The World Health Organization World Population Aging 2019 census reported population aging as a global phenomenon, with the proportion of persons aged 65 yr or over increasing from 6% in 1990 to 9% in 2019, and an expected increase to 16% by 2050. In Canada, 17.6% of the population is 65 yr or over, with a projected increase to 25% by 2050.1 This aging population will challenge health systems, including intensive care units (ICU). Increasingly, elderly patients and their surrogates will be faced with ICU treatment decisions. Clinicians will need to provide accurate information about outcomes following the use of life-sustaining interventions.

One such intervention is tracheostomy, which is performed in 9–15% of patients undergoing mechanical ventilation.2,3,4,5,6 Although variable among countries and ICU teams, the overall prevalence has increased. Tracheostomy in the ICU is mainly indicated for patients with acute respiratory failure, neuromuscular disease, or trauma with expected prolonged duration of mechanical ventilation and difficult or protracted course of weaning.3,6 The proposed benefits of tracheostomy compared with prolonged translaryngeal endotracheal intubation include reduction in laryngeal injury, facilitation of oral hygiene and pulmonary toilet, and improved patient comfort with reduced sedative requirements.7 While studies on ICU tracheostomy have reported expedited weaning from mechanical ventilation, reductions in ventilator-associated pneumonia, and reduced ICU length of stay,8,9 subsequent larger and international studies have failed to replicate these findings.2,10 Tracheostomy has not been shown to reduce mortality,2 and its insertion and subsequent management have been associated with risks and complications. Currently, there are few data on outcomes following tracheostomy in elderly critically ill patients. Single-center studies have described higher rates and duration of ventilator dependence and poor quality of life due to functional impairment. Reported mortality rates range from 22% to 74% with a low rate of successful decannulation (0–10%) and the possibility of permanent tracheostomy.11,12,13,14,15,16,17

With global aging and the proportional increase in elderly patients requiring mechanical ventilation and ICU support, it is important to determine tracheostomy outcomes in this group of patients. Weighing the benefits of tracheostomy with the expected risks and outcomes will affect decision-making by the healthcare team, patients, and their families. In this study, we sought to describe the outcomes of critically ill elderly patients who underwent tracheostomy in the ICU. The primary outcomes were ICU and hospital mortality, and secondary outcomes were lengths of stay, in-hospital decannulation rate, discharge destination, discharge functional status, and frailty.

Methods

We conducted a historical cohort study in three ICUs at two tertiary-care academic hospitals in Toronto, Canada. They were the medical-surgical ICU at Sinai Health System, and the medical-surgical ICU and cardiac ICU at Sunnybrook Health Sciences Centre. These three ICUs provide care for medical, surgical, trauma, neurology, and post-cardiac surgery patients. In both hospitals, mechanically ventilated patients are cared for exclusively in the ICU. All patients aged 70 and above, who received tracheostomies during their ICU stay between 1 January 2010, and 1 June 2016, were included in a retrospective chart review. The study period was selected to reflect contemporary practice and outcomes, information which would be most informative for clinicians and patients. Patients were excluded if they were admitted with a pre-existing tracheostomy or if they had a tracheostomy for failed intubation or a planned surgical procedure. The Institutional Research Ethics Boards at Mount Sinai Hospital (REB 13-0179-C) and Sunnybrook Hospital (CTO project ID 0757) approved the study and waived the need for consent.

A standardized case report form was used to collect data from electronic medical records. Interdisciplinary team progress notes in paper charts were reviewed if electronic records were incomplete. Demographic data including age, sex, reasons for hospital and ICU admission, and comorbidities were collected. Tracheostomy data included indication for tracheostomy, early (< 48 hr from tracheostomy) complications (such as bleeding, pneumothorax, pneumomediastinum, subcutaneous emphysema, esophageal injury), and late (≥ 48 hr from tracheostomy) complications (such as accidental decannulation, pneumomediastinum, trachea-esophageal fistula, and bleeding). We recorded time to decannulation and duration of mechanical ventilation (MV), which was defined as time from tracheostomy (day 0) to successful liberation (> 48 hr) from MV or death. We recorded the primary outcomes in the ICU and hospital as well as mortality. For nonsurvivors, demise following withdrawal of life support or decisions to withhold vasopressors, inotropes, or MV was noted. Data on the secondary outcome measures included length of stay, discharge destination, and discharge functional status including mobility, feeding, vocal function, swallowing assessment, and Clinical Frailty Score (CFS).18 In addition, data on the timing and the type of percutaneous feeding tube insertion were collected. Vocal function and swallowing outcomes were based on clinical evaluation and video swallowing assessment conducted by a speech-language pathologist close to the patient’s day of discharge.

