Introduction

Severely head-injured patients, particularly those with associated multisystem injuries, often require tracheostomy at some point during their inpatient course. Previous retrospective reports assessing the influence of tracheostomy timing in head-injured patients have provided mixed results, and prospective randomized studies are rare [1]. Traditionally, tracheostomy has been recommended in order to prevent airway complications such as tracheal stenosis and laryngeal pathology. Early tracheostomy specifically, has been advocated as a means to improve outcome, with various studies suggesting that early tracheostomy may decrease the incidence of pneumonia, reduce ICU days, and shorten overall length of stay [2–5]. Data also support early weaning of mechanical ventilation within 48 h of early tracheostomy [2, 6]. On the other hand, tracheostomy itself may result in various procedural complications “early” in a patient’s course [7–13]. Some evidence also exists to suggest that early tracheostomy does not, in fact, improve outcome [8, 14]. As a result, the timing of tracheostomy in these critically ill patients remains controversial [2–6, 15–20]. In this study, we retrospectively reviewed a database of 3,104 severely head-injured patients to determine the impact of tracheostomy timing on various hospital outcomes.

Methods and Materials

Study Population

This investigation was a retrospective, 15-year observational cohort study from January 1990 until December 2005. Data were obtained for patients admitted to participating Pennsylvania Trauma Society hospitals as collected by the Pennsylvania Trauma Society Foundation (PTSF) trauma registry. The PTSF was established by the Pennsylvania Emergency Services Act of 1985 to accredit trauma centers in Pennsylvania and is currently composed of 26 designated Level I and Level II trauma centers. In addition to its accreditation mission, the PTSF also collects data from trauma centers to determine the demographics of trauma throughout the state and to provide a measure of quality and outcome.

All the data elements of the Major Trauma Outcomes Study of the American College of Surgeons Committee on Trauma are included in the registry [21]. Patients entered into the database include all trauma deaths, all intensive care unit admissions, all admissions with a length of stay greater than 72 h, all trauma patients transferred in or out of a participating institution and all patients dead on arrival. As part of their charter, trauma centers are mandated to report the requested data elements to the PTSF and their continued accreditation depends on compliance. Database fidelity is monitored by annual external audits of each participating site by the PTSF. In addition, most programs also perform routine internal audits. As of December 31, 2007, the sum total of all trauma patients submitted to the PTSF database since its inception in 1985 was 460,471.

For the above study period, we obtained the data records on all patients with a Glasgow Coma Score (GCS) of less than or equal to eight and evidence of a head injury. We then selected all patients that had undergone a tracheostomy at some point during their hospitalization. To make the study group more homogenous and because there were very few patients with isolated head injury in this cohort, we then chose to focus our analysis on those patients that had an associated injury to at least one other body system. This group was then divided into an Early Tracheostomy Group (ETG), defined as having a tracheostomy performed during hospital days one through seven, and a Late Tracheostomy Group (LTG), defined as having a tracheostomy performed greater than 7 days after admission.

Baseline demographics and presenting characteristics, including injury type and pre-existing conditions were recorded. Clinical status at presentation in the form of GCS and Injury Severity Score (ISS) was also documented. ISS is an anatomical scoring system that provides an overall score for patients with multiple injuries. Each injury is assigned an Abbreviated Injury Scale (AIS) score and is allocated to one of six body regions (Head, Face, Chest, Abdomen, Extremities (including Pelvis), and External). Only the highest AIS score in each body region is used. The three most severely injured body regions have their score squared and added together to produce the ISS score.

Outcome Measures

The following clinical variables collected by the PTSF, discharge status, number of ICU days, number of hospital days, functional status, and types of adverse events were recorded as outcome measures.

Functional status as acquired by the PTSF is based on a modification of the Functional Independence Measure (FIM) Score. The PTSF modification assesses five categories: feeding, locomotion, expression, transfer mobility, and social interaction. Each category is scored on a four-point scale, with a score of 4 meaning complete independence, 3 meaning independence with a device, 2 meaning modified dependence, and 1 meaning complete dependence. For the purposes of our analysis, the score was dichotomized, with a total score of 5 to 10 considered “dependent” and a score of greater than 10 considered “independent.”

