Hemorrhagic stroke is a devastating cerebrovascular event associated with an immense degree of morbidity and mortality.^13,24,25 Mortality in the subset of patients with hemorrhagic stroke who require mechanical ventilation is especially high, ranging in the literature from 40% to 80%.^22,23 These patients generally require stabilization within an ICU, which increases their risk of nosocomial infection and medical complications.²⁰ An estimated 19% of ICU patients develop an infection during their ICU stay compared to 5% of hospitalized patients in general. This observation may be attributed to the need for temporary biomedical interventions, such as endotracheal tubes, indwelling urinary catheters, and vascular access lines, as well as the complex, multitiered interactions necessary between various providers in the ICU. Thus, efforts to characterize early interventions that may minimize the duration of hospital admission, particularly the ICU length of stay, and potential for medical complications while increasing functional outcomes and survival are of paramount interest.

Endotracheal intubation is often necessary in patients with hemorrhagic stroke who are critically ill; however, stroke patients are often intubated for the purpose of airway protection as opposed to a true requirement for mechanical ventilation.^22,23,28 In the general ICU setting, controversy exists regarding the decision to convert to tracheostomy and its optimal timing. Tracheostomy is customarily considered if it is anticipated that a patient will require long-term intubation or cannot be safely extubated.³ Some practitioners advocate for early conversion to tracheostomy because of its potential benefits over translaryngeal intubation, such as decreased risk of laryngeal and oropharyngeal lesions, facilitation of nursing care and lower-airway suctioning, increased patient comfort, and reduced sedation requirements.^2–5,21 However, the decision to proceed with tracheostomy must be considered in light of potential but uncommon complications such as pneumothorax, stomal site infection, tracheal stenosis, or tracheoesophageal fistula.

In addition to airway protection, adequate nutrition is essential for patients recovering from stroke. Malnutrition exacerbates neurological injury and leads to longer hospitalizations and worse outcomes.^11,15 To address this issue, patients are given enteral feeding, often by nasogastric tube. Prolonged feeding by nasogastric tube is associated with various complications including nasal lesions, chronic sinusitis, gastroesophageal reflux, and aspiration pneumonia.¹⁰ Percutaneous endoscopic gastrostomy (PEG) is an alternative method available to provide feeding in critically ill patients, which may avert the aforementioned complications. A meta-analysis demonstrated that PEG utilization reduces the likelihood of feeding failure (e.g., clogging or dislodgement of tube). PEG is generally considered when a patient is anticipated to have diminished or absent oral intake for an extended period. In the setting of hemorrhagic stroke, this may be due to a reduced level of consciousness secondary to neurological injury or to persistent postictal dysphagia. Much as they do with tracheostomy, providers making the decision to place a PEG tube must balance the perceived risks and benefits on a case-by-case basis. In this setting, the aim of our study was to review the tracheostomy and PEG placement data in our patients with hemorrhagic stroke in order to identify factors associated with earlier placement and to evaluate outcomes.

Methods

The University of Kentucky Institutional Review Board reviewed and approved the study protocol. The neurosurgery registry was screened for consecutive adult patients admitted between June 1, 2011, and June 1, 2015, using ICD-9 codes. In this study, hemorrhagic stroke was defined as either subarachnoid hemorrhage (SAH; ICD-9 430) or intracerebral hemorrhage (ICH; ICD-9 431). Among individuals with these ICD-9 designations, those requiring mechanical ventilation were selected for inclusion in the study. Our institutional practice is to perform tracheostomy and/or PEG for patients with hemorrhagic stroke if initial extubation fails and they have confounding factors that would lead to a high risk of early reintubation. We do not have strict guidelines for this but instead make decisions in a multidisciplinary setting for each patient. Patients with social discharge issues resulting in hospital stays > 100 days were excluded from analysis. We also excluded patients who left against medical advice, because their treatments were incomplete. Patients who died within 72 hours of admission (i.e., those admitted with nonsurvivable injuries) were excluded from analysis because they were not ever considered for tracheostomy or PEG tube placement.

