ARTICLE IN PRESS

Journal of Cardiothoracic and Vascular Anesthesia 000 (2022) 18

Contents lists available at ScienceDirect

Journal of Cardiothoracic and Vascular Anesthesia

journal homepage: www.jcvaonline.com

Original Article
Impact of Early Tracheostomy on Clinical Outcomes in
Trauma Patients Admitted the to Intensive Care Unit:
A Retrospective Causal Analysis
Chi Peng, MPH*,3, Liwei Peng, MDy,3, Fan Yang, MDz,3,
Hang Yu, MDx,3, Peng Wang, MD y
, Chao Cheng, MS,2y,
y y,1
Wei Zuo, MS , Weixin Li, MD , Zhichao Jin, PhD*
*
Department of Health Statistics, Naval Medical University, Shanghai, China
y
Department of Neurosurgery, Tangdu Hospital, Air Force Medical University, Xi’an, China
z
Institute of Pathology and Southwest Cancer Center, Southwest Hospital, Third Military Medical University
(Army Medical University) and Key Laboratory of Tumor Immunopathology, Ministry of Education of China,
Chongqing, China
x
Emergency Department, Changhai Hospital, Naval Medical University, Shanghai, China

Objectives: To assess the indications, timing, and clinical outcomes that result from the early tracheostomy (ET) administration, by causal infer-
ence models.
Design: A retrospective observational study.
Setting: Multiinstitutional intensive care unit in the United States
Participants: The study comprised 626 trauma patients.
Interventions: An ET versus late tracheostomy (LT).
Measurements and Main Results: Trauma patients with tracheostomy were identified from 2 public databases named Medical Information Mart
for the Intensive Care-IV and eICU Collaborative Research Database. Tracheostomy was defined as early (
7 days) or late (>7 days) from inten-
sive care unit admission. A marginal structural Cox model (MSCM) with inverse probability weighting was employed. For comparison, the
authors also used time-dependent propensity-score matching (PSM) to account for differences in the probability of receiving an ET or LT. A total
of 626 eligible patients were enrolled in the study, of whom 321 (51%) received a ET. The MSCM and time-dependent PSM indicated that the
ET group was associated with reduced ventilation-associated pneumonia (VAP) and a shorter mechanical ventilation (MV) duration than the LT
group. Yet, mortality did not show any difference between the two groups.
Conclusions: The authors’ study observed that ET was not associated with reduced mortality in trauma patients, but it was associated with
reduced VAP risk and MV duration. The results warrant further validation in randomized controlled trials.
Ó 2022 Elsevier Inc. All rights reserved.

Key Words: causal analysis; clinical outcome; intensive care unit; trauma; tracheostomy timing

This study was supported by The San Hang Program of the Naval Medical TRACHEOSTOMY IS A COMMONLY performed surgi-
University.
1 cal procedure in the intensive care unit (ICU) to secure the air-
Address correspondence to Weixin Li, MD, Department of Neurosurgery,
Tangdu Hospital, Air Force Medical University, No.1 Xinsi Road, Xi’an, way in patients with multiple traumas, especially in patients
China, 710038.2Zhichao Jin, PhD, Department of Health Statistics, Naval who need prolonged mechanical ventilation (MV).1,2,3 Two
Medical University, No. 800 Xiangyin Road, Shanghai, China, 200433. large, multicenter, randomized trials mainly included mixed
E-mail addresses: tangdunaowai@163.com (W. Li), jinzhichao@smmu. critically ill patients who needed prolonged MV.4,5 These 2
edu.cn (Z. Jin).
3 studies included all patients who were admitted to the ICU—
Chi Peng, Fan Yang, Liwei Peng, and Hang Yu contributed equally to this
work and are co-first authors. not limited to trauma patients—and, consequently, there may

https://doi.org/10.1053/j.jvca.2022.12.022
1053-0770/Ó 2022 Elsevier Inc. All rights reserved.
 ARTICLE IN PRESS
2 C. Peng et al. / Journal of Cardiothoracic and Vascular Anesthesia 00 (2022) 18

