Outcomes of percutaneous versus surgical tracheostomy in an Australian Quaternary Intensive Care Unit: An entropy-balanced retrospective study

Abstract

Background:

Studies comparing percutaneous tracheostomy (PT) and surgical tracheostomy (ST) complications in the critically ill patient population with high acuity, complexity, and severity of illness are sparse. This study evaluated the outcomes of elective PT versus ST in such patients managed at a quaternary referral center.

Aims:

The primary aim was to detect a difference in hospital mortality between the two techniques. The secondary aims were to compare Intensive Care Unit (ICU) mortality, complications (including stoma site, tracheostomy-related, and decannulation complications), ICU and hospital length of stay, and time to decannulation.

Methods:

This was a single-center retrospective observational study of ICU admission from August 2018 to August 2021. Patients were included if an elective tracheostomy was performed during their ICU admission. Patients with a pre-existing tracheostomy and those who underwent an obligatory tracheostomy requirement (e.g. total laryngectomy) were excluded. Cohorts were matched using Hainmueller’s entropy balancing. Binary data were evaluated using logistic regression and continuous data with ordinary least squares regression.

Results:

349 patients with a tracheostomy were managed in the ICU during the observation period. They were predominantly males (75% in PT; 67% in ST), with a mean age in the PT and ST group of (47; SD = 18) and (55; SD = 16), respectively. After exclusion, 135 patients remained, with 63 in the PT group and 72 in the ST group. Patients receiving ST were significantly older with a higher Body Mass Index (BMI) than the PT group. There were no significant differences in gender, Acute Physiological And Chronic Health Evaluation (APACHE) III, and the Australian and New Zealand Risk Of Death (ANZROD) between the two groups. There was no difference in hospital mortality between groups (OR 0.91, CI 0.26–3.18, p = 0.88). There were also no differences in ICU mortality, ICU and hospital length of stay, and time to decannulation. PT was associated with a greater likelihood of complications (OR 4.19; 95% CI 1.73–10.13; p < 0.01). PT was associated with a greater risk of complications in those who had this performed early (<10 days of intubation) as well as late (>10 days of intubation).

Conclusions:

Percutaneous tracheostomy was associated with higher complications compared to surgical tracheostomy. They were related to tracheostomy cuff deflation, stomal site bleeding and infection, sputum plugging, and accidental and failed decannulation. These findings have identified opportunities to improve patient outcomes.

Background

Tracheostomy is a common procedure in intensive care, typically for patients requiring extended duration of mechanical ventilation in the setting of prolonged respiratory failure, with hundreds of thousands performed worldwide each year. ¹ The decision to offer tracheostomy is multifaceted, encompassing considerations of indication, suitability, safety, informed consent, technical competence, and multidisciplinary aftercare. ²

The choice between surgical tracheostomy (ST) versus percutaneous-dilatational technique adds further complexity, influenced by dynamic factors including neck anatomy and intercurrent pathologies such as obesity, subcutaneous emphysema, and coagulopathy. ³ The evidence guiding this choice continues to evolve, and the clear superiority of one technique over the other has yet to be definitively established.

In previous meta-analyses ⁴ and Cochrane reviews, ⁵ mortality and serious adverse events were comparably rare between approaches. Percutaneous tracheostomy (PT) is commonly associated with a shorter length of stay with fewer periprocedural complications (wound infections, delayed wound healing, tracheal stenosis, fistula, unfavorable scarring). ⁶ It is worth noting that the mortality benefit identified by the meta-analyses ⁴ and systematic review by Delaney et al. was based on sub-group analysis and possibly contributed by a signal of reduced wound infection and clinically significant bleeding. This doesn’t generally mean that the results of these studies can be generalized across different ICUs with varying case-mix. Contradicting outcomes from COVID-19 datasets demonstrate no difference in mortality, complication, or decannulation rates between the techniques. ⁷ It’s also important to remember that, in the early stages of the pandemic, clinical practice changed, traditional guidelines for PT and ST were not always followed, and there was evidence of a higher rate of tracheostomy in patients with COVID-19 than in those without.