Patient data were retrieved from our institution’s electronic medical records and stored on a secure REDCap (Research Electronic Data Capture) database. The following variables were collected from patients’ records: age, sex, admission diagnosis (SAH or ICH), Hunt and Hess grade for SAH patients and ICH score for ICH patients, method of treatment (medical vs surgical for ICH, endovascular vs clipping for aneurysmal SAH), hospital day of tracheostomy and/or PEG, ICU length of stay, overall hospital length of stay, and discharge disposition. When absent from patients’ records, Hunt and Hess grades and ICH scores were retrospectively calculated based on charted data. The presence of the following comorbidities on admission was recorded as yes or no: hypertension, diabetes mellitus, chronic obstructive pulmonary disease, and pneumonia. Last, occurrence of the following complications during hospitalization was recorded as yes or no: deep venous thrombosis (DVT), bacteremia, ventriculitis, pneumonia, and urinary tract infection (UTI). Any complications arising from the performance of tracheotomy or PEG tube placement were also recorded.

Data were analyzed using RStudio (RStudio, Inc.). Logistic regression and multiple linear regression modeling were used to 1) analyze risk factors for tracheostomy and/or PEG, and 2) evaluate the impact of tracheostomy and/or PEG timing on patient outcomes. A logarithmic transformation was applied to nonnormally distributed variables prior to the analysis. Odds ratios, p values, and confidence intervals were calculated, with p < 0.05 considered statistically significant.

Results

Three hundred sixty-six patients with hemorrhagic stroke requiring mechanical ventilation were treated at our institution between 2011 and 2015. One hundred twenty-six of these patients had nonsurvivable injuries (i.e., the patient died within 72 hours of admission) and were excluded from analysis. After excluding fatalities that occurred within the first 72 hours, 240 patients remained: 113 of these individuals were diagnosed with SAH and 127 were diagnosed with ICH (Table 1).

We performed an analysis to identify patient risk factors associated with tracheostomy and PEG placement. In our study population, the presence of pneumonia on hospital admission (OR 4.98 [95% CI 3.68, 6.29]; p = 0.02) and the diagnosis of SAH (OR 1.90 [95% CI 1.31, 2.49]; p = 0.03) were significantly associated with undergoing tracheostomy. Presence of pneumonia on admission was significantly associated with PEG tube placement (OR 5.92 [95% CI 3.92, 6.09]; p = 0.01). With few exceptions, patient risk factors did not appear to be associated with the timing of tracheostomy and/or PEG tube placement in our population. Chi-square analyses revealed no significant differences in terms of prognostic score (Hunt and Hess grade and ICH score for SAH and ICH, respectively) between patients with hemorrhagic stroke undergoing early tracheostomy and/or PEG (i.e., within 1 week), and late tracheostomy and/or PEG (i.e., after the 1st week) (Tables 2 and 3). Furthermore, no significant differences were revealed in terms of age or sex distributions between the patients undergoing early versus late tracheostomy and/or PEG. However, in patients with ICH, the presence of pneumonia on admission was more commonly reported in the group undergoing late (p = 0.05) than the one with early tracheostomy. Also in patients with ICH, hypertension was a more commonly reported comorbidity in the group undergoing late PEG placement compared to those undergoing early PEG placement (p < 0.01). Low individual observed sample sizes for endovascular embolization or surgical clipping of aneurysms limited statistical power to analyze these variables as factors related to the performance or timing of tracheostomy or PEG. No significant association was found between the prognostic score of patients’ strokes and placement of tracheostomy or PEG.