have been some heterogeneity in the study population. In par- laboratory parameters, injury details, and multiple scoring sys-
ticular, tracheostomy timing is an important issue with trauma tems on the first day of ICU admission. Time-varying variables,
patients, who are a unique subpopulation of critically ill including PaO2, PaCO2, and GCS, were measured during the
patients.6,7 Previous studies assessing early tracheostomy (ET) entire ICU stay, with multiple measurements, and the lowest
versus late tracheostomy (LT) among trauma patients have daily values of PaO2 and GCS were included in the analysis,
demonstrated controversial results on clinical outcomes.8-12 accompanied by the highest daily value of PaCO2.
To the authors’ knowledge, these publications were limited
by the heterogeneity of included patients, the variability of ET Primary and Secondary Outcomes
definition, single-hospital sampling, insufficient statistical
power, and a lack of control for time-varying confounders. The primary outcome was all-cause mortality. The second-
Furthermore, all previous retrospective studies failed to incor- ary outcomes included surgical site infection, respiratory com-
porate the tracheostomy as the time-dependent variable, plications (ie, ventilation-associated pneumonia [VAP]), acute
unavoidably leading to survivor treatment bias.13 In fact, as a respiratory failure, acute respiratory distress syndrome), and
time-dependent variable, the treatment with tracheostomy other complications (defined as acute kidney injury, deep
depends on PaO2, PaCO2, and Glasgow Coma Scale venous thrombosis, pulmonary embolism, severe sepsis and
(GCS).14,15 Consequently, PaO2, PaCO2, and GCS are all shock, ICU length of stay and hospital length of stay, MV
causes and results of tracheostomy. However, it may not be duration, and sedation duration).
ethical to perform a randomized controlled trial to explore the
causal relationship; thus, the authors performed an analysis of Statistical Analysis
a hypothetical trial using observational longitudinal data to
achieve causal inference. The continuous variables were presented as median with
IQR, and the categorical variables as total number and percent-
Materials and Methods age. The proportions were compared using chi-square test or
Fisher exact test, whereas the continuous variables were com-
Data Source and Ethics Approval pared using the t test (normally distributed) or Wilcoxon rank
sum test (not normally distributed), as appropriate.
The study authors conducted a retrospective, longitudinal, In line with as-treated analyses, the marginal structural Cox
retrospective study of trauma patients from 2 sizeable critical model (MSCM) facilitates causal inference of the association
care databases, the Medical Information Mart for Intensive between a time-updated exposure (tracheostomy) and the out-
Care-IV version 1.016 and the eICU Collaborative Research come by accounting for both confounding and selection bias,
Database version 1.2.17 Because the study was an analysis of such as informative right-censoring.21 Details of MSCM can
the anonymized publicly available databases with preexisting be seen in the electronic supplementary materials (Supplemen-
Institutional Review Board approval, informed consent also tal Figure S1).
was waived. The study was reported in accordance with To further assess the association between ET and in-hospital
REporting of studies Conducted using Observational Rou- outcomes, time-dependent propensity score matching (PSM) was
tinely collected health Data statement.18 also adopted.22,23 The propensity score reflects the estimated
probability of initiation of tracheostomy in patients with trauma.
Study Population First, the propensity score was calculated based on a Cox propor-
tional hazards model to balance both baseline and time-varying
The inclusion criteria were as follows: patients with (1) covariates across patients treated with tracheostomy or not within
trauma and/or (2) had undergone tracheostomy. Patients with 7 days. The outcome variable in this model was time to tracheos-
chronic obstructive pulmonary disease, burns, late effects of tomy. Next, the authors applied a 1:1 risk-set matching on the
trauma, and those who stayed in the ICU for <48 hours were propensity score using the nearest-neighbor matching algorithm,
excluded from the study. Moreover, for patients with ICU with a maximum caliber of 0.01 of the propensity score.24 In this
readmissions, only data from the first ICU admission of the way, every patient with tracheostomy at any given time point (0-
first hospitalization were included in the analysis. The patients 7 days after ICU admission) was separately and sequentially
who had a tracheostomy performed <7 days from their ICU matched to a patient with a similar propensity score who did not
admission were categorized as the ET group, with the remain- receive a tracheostomy within the time point.
ing patients making up the LT group.19,20
Sensitivity Analysis
Data Collection
In the sensitivity analysis, to increase the robustness of the
In this study, detailed demographic dataincluded age, sex, authors’ study, they performed prespecified subgroup analyses
race, body mass index, admission type, and smoking history. stratified by age, sex, body mass index, oxygenation index,
Coexisting disorders were also collected. The Charlson Comor- injury location, CCI, GCS, and ISS to explore in-hospital out-
bidity Index (CCI) was calculated from its component variables. comes among study groups with different characteristics.
In addition, the authors extracted data containing vital signs, Moreover, the patterns assumed missing to be completely at
 ARTICLE IN PRESS
C. Peng et al. / Journal of Cardiothoracic and Vascular Anesthesia 00 (2022) 18 3