We then analyzed the impact of tracheostomy timing (early vs. late) on discharge status (alive vs. dead), functional status at discharge (independent vs. dependent), number of ICU days, and total number of hospital days. In addition, we assessed the impact of tracheostomy timing on four classes of adverse events or “occurrences” as defined by the PTSF: pulmonary, cardiovascular, infectious, and neurological. Adverse pulmonary occurrences recorded in the PTSF database include adult respiratory distress syndrome (ARDS), acute respiratory failure, aspiration/aspiration pneumonia, atelectasis, pleural effusion, pneumonia, pneumothorax, and pulmonary embolus. Cardiovascular occurrences were listed as cardiopulmonary arrest, acute arterial occlusion, major arrhythmia, deep venous thrombosis, and myocardial infarction. Infectious occurrences included empyema, sepsis, septicemia, and acute sinusitis. Adverse neurological events included central nervous system infections, progression of original neurological insult, and seizures.

Finally, since our initial primary endpoints yielded mixed results, we created a composite five-point ranked ordinal scale based on a combination of discharge status, functional status, and hospital days, to arrive at a summary patient outcome measure, referred to as the Composite Outcome Score (COS). The scale parallels the Glasgow Outcome Scale (GOS), with a grade of 1 being the worst and a grade of 5 being the best (Table 1). Grade 5 patients were discharged alive and independent within 30 days of admission. Grade 4 patients were discharged alive and independent more than 30 days after admission. Grade 3 patients were discharged alive and dependent less than 30 days after admission. Grade 2 patients were discharged alive and dependent more than 30 days after admission. Grade 1 patients died during their acute hospitalization.

Table 1 Composite outcome scale
Full size table

Statistical Analysis

The primary outcomes for the study were discharge status (alive vs. dead), functional status at discharge (independent vs. dependent), and ICU and hospital days dichotomized at their respective median values. The chi-square test was used to compare the timing of tracheostomy with dichotomous variables and the Mantel–Haenszel chi-square test was used to compare the timing of tracheostomy with ordinal variables. Comparisons with a P value < 0.20 from these chi-square tests were considered potential confounders, i.e., the distributions of timing of tracheostomy groups differed, to be controlled for in a multivariable analysis. These potential confounders were Injury Severity Score (ISS) (stratified as <15, 15–25, 26–40 and >40), age (dichotomized as 18–50 and >50 based on clinical grounds), pre-existing conditions (dichotomized as 0–1 pre-existing condition and >1 pre-existing condition), race (White, Black, other), sex, and injury type (blunt, penetrating, burn), as well as presence or absence of face, neck, thorax, abdominal, or spine injuries. Multiple logistic regression models were fit to assess the association of the timing of tracheostomy with each of the primary outcomes adjusting for the potential confounders, with results quantified using adjusted odds ratios (OR) and 95% confidence intervals (CI).

Results

In the approximately 15-year period from January 1990 until December 2005, 20,619 patients were admitted to PTSF participating hospitals with a presenting GCS ≤ 8 and evidence of a head injury. Of these 3,277 patients were found to have had a tracheostomy. On further examination, only 173 of these patients had isolated head injuries, while the vast majority (3,104) had a head injury as well as an injury to another body system (multisystem injury). Given the preponderance of patients with multisystem injuries and to provide for a more homogenous study cohort, we, therefore, chose to analyze only those patients with a severe head injury in the presence of an additional injury.

1,577 patients were found to have received a tracheostomy within 1–7 days and were designated the Early Tracheostomy Group (ETG). 1,527 patients were found to have undergone a tracheostomy after 7 days, these were designated the Late Tracheostomy Group (LTG). A comparison of the group demographics and presenting characteristics is presented in Table 2. Racial distribution was similar among groups (P = 0.85), but age and gender distributions were significantly different (P = 0.002 and P = 0.01, respectively). The majority of patients in both groups suffered blunt injuries, although the distribution was different between groups (P < 0.01). ETG patients tended to be slightly worse clinically on admission when compared to LTG patients, as measured by differences in admission GCS and ISS.