Such conflicting results continue to emerge, including recent large retrospective studies reporting longer ICU length of stay ⁸ and increased mortality with PT. ⁹ As such, it remains relevant to interrogate outcomes for both techniques and identify quality and safety signals to guide contemporary practice. An evolving area with currently limited evaluation is the impact of differences in disease severity, comorbidities, and admission diagnosis in influencing the decision for percutaneous versus a surgical approach. In this context, we conducted an entropy balance retrospective observational study to compare outcomes following elective percutaneous and surgical tracheostomy performed during intensive care unit admission.

Methods

Aims

This study evaluated elective PT and surgical ST outcomes performed during ICU admission. The primary outcome was hospital mortality, and the secondary outcomes were ICU mortality, complication rates, hospital length of stay, and ICU length of stay. Complications were analyzed under three categories: (a) stomal site related: peri-stomal bleeding, granulation tissue dehiscence, and stomal wound infection; (b) tracheostomy related: sputum plugging and spontaneous cuff deflation, and (c) decannulation related: accidental and failed decannulation.

Study design and setting

This retrospective, single-center, observational study was undertaken at the Royal Adelaide Hospital in South Australia from August 2018 to August 2021. The Central Adelaide Local Health Network Human Research Ethics Committee (Reference Number 13343) approved this study with a waiver of consent. The Royal Adelaide Hospital is an 800-bed metropolitan quaternary referral hospital with a capacity of up to 60 ICU beds, accommodating approximately 3500 ICU admissions per year. Our ICU operates under a closed-collaborative model and provides full-service care across medical, surgical, cardiothoracic, trauma, spinal injury, and burns admissions.

All adult patients (age > 18 years) admitted to the ICU during the study period recorded as having a tracheostomy were identified using the Royal Adelaide Hospital Speech Pathology Tracheostomy Database. This database was cross-referenced against the Diagnostic Related Group case-mix coding database identifying patients with procedural codes for surgical or PT. Each entry was manually checked against medical records.

Patients were included if a tracheostomy was performed during their ICU admission. Patients were excluded if they were admitted with a pre-existing tracheostomy, underwent a surgical procedure with an obligate tracheostomy requirement (e.g. total laryngectomy) during their admission, required a tracheostomy as an emergent procedure, or underwent tracheostomy as a palliative care procedure. PTs were performed at the bedside in the ICU by intensive care physicians and senior advanced trainees competent with airway management (Supplemental information: ICU unit policy and tracheostomy procedural guidelines). All PT procedures were performed with a standardized single or serial dilatational technique with bronchoscopy guidance as recommended by the Australia and New Zealand Intensive Care Society (ANZICS). ¹⁰ The Blue Rhino^R G2-Multi Percutaneous Tracheostomy Introducer Set was used to form the stoma and subsequent serial dilatation to facilitate the tracheostomy insertion. STs were performed in theater by a specialty surgical team (ENT, Plastics, Oro maxillary Facial, or Acute Surgical & Trauma).

The study adhered to the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines. ¹¹

Data collection

ICU data were extracted a priori from a Sunrise Electronic Management Record and Patient Administration System (Sunrise EMR & PAS, Chicago, Il, US) and patients’ clinical paper records. Data collected included age, gender, BMI, APACHE III score, ANZROD mortality prediction outcomes, ¹² Cormack-Lehane laryngoscopic grade, ¹³ ICU admission category, indications for tracheostomy, duration of mechanical ventilation with a tracheal tube in situ, days from tracheostomy decision to insertion, insertion procedure documentation, time to decannulation, length of stay (ICU and hospital), mortality (ICU and hospital), documentation explaining the choice of technique, and complications (Definitions in Table 1) in early and late tracheostomy (including tracheostomy site bleeding, sputum plugging, formation and associated complications of granulation tissue formation, stomal wound infection, accidental decannulation, spontaneous cuff deflation, or failed decannulation. We dichotomized our complications into early (⩽10 days intubated prior) and late (>10 days intubated prior) to align ourselves and standardize our research findings with previously published local and international research.^14,15

Table 1.

Definitions of tracheostomy complication.