Further analysis was conducted to assess whether any association existed between timing of tracheostomy or PEG tube placement and 1) ICU length of stay or 2) overall length of stay for patients surviving to discharge. Analysis revealed that for every 2-fold delay in timing of tracheostomy, patients’ ICU length of stay increased by 29.50% (p < 0.01). Moreover, for patients who had a tracheostomy after hospital day 14, there was a 49.40% increase in ICU length of stay compared with those undergoing tracheostomy within the 1st week (p < 0.01). The average ICU length of stay for patients undergoing tracheostomy within the 1st week was 15.59 days; in contrast, for patients undergoing tracheostomy after hospital day 14, the average ICU length of stay was 23.29 days (Fig. 1A). For overall length of hospitalization, each 2-fold delay in timing of tracheostomy was associated with an 18.90% increase in length of stay (p = 0.02). For patients undergoing tracheostomy after hospital day 14, there was a 25.80% increase in overall length of stay compared to those undergoing tracheostomy within the 1st week (p = 0.04). The average overall length of stay for patients undergoing tracheostomy within the 1st week was 25.19 days; in contrast, for patients undergoing tracheostomy after hospital day 14, the overall length of stay was 31.70 days (Fig. 1A). Similarly, every 2-fold delay in timing of PEG tube insertion was associated with an 18.90% increase in overall hospital length of stay (p = 0.01). Compared to patients undergoing PEG placement in the 1st week, patients who had PEG tubes placed after hospital day 14 had a 23.02% increased length of stay. The average overall length of stay for patients with PEG tubes placed in the 1st week was 25.81 days; in contrast, the average overall length of stay for those who had PEG tubes placed after hospital day 14 was 31.76 days (Fig. 1B).

Histograms depicting mean length of stay (in number of days) during ICU admission and overall hospitalization. Panel A demonstrates these values for tracheostomy; panel B demonstrates these values for PEG. Asterisk denotes statistical significance.

Additionally, we examined the relationship between timing of tracheostomy or PEG tube placement and patient survival and rates of complications during hospitalization. Our regression analyses indicated that timing of tracheostomy and PEG tube placement was not associated with mortality rate. However, chi-square analysis did indicate a statistically significant difference between the numbers surviving to discharge in ICH patients undergoing early tracheostomy versus late tracheostomy. No associations were found between timing of tracheostomy and PEG and the occurrence of DVT, bacteremia, pneumonia, or UTI during patients’ hospital courses. Importantly, we were unable to quantify aspiration events or differentiate between ventilator-associated pneumonia (VAP) and other etiologies of pneumonia from the retrospective medical chart review; therefore, we recorded the occurrence of pneumonia during admission without distinguishing the etiology. Too few cases of ventriculitis occurred (n = 4) for statistical analysis.

Complications associated with tracheostomy and PEG in this population were infrequent (n = 4). Two patients had complications related to tracheostomy: one patient developed vocal cord paralysis and another had stomal site hemorrhage. Two patients had complications related to PEG: one patient developed ileus and had persistent nausea and emesis after PEG placement; another sustained a liver laceration during placement, developed a hematoma, and later required removal of the PEG tube due to leakage.

Discussion

The decision to convert to tracheostomy and/or PEG and the timing of these procedures is controversial. Some suggest that, in appropriate patients, an early tracheostomy is preferable to extended translaryngeal intubation with tracheostomy at a later date if necessary.^1,2,4,5,21 The term “early” is defined differently across various published series; often it refers to tracheostomies performed within 1 week of initial intubation.⁹ Conclusions from analyses of early tracheostomy in mixed ICU populations have been conflicting. A randomized controlled trial conducted in the United Kingdom with 899 patients from a variety of medical and surgical ICU services reported no benefit in mortality or other major clinical outcomes for patients undergoing early compared to late tracheostomy.³¹ In contrast, retrospective analyses of mixed ICU populations have reported trends of decreased time spent on a ventilator and shortened ICU length of stay.^1,2,4,5,9