Table 2 Demographics and presenting characteristics
Full size table

Univariate Analysis

Results of an initial univariate analysis of the primary outcome measures are presented in Table 3. LTG patients had significantly longer hospital stays than ETG patients. Similarly, LTG patients spent a longer time in the ICU. A greater percentage of ETG patients were discharged in an independent state as compared to LTG patients. However, LTG patients were significantly more likely to be discharged alive as compared to ETG patients. Finally, LTG patients had significantly higher pulmonary, cardiac, infectious, and neurologic adverse occurrences compared to the ETG patients.

Table 3 Outcome measures: univariate analysis
Full size table

Multivariate Analysis

Length of Stay (Table 4)

LTG patients were four times more likely to have a longer ICU stay than ETG patients. Age, ISS group, pre-existing conditions, facial, neck, and thoracic injuries were also predictive of length of ICU stay. In a similar analysis of total hospital days, LTG patients were three times more likely to have a longer stay than ETG patients. ISS group, race, and facial injuries were also predictive of length of total hospital stay.

Table 4 Multivariate analysis: longer length of stay
Full size table

Functional Status (Table 5)

ETG patients were almost 1.5 times more likely to be functionally independent at discharge than LTG patients. Younger age and male gender were predictive of functional independence. Functional independence was also higher in groups with ISS scores less than 15, between 15 and 25, and 26–40, in comparison to the group with an ISS of greater than 40. On the other hand, patients with blunt injuries were almost three times as likely as patients with other mechanisms of injury to be dependent at discharge.

Table 5 Multivariate analysis: functional independence at discharge
Full size table

Mortality (Table 6)

In contrast to the advantages seen to ETG patients in terms of length of stay and functional independence, LTG patients demonstrated a survival benefit. LTG patients were slightly more than twice as likely to be discharged alive in comparison to ETG patients. Younger patients (ages 18–50) were 2.6 times as likely to be discharged alive as compared to patients greater than 50 years of age. Black race was an unfavorable predictor of survival. Patients with thoracic injuries were more likely to die when compared to those without such injuries. However, facial injury patients have a higher chance of survival.

Table 6 Multivariate analysis: discharge alive
Full size table

Pulmonary Occurrences (Table 7)

LTG patients were approximately twice as likely to suffer from an adverse pulmonary occurrence as compared to their ETG counterparts. Male gender and black race were also associated with a higher rate of adverse pulmonary occurrences. Patients with ISS scores less than 15, between 15 and 25, and 26–40 were less likely to sustain a pulmonary incident compared to those with an ISS of greater than 40.

Table 7 Multivariate analysis: adverse occurrences
Full size table

Cardiac Occurrences (Table 7)

Incidence of cardiac events was almost one and a half times as likely in LTG group. Age lower than 50, presence of pre-existing condition and an ISS score of 15–25 versus >40 were also associated with a lower rate of adverse cardiac occurrences.

Infectious Occurrences (Table 7)

LTG patients were about one and a half times more likely to have an infectious occurrence. Whites were less likely to incur an infection. An ISS score of more than 40 was predictive of a higher infection rate in comparison to scores less than 15, between 15 and 25, and 26–40.

Neurologic Occurrences (Table 7)

Patients with LTG were twice as likely to suffer an adverse neurologic incident as compared to ETG patients. Black race was associated with more frequent adverse neurologic occurrences. Having an ISS score between 26 and 40, versus >40 was associated with fewer neurologic occurrences. Thoracic injuries were also associated with fewer occurrences.