random, so a multiple imputation approach was used to iterate significance was considered to be at 2-sided p < 0.05. All anal-
the original data (Supplemental Table S1).25 Ultimately, yses were performed with R version 4.0.2 (R Foundation for
unmeasured confounding may have biased the estimates from Statistical Computing).
this observation study; thus, the E-value also was computed to
further evaluate the robustness of the findings.26,27 The E-
Results
value indicated how strongly an unmeasured confounder
would have to be associated with both the use of ET and the Study Cohort
outcomes of interest to reduce the observed effect to the null,
conditional on the measured covariates. Finally, because the Among 16,360 trauma patients, a total of 865 subjects
definition of ET varied across articles, the authors performed a underwent tracheostomy. In accordance with the exclusion cri-
sensitivity analysis that included 5 days as a cut-off point to teria, 626 patients were included in the study cohort. Of this
analyze whether there were differences in outcomes between cohort, 321 had an ET, whereas the remaining 305 patients
ET and LT. The reason why the authors chose 5 days as the had LT treatment (Fig 1). The baseline characteristics of the
time cut-off point refers to previous literature.9 Statistical ET and LT groups are described in Table 1 and Supplemental

Fig 1. Flowchart of eligible participants.

 ARTICLE IN PRESS
4 C. Peng et al. / Journal of Cardiothoracic and Vascular Anesthesia 00 (2022) 18

Table 1
Baseline Characteristic of the Included Patients Stratified by Tracheostomy Timing During ICU Stay

Variables All (n = 626) ET (n = 321) LT (n = 305) p Value

Demographics
Age, median (IQR), y 57.00 (39.00-71.00) 57.00 (36.00-71.00) 58.00 (40.00-71.00) 0.343
Male sex, n (%) 432 (69.01) 217 (67.60) 215 (70.49) 0.487
Smoking history, n (%) 218 (34.82) 127 (39.56) 91 (29.84) 0.014
CCI, median (IQR) 3.00 (1.00-6.00) 3.00 (1.00-6.00) 3.00 (1.00-5.00) 0.552
Vital signs on the day of ICU admission, median (IQR)
Temperature, ⁰C 37.10 (36.60-37.50) 37.10 (36.70-37.50) 37.00 (36.50-37.50) 0.101
Mean arterial pressure, mmHg 80.00 (73.00-89.00) 78.00 (72.00-86.00) 81.00 (74.00-91.00) 0.006
Heart rate, beats/min 88.00 (77.00-102.00) 86.00 (75.00-99.00) 90.00 (79.00-104.00) 0.008
Respiratory rate, breaths/min 19.00 (17.00-22.00) 18.00 (17.00-21.00) 20.00 (17.00-23.00) < 0.001
Laboratory findings on the day of ICU admission, median (IQR)
PaO2, mmHg 86.00 (61.00-127.00) 87.00 (63.00-127.00) 81.00 (58.00-127.00) 0.242
SaO2, % 98.00 (96.00-99.00) 98.00 (96.00-99.00) 98.00 (96.00-99.00) 0.615
PaO2/FIO2 202.50 (118.00-318.00) 199.50 (117.00-313.00) 218.00 (123.75-335.50) 0.582
PaCO2, mmHg 46.00 (41.00-53.00) 46.00 (40.00-54.00) 46.00 (42.00-52.50) 0.458
HCO3, mmol/L 21.20 (19.00-23.28) 21.20 (19.50-23.00) 21.20 (18.90-23.30) 0.981
PH 7.40 (7.30-7.40) 7.40 (7.30, 7.40) 7.40 (7.30, 7.40) 0.038
BE, mmol/L -1.48 (-4.34 to 0.50) -1.00 (-3.67 to 1.00) -2.20 (-5.05 to 0.00) < 0.001
TCO2, mmol/L 24.00 (21.00-27.00) 25.00 (22.00-27.00) 24.00 (21.00-26.00) < 0.001
Type of injury, n (%) 0.225
Head 140 (35.81) 60 (31.91) 80 (39.41)
Chest 47 (12.02) 26 (13.83) 21 (10.34)
Spinal cord 146 (37.34) 77 (40.96) 69 (33.99)
Abdomen 58 (14.83) 25 (13.30) 33 (16.26)
Mechanism of injury, n (%) 0.052
Unintentional falls 66 (17.69) 38 (17.51) 28 (17.95)
Motor vehicle crashes 291 (78.02) 165 (76.04) 126 (80.77)
Firearm injuries 16 (4.29) 14 (6.45) 2 (1.28)
Scoring systems
SOFA 6.00 (4.00-9.00) 6.00 (4.00-8.00) 7.00 (5.00-10.00) < 0.001
APSIII 62.00 (44.00-79.00) 53.00 (38.00-76.00) 67.00 (53.00-82.00) < 0.001
GCS < 0.001