Complications	Description
Peri-stomal bleeding	Bleeding from the tracheal stoma created for insertion of tracheostomy
Granulation tissue dehiscence	Breakdown of granulation tissue surrounding previously formed stoma
Stomal breakdown	Stomal wound dehiscence caused by infection
Sputum plugging	Partial or complete obstruction of tracheostomy tube
Spontaneous tracheostomy cuff deflation	Tracheostomy tube balloon deflation within the trachea and inability to maintain cuff pressure
Accidental decannulation	Unplanned removal of tracheostomy tube
Failed decannulation	Decannulation causing respiratory distress requiring reinsertion of tracheostomy or reintubation

Statistical analysis

No formal power calculations were attempted for this study as a convenience sample was utilized. Hainmueller’s entropy balancing ¹⁶ procedure was used to construct a set of weights that was applied to each observation in the ST group so that the distributional balance of patient demographics (for age, gender, BMI, APACHE III, and admission type) between the two groups was achieved. Admission types were categorized as neurology, respiratory, cardiac, and trauma. We favor entropy balancing over the propensity score matching approach commonly used in many observational studies, as it does not discard samples with low matching scores, avoiding consequent information loss.

Once the distributions of patient demographic variables were matched, entropy balancing weights were applied to regression models. Ordinary least squares regression was used to estimate the effect of the tracheostomy technique on length of stay and time to decannulation, and logistic regression was used for calculating the impact on the odds of mortality (at ICU or hospital) and post-procedure complications. All regressions were adjusted for patient demographics. Baseline data are presented as mean ± standard deviation or n (%) as appropriate. Odds ratios with a 95% confidence interval were reported for logistic regressions, and regression coefficients were reported for the ordinary least square regressions. Both 95% confidence interval and p-values were computed based on robust standard errors estimated in both regressions. A p-value < 0.05 was statistically significant.

Results

Study cohort

Between August 2018 and August 2021, 349 tracheostomized patients were admitted to the ICU. Of these, 190 patients met the inclusion criteria, while 159 patients were excluded for emergency and elective tracheostomies performed in the operating theater. A further 55 of these patients were subsequently excluded due to missing BMI data, after which the final analyzed cohort of 135 patients comprised 63 PT patients and 72 ST patients (Figure 1).

Figure 1.

Study cohort flowchart depicting derivation of analyzed sample and overall PT & ST cases.

Elective ICU tracheostomies = admitted ICU patients receiving PT in the ICU or transferred to the operating theater for ST.

Emergency & Elective ICU tracheostomies = patients admitted to ICU with ST.

PT = percutaneous tracheostomy. ST = surgical tracheostomy.

Mode = Indicating the circumstance (elective or emergency) and location (ICU or OT).

Comparative causes of percutaneous tracheostomy complications.

Baseline characteristics

Table 2 depicts the baseline characteristics of the two groups of patients. Patients receiving ST were older (56 vs 47 years, p < 0.01) and had significantly higher BMI [30.2 (SD 11.64) vs 26.4 (SD 6.39), p = 0.02] than the PT group. The most frequent indication for tracheostomy in both groups was slow ventilatory wean (75% of PT, 48% of ST), followed by failed extubation (16% of PT, 22% of ST). There were no significant differences in gender, APACHE III score, and ANZROD risk of in-hospital death between the two groups. APACHE III scores were notably high in the PT group (60.90; SD = 26.71) and the ST group (59.04; SD = 21.78).

Table 2.

Baseline characteristics of tracheostomy patients.

	Percutaneous tracheostomy (n = 63)	Surgical tracheostomy (n = 72)	p-Value
Variables
Age, mean (SD)	47.37 (18.00)	55.75 (16.86)	<0.01
Body mass index, mean (SD)	26.37 (6.39)	30.17 (11.64)	0.02
APACHE III score, mean (SD)	60.90 (26.71)	59.04 (21.78)	0.66
ANZROD, mean (SD)	0.19 (0.20)	0.15 (0.17)	0.25
Gender			0.31
Female, n (%)	16 (25)	24 (33)	-
Male, n (%)	47 (75)	48 (67)	-
Tracheostomy indication			0.01
Slow ventilatory wean, n (%)	47 (75)	35 (48)	-
Failed extubation, n (%)	10 (16)	16 (22)	-
Difficult airway, n (%)	3 (4)	4 (6)	-
Upper airway obstruction, n (%)	1 (2)	9 (13)	-
Surgical indication, n (%)	2 (3)	8 (11)	-
Duration of intubation, days			0.48
⩽10, n (%)	37 (59)	38 (53)	-
>10, n (%)	26 (41)	34 (47)	-

Patients receiving early (⩽10 days) versus late tracheostomy (>10 days) were evenly distributed, both within and between groups. ICU admission categories are presented in Table 3. One patient diagnosed with COVID-19 underwent ST.