Because the underlying pathophysiology and indications for intubation in patients with hemorrhagic stroke who are critically ill differ in important ways from those in nonneurological ICU patients,^22,23,28 the timing of tracheostomy may have more nuanced effects on the course of care and outcomes in the former group that should be more thoroughly investigated. In that vein, it has been proposed that early conversion to a tracheostomy for stroke patients could lead to earlier cessation of sedation, permitting these individuals to awaken, wean from mechanical ventilation, and engage in active rehabilitation regimens more quickly.²³ Unfortunately, prospective and randomized data regarding early tracheostomy specifically in stroke patients, particularly those with hemorrhagic stroke, are limited.³ One randomized trial of early tracheostomy in 60 stroke patients (including both hemorrhagic and ischemic stroke), the Stroke-Related Early Tracheostomy versus Prolonged Orotracheal Intubation in Neurocritical Care Trial (SETPOINT), indicated that early tracheostomy is safe and was associated with reduced need for sedatives.³ Most of the data that have been published come from subgroup analyses of larger mixed cohorts, which have reported findings such as reduced ICU length of stay, less time spent on ventilator, and even reduced mortality with early tracheostomy.^{3,18,19,23,27} One retrospective analysis specifically examining the timing of tracheostomy in poor-grade SAH (i.e., World Federation of Neurosurgical Societies classification III–V) found reduced incidence of pneumonia, shorter duration of mechanical ventilation, and shorter duration of endotracheal cannulation for patients undergoing early tracheostomy.⁹ The SETPOINT2 randomized trial, a follow-up to SETPOINT, is currently underway and seeks to elucidate the impact of early tracheostomy on functional outcome in stroke patients (as quantified via the modified Rankin Scale) and other clinical parameters such as ICU length of stay, mortality, and duration of sedation.²³

Literature regarding early PEG placement is similarly mixed, and clinical practice often diverges from clinical guidelines with regard to recommendation for and timing of PEG placement. In one analysis of the Nationwide Inpatient Sample, the median timing of PEG placement between various hospitals was reported to range from approximately 3 days to 3 weeks from date of admission.⁸ Reasons for earlier PEG are varied, but a commonly cited rationale is that earlier PEG placement can potentially facilitate expeditious discharge from the hospital and subsequent admission to an acute rehabilitation or skilled nursing facility.²⁹ Besides the perceived logistical benefits, some studies have described clinical outcome benefits associated with early PEG. For example, in a series of patients with stroke or head injuries requiring mechanical ventilation, early PEG (i.e., within 24 hours of intubation) resulted in decreased rates of VAP compared with nasogastric feeding.¹² In contrast, the FOOD (Feed Or Ordinary Diet) trial, a randomized controlled trial of stroke patients, concluded that early PEG had no benefit and possibly increased the likelihood of death or poor outcome when compared with nasogastric tube feeding.⁶ Despite the existence of observational and randomized studies regarding timing of tracheotomy and PEG in various populations, few studies have specifically examined their role in the treatment of patients with hemorrhagic stroke.

Identification of patient factors associated with tracheostomy and PEG placement may facilitate the selection of appropriate candidates and optimize early performance of tracheostomy and PEG in patients who might derive the most benefit. We performed a logistic regression analysis to evaluate patient factors associated with tracheostomy and PEG placement in this population. Our logistic regression analyses indicated that the presence of pneumonia on admission was associated with increased likelihood of tracheostomy and PEG tube placement. The presence of pulmonary comorbidities has been cited previously as a predictor of tracheostomy placement.³⁰ To our knowledge, pneumonia or other pulmonary comorbidities have not been reported as an independent risk factor for PEG placement. A few clinical scoring tools such as the SETscore (stroke-related early tracheostomy score), which can be applied to patients with hemorrhagic stroke, have been developed to predict duration of mechanical ventilation and determine the need for tracheostomy.²² The SETscore assigns points in 3 domains: neurological function, neurological lesion, and general organ function. A detailed breakdown of these parameters and respective point values is available in the original article. Reportedly using a cutoff of 8 points, this scoring model predicts neuro-ICU length of stay, ventilator time, and need for tracheostomy with a sensitivity of 64% and a specificity of 86%. We did not collect the requisite parameters to test this scoring tool in our population.