We calculated descriptive statistics for all variables of interest. Continuous measures such as age are summarized using means and standard deviations (SD) or medians with interquartile ranges [IQRs], whereas categorical measures are summarized using counts and percentages. The outcomes of mortality and decannulation were binary in nature and therefore analyzed using logistic regression models. Prior to model development, we assessed the set of predictor variables of interest for the presence of multicollinearity using tolerance statistics. We used a tolerance value of < 0.4 as the cut-off point for the presence of multicollinearity. In such cases, only one member of a correlated set was retained for the multivariable model. Multicollinearity was not found to be a concern. The results of the logistic regression models are presented as odds ratios (ORs) and their associated 95% confidence intervals (CIs). The choice of variables to include in the logistic regression models was based on clinical importance and statistical limitations. The number of variables permitted in a logistic regression model was based on the number in the smaller outcome category divided by ten (Regression Modeling Strategies, Frank Harrell, 2001, Springer-Verlag New York, Inc., New York, NY, USA). For decannulation, this allowed a total of six variables. The predictors for consideration in both models included age, frailty on admission, comorbidities, hospital admission, ICU length of stay, and hospital length of stay. Hospital length of stay was dropped in favor of ICU length of stay. The model on decannulation also included the reason for tracheostomy. No other variables were considered for inclusion in the analyses. All statistical analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC, USA).

Results

Patient demographics

Overall, 270 critically ill patients ≥ 70 yr had a tracheostomy procedure during the study period. The mean (SD) age was 81 (6) yr, and 59% were men (Table 1 and Electronic Supplementary Material [ESM] eTable 1). The most common indication for ICU admission was acute respiratory failure (147/270, 54%). Patients were intubated for a median [IQR] of 14 [8–21] days prior to tracheostomy, and 41% (110/270) had at least one previous intubation during the index hospital admission (Table 2). The most common indication for tracheostomy was prolonged intubation (202/270, 75%). Approximately half (147/270, 54%) of tracheostomies were performed with an open procedure, and the remainder (123/270, 46%) were performed percutaneously.

Table 1 Patient characteristics
Full size table
Table 2 Intubation and tracheostomy details
Full size table

Tracheostomy complications

Both early and late complications were rare (Table 2). There were 3/270 (1%) cases of bleeding during the tracheostomy procedure, and 3/270 (1%) cases of early tracheostomy site bleeding within 48 hr following the procedure; all occurred following open procedures. With regard to late complications (more than 48 hr following procedure), there were 4/270 (1%) accidental decannulations, 4/270 (1%) cases of tracheomalacia, 8/270 (3%) cases of bleeding at the tracheostomy site, and 6/270 (2%) cases of skin ulceration around the tracheostomy site. None of the documented complications were directly associated with mortality.

Outcomes

Intensive care unit and hospital mortality were 26% (68/270) and 46% (125/270), respectively (Table 3). Overall, 19% (52/270) required readmission to the ICU. Of the 68 patients who died in the ICU, 28 (43%) withdrew or withheld life-sustaining therapies (vasopressors or MV). Figure 1 summarizes the patient outcomes.

Table 3 Hospital outcomes
Full size table
Fig. 1
figure 1

Outcomes

Full size image

The median [IQR] ICU and hospital lengths of stay were 31 [17–53] and 81 [46–121] days, respectively. Only 15% (41/270) were liberated from MV during the index ICU admission; the median [IQR] duration of MV following tracheostomy in these patients was 7 [3–19] days (Table 3). Overall, only 25% (67/270) of patients were successfully decannulated during the index hospital admission. The median [IQR] time from tracheostomy to decannulation was 41 [25–68] days. Of the 145 patients who survived the hospital admission, 56% (82/145) were discharged with the tracheostomy in situ; and 27% (22/145) still required MV support.