Table 3.

ICU admission category for included patients.

Admission category	Percutaneous tracheostomy (n = 63)	Surgical tracheostomy (n = 72)
Neurology	27 (42)	12 (17)
Cerebrovascular accident	19 (30)	7 (10)
Spinal infections	1 (1.5)	2 (3)
Others	7 (11)	3 (4)
Respiratory	10 (16)	20 (28)
Pneumonia	6 (10)	7 (10)
Upper airway obstruction	1 (1.5)	12 (17)
Malignancy	1 (1.5)	1 (1)
Asthma	1 (1.5)
Haemoptysis	1 (1.5)
Cardiac	5 (8)	3 (4)
Acute coronary syndrome	4 (6)	2 (3)
Others	1 (2)	1 (1)
Trauma	15 (24)	24 (33)
Traumatic brain injury	6 (9.5)	8 (11)
Spinal	1 (1.5)	1 (5)
Polytrauma	8 (13)	15 (21)
Others
Other	6 (9.5)	13 (18)

ICU: Intensive Care Unit.

All results are n (%) unless otherwise stated.

Outcomes in percutaneous tracheostomy versus surgical tracheostomy

There was no difference in hospital mortality between percutaneous and surgical tracheostomy (OR 0.91, CI 0.26–3.18, p = 0.88). The type of technique did not impact the ICU mortality (OR 0.67, CI 0.10–4.62, p = 0.69; Table 4). PT was associated with a greater likelihood of complications (OR 4.19; 95% CI 1.73–10.13; p < 0.01). In those who had early tracheostomy (within 10 days of intubation), complication rates were higher in those undergoing PT (OR 9.95, CI 2.50–39.59, p < 0.01). In those patients who had late tracheostomies (10 days after intubation), the complication rates were higher in the PT group as well, with a comparably lower odds ratio. No significant differences between groups were noted in the ICU length of stay, hospital length of stay, or time to decannulation.

Table 4.

Outcomes of surgical versus percutaneous tracheostomy.

Outcomes	Percutaneous tracheostomy (n = 63)	Surgical tracheostomy (n = 72)	If percutaneous tracheostomy used		p-Value
			Odds ratio	95% CI
Primary outcome
Mortality, hospital, a n (%)	5 (8)	10 (14)	0.91	0.26–3.18	0.88
Secondary outcomes
Mortality, ICU, a n (%)	2 (3)	5 (7)	0.67	0.10–4.62	0.69
Complications, total, a n (%)	25 (40)	16 (27)	4.19	1.74–10.13	<0.01
⩽10 days intubated prior, n (%)	16 (43)	8 (21)	9.95	2.50–39.59	<0.01
>10 days intubated prior, n (%)	9 (35)	8 (24)	4.24	1.02–17.66	0.05
			Regression coefficient	95% CI
Length of stay (days) b
ICU, mean (SD)	23 (15)	25 (19)	−2.54	−8.05 to 2.97	0.36
Hospital, mean (SD)	55 (36)	56 (31)	−2.51	−16.31 to 11.29	0.72
Time to de-cannulate (days), mean (SD)	23.48 (23.38)	23.64 (16.19)	−1.97	−10.87 to 6.92	0.66

ICU: Intensive Care Unit; APACHE III: Acute Physiological Assessment and Chronic Health Evaluation III; ANZROD: The Australian and New Zealand Risk of Death.

^aLogistic regression analysis. Odds ratios are reported.

^bOrdinary least squares regression analysis. Regression coefficients are reported.