Composite Outcome Scale (Table 8)

As previously discussed, since analysis of tracheostomy timing with regard to the three primary outcome variables (length of stay, functional independence, and mortality) showed mixed results, we chose to combine all three measures into a Composite Outcome Scale (COS, Table 1). When analyzed according to the COS there was a nonsignificant trend toward a higher likelihood of a good outcome in ETG patients. This analysis also showed that patients in relatively good condition on admission as indicated by a favorable ISS, tended to have better outcomes on the COS. When the COS analysis was repeated using only those patients with ISS < 15 and 15–25 (good prognosis patients), LTG patients were found to be approximately 50% less likely to have a good COS outcome compared to ETG patients.

Table 8 Multivariate analysis: composite outcome scale (COS) COS (Table 1) is dichotomized and a good outcome (COS 4, 5) is modeled
Full size table

Discussion

The CDC estimates that over 200,000 individuals are hospitalized annually for traumatic brain injury [22]. For a portion of patients, airway access will be crucial to the management of their neurotrauma. This is particularly evident when the patient presents with a severe brain injury (GCS < 8). It is expected that the sickest patients will require the lengthiest ventilator dependent times [23] and having a neurologic insult puts these patients at a higher risk of undergoing tracheostomy due to the prolonged need for mechanical ventilation [24]. Current algorithms of care are presented with a dilemma as to when best to employ tracheostomy in this setting [25–27]. Various outcome measures have been studied in order to aid decision making. These have included traditional measures such as overall clinical outcome and length of hospital or ICU stay, as well as more respiratory specific measures, such as incidence of laryngeal damage, degree of improvement in respiratory mechanics, rate of ventilator-associated pneumonia, and ease of caretaking [28]. Several studies have compared tracheostomy with prolonged translaryngeal intubation [5, 25, 29, 30]. In addition, early tracheostomy has been compared to late tracheostomy both retrospectively and in some relatively small prospective series [31–33]. In general, studies suggest a higher incidence of positive outcomes in patients with tracheostomy, but this is by no means universal [4, 31, 32, 34, 35]. The influence of tracheostomy timing on mortality has been particularly mixed and functional outcome has not been well studied. Some reports have found improvement in length of stay parameters, but no change in mortality [36, 37]. In a meta-analysis by Griffiths et al., early tracheostomy was found to reduce the duration of artificial ventilation as well as intensive care unit length of stay [38]. However, this review also found no effect on mortality. Bouderka et al., in small cohort of patients, showed that early tracheostomy in head injury decreased both, total days of mechanical ventilation, and mechanical ventilation time after the development of pneumonia [25]. Yet, overall rates of pneumonia and mortality were not different between the two groups. On the other hand, in a small prospective study, Chintamani et al. found fewer intubation associated complications and improved mortality in patients receiving early tracheostomy as compared to prolonged mechanical ventilation [39]. Despite these multiple studies there remains no consensus to guide timing of tracheostomy these critically ill patients [14, 40, 41]. The American College of Chest Physicians’ Consensus Conference on Artificial Airways in Patients Receiving Mechanical Ventilation recommends tracheostomy in anticipation of intubation for greater than 21 days, but stops short of recommending specific timing prior to 21 days. Recent literature appears undecided as to contracting this window in favor of early tracheostomy.

In this study, we sought to examine the role of tracheostomy timing in a specific, well-defined sub-population, namely those patients with severe traumatic brain injury associated with other (multisystem) injuries. We chose this population largely due to the substantial number of patients available and the relatively high frequency with which tracheostomy was performed. In addition to the traditional outcome measures of mortality and hospital length of stay employed in the majority of studies regarding tracheostomy timing, we added a measure of functional independence (the FIM score—Functional Independence Measure). Functional status is obviously an important aspect of patient outcome and one that is often overlooked in studies of tracheostomy timing, which tend to focus on length of stay and/or length of ventilator dependence. In the setting of neurological injury, functional outcome is of paramount importance.

Our results examining rates of pulmonary, cardiac, infectious, and neurological adverse occurrences demonstrated an advantage for early tracheostomy in all categories. However, initial analysis of the three primary outcome measures (length of stay, functional independence, and survival) suggested an inconsistent benefit for early tracheostomy. While ETG patients were more likely to be functionally independent and have a shorter length of stay, LTG patients were slightly more than twice as likely to be discharged alive. From these three outcomes, we, therefore, created a Composite Outcome Scale, whereby a good outcome was defined in a “patient-centric” manner, as independent survival regardless of length of stay. Using this scale there was a trend toward a higher likelihood of a good outcome in the ETG. However, this trend was not statistically significant (Table 8).

