and brain neoplasm (65 patients). A complete list can be seen in Table 2.

Table 2 Primary diagnosis of neurological patients intubated in the NCCU
Full size table

Extubation and Reintubation

Of the 1,265 patients intubated for a neurological reason, 844 (67%) were successfully extubated, 292 (23%) were extubated to comfort care, and 129 (10.2%) were reintubated at some point during their hospital stay. Of those 129 patients who were reintubated, 77 were reintubated within 72 h of extubation, making the overall extubation failure rate 6.1%. Of the 129 who were reintubated, 99 were reintubated for issues related to their primary brain injury, independent of spinal cord or peripheral nerve injury. Additionally, 181 (14.3%) patients had a tracheostomy during their hospital stay (77 had a tracheostomy, were successfully liberated from the ventilator, and were part of the 844 who were successfully liberated; 104 had a tracheostomy and were ventilator-dependent at the end of their hospital stay and were part of the group who were reintubated.) There were 11 patients who self-extubated among the (n = 99) brain injury only group and 13 in those reintubated for any neurologically related disease (n = 129). Of the 129 patient with any Neurological injury, 15 died after reintubation, 104 had a tracheostomy, and 10 survived and were extubated without a tracheostomy (Fig. 1).

Fig. 1
figure 1

Extubated patients and their extubation outcomes (n = 1265)

Full size image

Reintubation and Primary Brain Injury

Of the 129 patients who were reintubated for neurologically related injury, 99 (77%) had isolated brain injury only. The others had spinal cord injury [23] or peripheral nerve injury or neuromuscular disorders [7]. The most common admitting diagnoses of those reintubated for primary brain injury were: subarachnoid hemorrhage, intracranial hemorrhage, ischemic stroke, and brain neoplasm (see Table 3). Of those patients with primary brain injury, 75 were reintubated within 72 h of extubation. The extubation failure rate of this group as a percentage of the total intubated population was 6.1. Within this group we had six cases of stridor. One occurred more than 24 h after extubation. The others occurred with the first 12 h after extubation (see Table 4).

Table 3 The primary diagnosis of patients with acute brain injury only who required reintubation in the NCCU
Full size table
Table 4 Number of patients reintubated at different times secondary to a primary brain injury
Full size table

The primary cause for reintubation of patients with primary brain injury was respiratory distress associated with a change in mental status without signs of aspiration or pneumonia [62 patients (63%)]. These patients had an intact gag and cough but decreased respiratory rate and clinical as well as radiographic signs of atelectasis. The primary decision to intubate these patients was based on the clinical assessment of the attending physician; 25 of these patients demonstrated life-threatening oxygen desaturation prior to reintubation. All patients in this category were reported as having signs of respiratory distress described as air hunger, increased respiratory rate, the use of accessory muscles, and gasping. None of these patients had aspirated, had pneumonia as a primary diagnosis, or had been diagnosed with pneumonia according to the NNIS criteria [26]. Other causes for reintubation of patients with primary brain injury were pneumonia, aspiration, and general medical emergencies related or not related to the progression of the primary neurological injury (Table 5).

Table 5 Reasons for reintubation in patients with a primary brain injury
Full size table

Aspiration and Pneumonia

Aspiration was identified by the providers in six reintubated patients (6%), and pneumonia was diagnosed in 10 reintubated patients (10%). Retrospective review using the NNIS criteria identified two additional patients who met the criteria for nosocomial pneumonia. The total pneumonia rate combining aspiration and nosocomial pneumonia was 18% (n = 18).

Other Medical Conditions

A small number of patients with altered mental status and other acute medical conditions also were reintubated: one patient (1%) with hematemesis, two patients (2%) with pulmonary edema secondary to congestive heart failure, one patient (1%) with cardiac arrest, and six patients (6%) with stridor.

Discussion

Extubation failure is a problem faced by all critical care physicians. Reintubation leads to increased LOS and is associated with increased morbidity and mortality. There are a number of risks associated with prolonged intubation including an increased risk of VAP, increased mortality, and increased LOS [4, 27–29]. Providers of critical care are extubating patients from mechanical ventilation earlier than in the past, with the goal of reducing morbidity, mortality, and the costs related to VAP and ventilator management [13, 20, 24]. However, extubation failure is problematic, as reintubation is also associated with increased LOS and increased morbidity and mortality [4, 27–29]. Understanding the characteristics of intubated patients with primary brain injury that contribute to the need for reintubation will help to reduce morbidity, mortality, and costs in the NCCU [24]. This is the first retrospective study of a large case series of encephalopathic patients with primary brain injury who required reintubation. The results have allowed us to identify issues pertaining to primary brain injury that lead to reintubation and eventual respiratory failure. Our study population differs from those observed in non-neurological medical ICUs and general surgical ICUs. Few, if any, of our patients had delirium or encephalopathy caused by something other than one of four primary diagnoses: traumatic brain injury, subarachnoid hemorrhage, intracranial hemorrhage, or brain tumor. The classical ICU causes of delirium, such as sepsis or infection, were under-represented in our population [4, 6, 15, 22, 28, 29]. We believe that we have isolated a population that is unique to the NCCU setting. Patients in our study failed extubation or were later reintubated secondary to their neurological injury, not due to increased work of breathing resulting from direct injury to their lungs. This study will allow us to plan future prospective trials.