Complications by treatment group are presented in Table 5. Spontaneous cuff deflation was the most frequent complication in both groups, observed at 24% and 11% in the PT and ST groups, respectively. Peri-stomal bleeding was noted to be the second most common complication, occurring evenly (8%) in both groups. Other complications observed solely within the PT group included granulation tissue dehiscence (1.5%), sputum plugging within tracheostomy (1.5%), and failed decannulation (3%). Accidental decannulation was solely observed within the ST group. Relatively similar rates of stomal wound infection were observed within both groups (ST = 1%, PT = 1.5%). Some patients reported multiple complications; out of 41 patients, there were 47 reported complications.

Table 5.

Tracheostomy complications with percutaneous versus surgical technique.

Complications	PT (n = 63)	ST (n = 72)
	25 (40)	16 (27)
Stoma site
Peri-stomal bleeding	5 (8)	6 (8)
Granulation tissue dehiscence	1 (1.5)	-
Stomal wound infection	1 (1.5)	1 (1)
Tracheostomy-related
Sputum plugging within tracheostomy	1 (1.5)	-
Spontaneous tracheostomy cuff deflation	15 (24)	8 (11)
Decannulation
Accidental decannulation	-	2 (3)
Failed decannulation	2 (3)	-

ICU: Intensive Care Unit.

All results are n (%) unless otherwise stated.

Discussion

Key findings

In this single-center retrospective observational study, we report no difference in ICU or hospital mortality between patients undergoing percutaneous and surgical tracheostomy to facilitate weaning from mechanical ventilation. The ICU and hospital length of stay and decannulation time were similar between groups. PT was associated with an approximate three-fold increased risk of complications.

Spontaneous cuff deflation was the most frequent complication with both techniques but occurred more frequently in the PT group, with almost a quarter of patients receiving a percutaneous tracheostomy experiencing a cuff leak (Table 5, Figure 2). This unique finding has never been reported in previous tracheostomy literature^4,6,7,19 outside of recent COVID-19 studies.

Figure 2.

The graph above demonstrates spontaneous tracheostomy cuff deflation as a common complication predominantly evident in the percutaneous group.

The exact cause for this phenomenon, contrary to complication reporting in international datasets, ¹ is unclear, with a multitude of possibilities. None are conclusive; thus, we cautiously speculate and consider several key possibilities. Firstly, acquired anatomical aberrations from tracheomegaly secondary to prolonged cuff overinflation resulting in cuff leaks have been previously reported amongst severe COVID-19 patients.^{17 –19} Secondly, an inappropriately sized tracheostomy tube causes intra-tracheal sizing mismatch, resulting in cycles of cuff inflation-deflation-reinflation that may predispose to cuff rupture or cuff hyperinflation, resulting in tracheomegaly. Thirdly, equipment malfunction involving the pilot balloon or tracheostomy cuff may occur due to manufacturing defects or peri-procedural damage, respectively. The final common step of tracheostomy introduction via a narrow orifice may cause micro-abrasion from tracheal ring contact, impacting cuff integrity and causing decay and weakness over time. ²⁰ The formation of a tracheal window with direct visualization of the cuff and tracheostomy entry into the trachea ²¹ mitigates such incidences with surgical tracheostomies.

Cuff leaks impact multi-faceted components in safe tracheostomy practices. These include patient recovery, medical and nursing care, and the hidden cost implications. At-risk patients are exposed to micro-aspirations and, consequently, ventilator-acquired pneumonia,^22,23 especially patients with bulbar dysfunction and poor cough. ²⁴ The benefits of positive pressure ventilation are unoptimized, thus predisposing the risk of hypoxemic respiratory failure that can rapidly spiral into a peri-arrest situation with an unsecured threatened airway. ²⁵ Preventing cuff-related complications should begin with adequate training of healthcare professionals performing tracheostomy. Governance to investigate the cause of the leak is vital for systems improvement. Other preventive measures include vigilant pressure monitoring, correct tracheostomy sizes, and tailoring cuff inflation frequency and maintenance.