This runs contrary to many, but not all previous studies of tracheostomy after trauma or traumatic brain injury specifically, which have suggested a benefit for early tracheostomy. Most likely this lack of significance is related to the higher likelihood of death seen in the ETG group. This probably reflects a selection bias in the study population. Since the study is not randomized, early tracheostomy by its very nature will have been performed in patients without the definitive knowledge of their ability to survive. In contrast, LTG patients will have already survived at least 7 days, making their overall chances of survival higher. Although we made an effort to control for the patient’s illness severity by including admission ISS scores as independent variables in the multivariate analyses, these scores do not predict mortality with complete accuracy and other factors, not recognized or studied, may be important.

In a recent article by Schauer et al., utilizing a similar database, patients with a high probability of survival were found to be more likely to benefit from early tracheostomy [42]. The authors used the Trauma and Injury Severity Score (TRISS) to calculate the probability of survival (P). Patients with a low probability of survival (low Ps) had a higher rate of mortality with early tracheostomy. However, among patients with high Ps, those in the early tracheostomy group were found to have significantly reduced ICU days, total ventilator days, hospital days, and rates of pneumonia. Although our dataset did not include the TRISS specifically, presenting ISS was recorded. Based on this information, we performed a post-hoc analysis including just the patients with relatively good ISS (those in the ISS < 15 and 15–25 groups). Among this “good prognosis” group, we found that ETG patients were just over twice as likely to have a good outcome using our Composite Outcome Scale (Table 8). When we performed the same analysis for those with poor ISS, we found that tracheostomy timing had no effect on outcome (data not shown). This analysis, therefore, also supports the conclusion that the patients in better clinical condition are the ones most likely to benefit from early tracheostomy.

Clearly, this study is not without its limitations. The retrospective nature of the data makes selection bias a definite concern as described. However, we have attempted to control for intergroup differences in presenting characteristics through rigorous multivariate statistical methods and our post-hoc analyses have helped clarify some potentially concerning issues. The data itself are prospectively entered and participating facilities would not be expected to have knowledge of future data examinations or studies. In addition, data fidelity is an important part of the PTSF mission and regular external audits are standard. The FIM score may be relatively crude measure of independence, but this is a widely used, reproducible tool. The cut off dates of 7 days for tracheostomy and 30 days for hospital stay may be viewed as somewhat arbitrary. The definition of early tracheostomy was made prior to any data analysis, based on definitions used in the published literature. Subsequent, additional post-hoc analyses using either 5 days or 10 days yielded the same results (data not shown). Likewise, dichotomizing the hospital length of stay at 30 days was determined prior to the data analysis and changing this to 15 days or 45 days did not change the overall result (data not shown). Finally, since some results are based on post-hoc analyses there is an increased risk of a false negative conclusion or Type II error.

Conclusions

To date, this is the largest study to report the effects of tracheostomy timing on outcome after severe head injury. Not only does this report benefit from a considerable number of patients, but it also provides a window into the functional outcome of these patients that is not usually afforded in studies of this type. In summary, our results suggest that a strategy of early (hospital days one through seven) tracheostomy, particularly when performed on patients with a reasonable chance of survival (favorable ISS), results in a better overall clinical outcome (fewer adverse events, shorter length of stay, and higher likelihood of functional independence) than when the tracheostomy is performed in a delayed manner (greater than 7 days). These findings indicate a complex relationship between tracheostomy timing and outcome. Results may vary based on the outcome measure selected and may be further influenced by the severity of the patient’s injuries. A prospective randomized trial may help further define and clarify the role of tracheostomy timing in head-injured patients. However, it is important that such a study include a measure of functional status and perhaps only focus on those patients with a relatively high chance survival.