With regard to PEG, a similar clinical algorithm called the GRAVo (Glasgow Coma Scale score, Race, Age, ICH Volume) score has been developed to predict the need for PEG in patients with ICH,⁷ although further studies validating widespread application of this model are necessary. This scoring model assigns points for Glasgow Coma Scale score of ≤ 12, black race, age > 50 years, and ICH volume > 30 ml. Reportedly, a score of > 4 is associated with 12-fold higher odds of PEG placement compared to a score of ≤ 4 in patients with ICH. We also did not collect the parameters necessary to evaluate this scoring tool in our patients. Ideally, further investigations could potentially yield an algorithm to reliably identify patients with hemorrhagic stroke for whom both early tracheostomy and PEG placement may be especially beneficial.

In this study population, diagnosis of SAH was significantly associated with tracheostomy and PEG placement. In fact, our SAH patients underwent tracheostomy at a rate of 38.05% compared to 25.20% for ICH patients (Table 1). Likewise, SAH patients underwent PEG tube placement at a rate of 39.82% compared to 32.28% for ICH patients. At our institution, patients with SAH tended to be admitted to the neurosurgical service, whereas those with ICH were generally admitted to the neurology service. Different practitioners and decision-making processes between these services may have played a role in the marked propensity for SAH patients to undergo tracheostomy. The discrepancy in rates of tracheostomy and PEG incidences between our SAH and ICH populations highlights this observational study’s limitations, but it suggests a need for further investigation in this area.

We did find that the reported rate of hypertension in ICH patients undergoing late PEG tube placement differed significantly from that in ICH patients who received earlier PEG placement. More precisely, 7 of 11 ICH patients who underwent early PEG had a history of hypertension compared to 21 of 21 ICH patients who received late PEG. It is conceivable that hemodynamic stability could be related to the choice to defer PEG. However, we believe this could represent a random association rather than a causative issue. We also found that the reported rate of pneumonia on admission differed significantly between ICH patients undergoing early and late tracheostomy. This finding was quite close to the significance threshold (p = 0.05) and would require a larger study to confirm.

In patients surviving to discharge, earlier tracheostomy was associated with a shorter ICU stay. Our analysis showed that for each 2-fold increase in day of tracheostomy, the patient’s ICU stay was increased by 18.90%. This finding of reduced ICU length of stay has been described in some previous series of patients receiving neurocritical care.^18,19,27 A randomized trial of early tracheostomy in 60 stroke patients, the SETPOINT study, indicated that early tracheostomy is safe and was associated with reduced need for sedatives.³ However, that study did not find an association between early tracheostomy and reduced length of ICU stay, although there was a suggestion of benefits in mortality rate and incidence of complications. Due to its small size, that study lacked power and may have been unable to detect benefits demonstrated in previous analyses. A larger-scale randomized trial evaluating tracheostomy in stroke patients is needed; the SETPOINT2 study is underway, and aims to enroll more patients than SETPOINT and to examine a variety of clinical outcomes, including ICU length of stay.²³

We also found that earlier tracheostomy and PEG were associated with shorter overall hospitalization in our study population. Prior series have described shorter overall length of hospitalization in patients undergoing earlier versus delayed tracheostomy, although findings generally do not reach statistical significance.^2–5,21 A previous study reported that PEG placement was associated with shorter overall length of stay in the hospital when performed concomitantly with tracheostomy;¹⁶ however, that study did not specifically evaluate early PEG placement, nor was it specific to patients with hemorrhagic stroke. Guaranteeing adequate enteral nutrition in critically ill patients is an essential part of recovery.^11,15 Furthermore, stroke patients often require dysphagia evaluations, which may identify feeding difficulties, requiring PEG. Early PEG placement may alleviate discharge delays that can occur when prolonged feeding difficulty is diagnosed late in the inpatient hospital stay.