Review articles and specific texts identify the risk of aspiration pneumonia as the primary reason for reintubating or delaying extubation in populations with primary brain injury [17–19]. In our population with primary brain injury only, just 6% of reintubated patients aspirated, and only 12% had nosocomial pneumonia identified clinically prior to reintubation. Hence, 18% of patients had any kind of pneumonia before reintubation, as identified retrospectively based on the NNIS criteria or clinical description. These observations suggest that aspiration and nosocomial pneumonia while a risk is not the most common causes of reintubation in patients with primary brain injury.

In our study population, which included patients with all types of neurological injury, including spinal cord and peripheral nerve injury, the overall rate of reintubation was 7.9%. Of those with primary brain injury, 6.1% required reintubation within 72 h of extubation. This rate of extubation failure is unusually low. Given that this is a retrospective review, there is a risk of ascertainment bias. We attempted to correct for this bias using multiple databases to capture and identify all intubated patients as well as those reintubated. The databases used included our primary NCCU computerized data system, as well as independent databases maintained by our respiratory therapy service and our ICU administrators. Even if we significantly underestimated the extubation rate, it is unlikely that we would have a rate comparable to that of other surgical or medical ICUs. Although no study has directly addressed what extubation failure rate has the best cost–benefit ratio in the ICU setting, most authors have suggested that reintubation rates between 10 and 15% are justified [6]. The fact that our rate is so low suggests that a number of patients in our unit may not be receiving the benefit of early extubation [23]. In our unit, the VAP rates have been reported to be as high as 30 cases per 1000 ventilated hours. We have successfully reduced these numbers through the use of ventilator bundles that include spontaneous breathing trials and improved oral hygiene with chlorhexidine scrubs. This practice has resulted in a reduction of the VAP rates to as low as 11.7 cases per 1000 ventilator hours (unpublished data). Even though our in-hospital mortality rate was high in this population (21%), as compared to the 2.5–15% observed in other studies [6, 7], no documented deaths resulting from early extubation were recorded during the period of observation.

In our case series, the primary cause of respiratory failure that led to reintubation appeared to be neurological respiratory insufficiency. We believe that this phenomenon is the result of primary brain injury and represents a proclivity of the injured brain not to detect physiological cues of ongoing disrupted ventilator patterns such as changing CO2 sensitivity, inadequate inspiration, and decreased arterial oxygen levels. The clinical course is typified by atelectasis and decreasing minute ventilation resulting in ventilation/perfusion (V/Q) mismatch and eventual respiratory failure. In review, these patients did not have signs of infection as identified by fever, elevated white blood cell count, or identified infectious sources. In addition, many of these patients demonstrated rapid correction of their V/Q mismatch with the administration of positive end-expiratory pressure and ventilatory support. This pattern of respiratory failure that we termed “neurological respiratory insufficiency” was defined clinically as atelectasis on chest X-ray (without focal infiltrate, pulmonary embolus, or clinical pneumonia as defined by NNIS criteria) in conjunction with one of the following four criteria: respiratory rate less than 8, tidal volumes less than age-specific norms at the time of reintubation, PaCO2 greater than 48, or hypoxia (pulse oximetry less than 90% or ABG PaO2 less than 60 mmHg). This pattern was observed in 59 (60%) of the patients who were reintubated. Sixty-six percent of patients (39 of 59) who had no clearly defined precipitating events leading to intubation demonstrated neurological respiratory insufficiency. These parameters are less than satisfying in their accuracy and precision in predicting respiratory failure in our population of patients with primary brain injury. However, few cues are available to the neurointensivist to truly predict who will need reintubation. Future studies can help to build on our observations to enable more accurate prediction.

It is our belief that patients with intact cough and gag reflexes and ongoing encephalopathy related to their primary brain injury may tolerate a trial of noninvasive ventilatory support such as bilevel positive airway pressure (BiPAP) or CPAP designed to reduce atelectasis and facilitate recruitment of alveoli. We have observed that in many patients, the implementation of nighttime BiPAP is successful in limiting atelectasis with no apparent increase in the risk of pneumonia. A randomized trial is needed to assess the efficacy of this strategy.

The majority of patients in our case series were reintubated based on the clinical decisions of providers who noted that they were in respiratory distress secondary to a change in mental status. The finding that 63% of the patients reintubated had no clearly defined underlying mechanism for respiratory failure was unexpected. The clinical decision to reintubate patients in this setting may have been overly cautious. This possibility is supported by the observation that of the 63 patients who were reintubated without sign of a primary pulmonary problem, only 23% actually demonstrated signs of desaturation based on ABG or pulse oximetry monitoring. However, 42 of these 63 patients met our criteria for neurological respiratory insufficiency. Patients who were reintubated after 72 h all showed signs of hypoventilation.

Conclusion

Brain-injured patients who have a primary diagnosis of encephalopathy and have been intubated have a low rate of reintubation. One cause of reintubation in these patients is aspiration or nosocomial pneumonia. However, the primary cause for respiratory failure and reintubation is not generally related to pneumonia. It appears to be a pattern of disrupted ventilation resulting in hypoventilation typified by increasing atelectasis and decreasing minute ventilation in the setting of a primary brain injury. The primary impetus for reintubation appears to be related to the clinician’s impression of progressive respiratory distress and declining mental status as opposed to actual physiological signs of desaturation and low arterial oxygen saturation. The most common underlying physiological process leading to respiratory distress is a process of progressive atelectasis and decreasing minute ventilation that eventually causes a V/Q mismatch. This patient population may benefit from strategies that use noninvasive methods of ventilatory support, such as BiPAP and CPAP, with the goal of maintaining alveolar recruitment. These data are preliminary and of a retrospective nature. Future prospective studies are required to test these hypotheses and to confirm the observations reported here.