Prior to hospitalization, most patients were from home (227/270, 84%), orally fed (259/270, 96%), and independently ambulatory (163/270, 60%) (Table 1). The median [IQR] CFS was 4 [3–5], indicating they were vulnerable. Of 145 hospital survivors, 41 (28%) were discharged to another acute care hospital, 34 (23%) to chronic care centers, 39 (27%) to rehabilitation centers, six (2%) to weaning or long-term ventilation centers, 17 (12%) directly home, and seven (5%) to a palliative care unit. At the time of hospital discharge, most patients remained tube-fed (92/145, 63% of survivors), unable to ambulate with or without assistance (96/145, 66% of survivors), with a median CFS of 7 (severely frail) (Table 4). Comparison of CFS on admission and prior to discharge showed a significant deterioration (Fig. 2). Eighty of 145 survivors (55%) had at least one formal swallowing assessment performed prior to discharge; with 44% (35/145) having moderate to severe dysphagia. Almost half of all patients (117/270, 43%) had a percutaneous gastrostomy (PEG) or gastrojejunostomy tube placed during their hospitalization. At the end of hospitalization, only 25% (67/270) of patients were communicating verbally after tracheostomy decannulation or by using a speaking valve; the rest were limited by illness severity, need for MV, or low level of consciousness.

Table 4 Functional outcomes
Full size table
Fig. 2
figure 2

Change in Clinical Frailty Score. Connected line plot showing the change in Clinical Frailty Score (CFS) prior to ICU admission and prior to hospital discharge [mean (standard deviation) CFS, 4.25 (1.34) vs 7.26 (0.79), P < 0.001). The colored lines indicate individual patients and the bold line represents the mean CFS. ICU = intensive care unit

Full size image

Predictors of outcome

Age was a significant predictor of hospital mortality (OR, 1.05; 95% CI, 1 to 1.09; P = 0.05), while comorbidities, reasons for hospital admission, and frailty on admission were not found to be significant (ESM eTable 2). For decannulation, older age (OR, 0.94; 95% CI, 0.89 to 0.99; P = 0.02), greater frailty on admission (OR, 0.72; 95% CI, 0.55 to 0.94; P = 0.02), and prolonged intubation and MV as the indication for tracheostomy (OR, 0.44; 95% CI, 0.22 to 0.88; P = 0.02) were significantly associated with lower odds of decannulation (ESM eTable 3).

Discussion

Elderly patients who have a tracheostomy for prolonged MV have long ICU and hospital stays, a high mortality rate, and are unlikely to be discharged directly home. At the time of hospital discharge, survivors had a poor quality of life, with most receiving tube feeding, being unable to ambulate, and being severely frail.

Our results are comparable to those of other studies. Hospital mortality rates were comparable to those in more recent studies of tracheostomy both in the elderly and the overall population.5,16,19 In previous studies of patients with tracheostomy, short-term mortality was independently associated with age, comorbidities,12,14,16,19,20 and frailty.21,22 Elderly hospital survivors remain at risk for subsequent deterioration and mortality. Outcome studies of critically ill elderly patients describe 12-month mortality rates of 40–44%,22,23 and this rate is even higher in those with tracheostomies, ranging from 50% to 94%.12,15,17,24 Risk factors for long-term mortality include comorbidities, failure of tracheostomy decannulation, and frailty.17

Despite high levels of baseline functional independence, hospital survivors had loss of mobility, verbal communication, and oral feeding ability. On hospital discharge, the majority of patients required supportive care and were discharged to other acute care or chronic care facilities; only 4% were discharged home. This deterioration and debilitation has been previously described. Outcome studies of critically ill elderly patients describe severe and persistent functional dependency in survivors, who continue to require acute care with readmissions and supportive care after discharge even if they were previously independent and living at home.12,20,23,24

Of the 80 patients who had a formal swallowing assessment, 84% had varying degrees of dysphagia. This is consistent with rates of swallowing disorders reported in tracheostomy patients (11–93%).25,26,27 A total of 117 patients (43%) required insertion of percutaneous feeding tubes (Table 4). Most survivors (64%) remained solely dependent on tube feeding (Fig. 1). Currently, there are minimal data on feeding outcomes in the few studies of tracheostomy in the elderly population, but available data show extremely low rates of oral feeding (3–4%)12 with 71% undergoing gastrostomy tube insertion.11 These data contrast starkly with a study of 126 younger tracheostomized patients, whereby on hospital discharge, 86% were receiving nutrition orally and 25% via PEG.27