In ICH patients undergoing early tracheostomy, we found a significant difference in the percent of patients surviving to discharge compared to those undergoing late tracheostomy. In terms of actual frequencies, 12 of 12 ICH patients who underwent early tracheostomy survived compared to 11 of 15 ICH patients in whom late tracheostomy was performed. This finding barely reached the threshold of statistical significance (p = 0.05) and does not concur with our regression analysis, which found no significant association between mortality and timing of tracheostomy. Although these data suggest a possible association, a larger study would be needed to more adequately analyze this particular issue.

Some controversy exists with regard to placing a PEG tube in a patient who has undergone CSF shunting for hydrocephalus. A published gastrointestinal guideline cites ventriculoperitoneal shunting (VPS) for hydrocephalus as an absolute contraindication to performance of PEG;²⁶ alternatively, other published guidelines cite VPS as only a relative contraindication.¹⁴ In stark contrast, a systematic review by Oterdoom and colleagues concluded that VPS should not be considered a contraindication to PEG placement.¹⁷ We did not specifically collect data regarding the timing of CSF shunt or drain placement relative to tracheostomy or PEG; however, at our institution, we do not have any protocol regarding the delay of shunting relative to PEG placement. We do not perform PEG and VPS in the same procedural setting, but we have no minimum requirement for timing intervals between the two. Because there is no protocol on this, we believe shunting would have no impact on tracheostomy or PEG timing.

This study suffers from several limitations. It is limited by its retrospective, observational design. ICD-9 coding was used to identify patients for inclusion in the study, and it is possible that inaccuracies existed in the initial diagnoses and coding for these patients. Although we reviewed each patient’s chart to ensure accuracy, we cannot discount the possibility that some patients were never included due to improper coding (i.e., ICD-9 designations other than 430 or 431). The retrospective nature of this study additionally limited our ability to glean certain information that was not readily available in the medical records (e.g., precise etiology of pneumonia [VAP vs other]). Furthermore, although it was not the initial aim of the present study, it would be prudent to analyze subgroups of our hemorrhagic stroke population, such as patients with angiographically demonstrated aneurysmal SAH. However, given our relative total numbers, we did not believe post hoc separation of SAH by cause to be appropriate or statistically sound. In future studies, we will consider gathering information to perform subgroup analyses, such as in patients with aneurysmal SAH. Because this was an uncontrolled retrospective study of patients on multiple inpatient services, the natural practice biases of individual providers may have played a role in variances toward a decision for tracheostomy and/or PEG. Despite these limitations, we were able to demonstrate some clear relationships between underlying patient factors and decisions for performance of these procedures.

Conclusions

Hemorrhagic stroke is a devastating event. Early identification of patients who may benefit from aggressive resuscitative measures such as tracheostomy and/or PEG placement may optimize patient outcomes. Our study revealed a minimal rate of complications due to tracheostomy and PEG, identified patient factors associated with these interventions, and demonstrated a reduced length of stay for patients undergoing earlier tracheostomy and PEG placement. Of particular interest, our data demonstrate that individuals who had earlier tracheostomies tended to have shorter ICU admissions. Optimizing the care of critically ill patients and limiting the time admitted to the ICU can potentially help in controlling healthcare costs and reduce the likelihood of ICU-associated complications.

Disclosures

Dr. Fraser has direct stock ownership in Fawkes Biotechnology, and he is a consultant for Stream Biomedical and Medtronic.

Author Contributions

Conception and design: all authors. Acquisition of data: McCann. Analysis and interpretation of data: all authors. Drafting the article: McCann. Critically revising the article: Fraser, Hatton, Vsevolozhskaya. Reviewed submitted version of manuscript: all authors. Statistical analysis: McCann, Vsevolozhskaya. Study supervision: Fraser.