Relationship to previous studies

Our study expands upon the body of data derived from observational studies regarding tracheostomy in contemporary intensive care units in Australia. The outcomes of the extensive epidemiological observational study conducted in the State of Victoria by Casamento et al. ⁸ yielded contradictory findings, whereas the South Australian study conducted by Bihari et al. ²⁶ produced comparable findings.

Bihari et al. primarily investigated healthcare costs and outcomes for patients undergoing tracheostomy in a tertiary-level hospital in Adelaide, South Australia. Their findings on hospital mortality, ICU, and hospital length of stay from 386 patients who predominantly underwent PT (85%) are congruent with our results.

Contradicting this and our results are the outcomes published by Casamento et al., who performed a large retrospective observational epidemiological analysis involving 6010 patients in the State of Victoria over a decade from 2004 to 2014. While their incidences of tracheostomy halved in that period, the authors reported an overall increase in ICU and hospital length of stay and decreased ICU and hospital mortality between PT and ST groups. Similarities in patient groups (including medical, surgical, trauma, neurosurgical, spinal, and transplant patients) with higher median APACHE III scores and age groups in comparison to our study cohort were noted; APACHE III scores (55 [IQR = 43-74] vs 64 [IQR = 47-86] and age groups (55 [IQR = 36-67] vs 64 [IQR = 49-74] years).

In addition to their significantly larger sample size, these disparities could explain the differing results in the ICU and hospital length of stay. The decreased incidences from their trial conducted in 2004 to 2014 compared to our more recent observation could indicate a trend toward more nuanced patient selection with improved peri-tracheostomy care involving internal and outreach multidisciplinary tracheostomy teams.

In specific relevance to our predominant cohort of neurological tracheostomized patients, we note the congruency of this sub-population outcome with various international research data. Compared to the prospective randomized controlled trial (SETPOINT) by Bösel et al., ²⁷ we note their finding of no difference in the length of ICU stay parallel to our outcomes. On a more recent and larger scale, Premraj et al. ²⁸ metanalyses comprising 13 RCTs with a total of 17,000 patients investigating tracheostomy timing in critically ill patients with stroke were in accordance with our findings of no difference in hospital, ICU length of stay, and mortality outcomes.

Study implication

In our quaternary-level mixed-ICU cohort of critically ill patients (overall mean APACHE III = 60), the percutaneous and surgical tracheostomy groups had comparable mortality rates, ICU and hospital length of stay, and time to decannulation. The increased incidence of percutaneous tracheostomy complications predominantly attributed to spontaneous cuff deflation highlights the importance of vigilant cuff management and monitoring. Adequate training, ²⁹ regular cuff pressure assessments, proper tracheostomy selection, and individualized patient care are crucial for reducing the risk of cuff-related complications. By implementing preventive strategies and optimizing cuff management practices, including continuous cuff pressure monitoring, healthcare providers can minimize cuff-related complications and improve patient outcomes during percutaneous tracheostomy procedures.

Strength and limitations

Our study has several strengths. Under real-world conditions, we analyzed the outcomes of critically ill patients undergoing tracheostomy in a quaternary-level mixed medical-surgical ICU. To minimize the effect of confounders, we utilized a novel and robust matching technique in our analysis: entropy balancing. Entropy balancing is model-free and does not rely on first-stage fits to obtain score weights. Instead, it directly adjusts the weights to the known sample to achieve a high covariate balance. ¹⁴ Further, our choice of covariates represented demographic properties, prognostic indicators, and ICU admission indications affecting tracheostomy choices.

Our study also has several significant limitations. The first is its single-center, retrospective design, with possible selection bias, incorrect data extraction, and misclassifications. This design limits the generalizability of our results, and their external validity and outcomes reflect association rather than causation. However, the researchers minimized potential erroneous data entry by manually cross-checking multiple databases and close interrogation of case records. Secondly, the entropy matching did not account for provider characteristics; hence, unknown confounding factors are possible. Thirdly, individual providers’ bias regarding the tracheostomy technique was not explored. Lastly, we could not evaluate patient-related outcome measures such as time to oral intake and aspects of communication.

Conclusions

In our center, percutaneous tracheostomy was associated with a three-fold higher complication rate than surgical tracheostomy. Hospital and ICU mortality and length of stay were similar between techniques.