The high incidence of dysphagia is likely associated with our older study cohort, as age is a risk factor for swallowing impairment,28 largely due to elderly patients having more comorbidities and diminished functional reserve, and a greater predisposition to deconditioning during critical illness. The tracheostomy tube may impair swallowing by disrupting the normal protective reflexes; and this may be more pronounced in the elderly. Even after decannulation, there may not be immediate reversal of swallowing impairment. Hence, if unrecognized, it may result in complications such as aspiration pneumonia and nutritional deficiencies.29 Given that swallowing impairment can potentially be reversed by decannulation and rehabilitation,25,27 standardized assessment is important.

According to the Clinical Frailty Scale, at baseline half of our patients were managing well or were vulnerable, while a third were frail. Prior to discharge, all survivors were frail, with 82% severely to very severely frail. It was not unexpected that patients who were vulnerable and pre-frail progressed along the clinical trajectory of critical illness to severe frailty after ICU admission, and this is unlikely to be a result of tracheostomy alone. In our study, frailty on admission was significantly associated with lower odds of decannulation, and although not statically significant, there was a trend toward increased hospital mortality. Frailty has been well described in the critical care literature as an independent risk factor for hospital and long-term mortality, disability, extubation failure with need for tracheostomy, and discharge to long-term care facilities.22,30,31,32 While age is a risk factor for frailty,33 frailty can exist across all age groups as a marker of reduced biologic reserve, and increased vulnerability to stressors.31,34,35 A study by Johnson et al.21 also found significantly higher baseline frailty among nonsurvivors in a population of patients who underwent tracheostomy.

Patients and family members often express concerns that the tracheostomy will be permanent, and this has implications on quality of life and discharge disposition. Rates of successful decannulation in survivors vary from 26% to 100% in the general population,36,37,38,39,40 and from 0% to 13% in elderly populations.11,12,15,17 Our cohort’s decannulation rate may be higher than 43%, given that patients who were discharged to other facilities may have been decannulated at a later time. Decannulation practices vary among clinical teams and hospitals22 and there is a lack of standardization and guidelines for decannulation. Nevertheless, appropriate evaluation for decannulation is an essential component of rehabilitation. Decannulation is associated with better functional outcomes, improved dysphagia, and more effective communication, which in turn has a positive effect on mood and quality of life.41,42,43,44 A multidisciplinary approach with patient evaluation by a team including physicians, physiotherapists, speech-language therapists, and nurses along with decannulation protocols and decisional flowcharts is important in ensuring safe and successful decannulations.45,46,47

The strengths of this study include a relatively large number of elderly patients from two centers. Besides mortality, our study provides data on patient-centered quality of life indicators such as eating, mobility, and speech. The assessment of frailty also adds to the growing evidence that it is an important prognostic marker in critically ill patients. Our study also has limitations. It was conducted at two sites in Toronto, Canada, which may limit the generalizability of our findings. This study was a retrospective chart review, and as such, certain data extracted from the clinical notes (e.g., CFS and withdrawal or withholding of life-sustaining therapies) may not be accurate. We also did not have data on short-term follow-up of patients transferred to other centers, or long-term outcome data on all patients. In addition, we did not have information on the perspectives of patients and their caregivers regarding the value they place on discharge disposition, eating, and speaking. Prospective studies are needed to elucidate the long-term outcomes of tracheostomy in this population, as well as the physical and emotional symptom burden following tracheostomy. Lastly, dysphagia assessments and decannulation processes were based on local hospital practices in part due to lack of guidelines on the subject. Further research in these areas will allow for standardization and less heterogeneity in global practices.

Conclusions

In this historical cohort study, elderly patients who were tracheostomized as part of their ICU care had complex clinical courses involving ICU readmissions and long hospital stays, high in-hospital mortality, and consequent profound physical morbidity with infrequent discharge home. The illness trajectory had a significant impact on functional independence, including the ability to eat, speak, and participate in self-care. These findings and outcomes should be taken into consideration by clinicians when discussing the option of tracheostomy with patients and their families.