8 291 (49.49) 154 (50.83) 137 (48.07)
9-12 151 (25.68) 77 (25.41) 74 (25.96)
13-15 146 (24.83) 72 (23.76) 74 (25.96)
ISS 0.196
15 433.00 (69.17) 230.00 (71.65) 203.00 (66.56)
<15 193.00 (30.83) 91.00 (28.35) 102.00 (33.44)

Abbreviations: APSIII, acute physiology score III; BE, base excess; BMI, body mass index; CCI, Charlson Comorbidity Index; ET, early tracheostomy; FIO2,
fraction of inspired oxygen; PH, pondus hydrogenii; BE, base excess; TCO2, total CO2; GCS, Glasgow Coma Scale; ISS, Injury Severity Score; ICU, intensive care
unit; LT, late tracheostomy; SaO2, arterial oxygen saturation; SOFA, Sepsis-Related Organ Failure Assessment.

Table S2. Patients in the ET group were more likely to have a p = 0.020), GCS (OR for each 1-point increase 0.77;
smoking history (39.56% v 29.84%, p = 0.014) than the LT p = 0.045) were important predictors of ET. Other indepen-
group. On the first day of ICU admission, the ET group dent predictors included older age, higher CCI, and the
showed lower mean arterial pressure (78.00 v 81.00 mmHg; Sepsis-Related Organ Failure Assessment score. Details
p = 0.006), lower heart rate (86.00 v 90.00 beats/min; can be seen in Supplemental Table S3.
p = 0.008), lower respiratory rate (18.00 v 20.00 times/min; p The distributions of inverse probability weighting in the
< 0.001), and higher pH (7.40 v 7.40; p = 0.038) than the LT ET and LT groups are shown in Figure 2. By adjusting for
group. In general, patients in the ET group were less critically both baseline and time-varying covariates, the associations
ill than in the LT group (Sepsis-Related Organ Failure Assess- between ET and clinical outcomes were investigated in the
ment 6.00 v 7.00; p < 0.001) (acute physiology score III 53.00 MSCM. MSCM results showed that an ET was not signifi-
v 67.00; p < 0.001) than the LT group. cantly associated with reduced mortality (30-day mortality:
hazard ratio [HR] 1.21; p = 0.563; 90-day mortality: HR
Marginal Structural Cox Model 0.78; p = 0.355; 365-day mortality: HR 0.77; p = 0.316) in
this study. Similarly, the study did not identify any significant
The first step in fitting MSCM demonstrated that PaO2 beneficial effects of an ET on 30-day mortality in different
(odds ratio [OR] for each 10-mmHg decrease 0.88; subgroups (Supplemental Figure S2). After adjustment for
p = 0.044), PaCO2 (OR for each 10 mmHg increase 1.10; confounders by using MSCM, it was found that VAP and MV
 ARTICLE IN PRESS
C. Peng et al. / Journal of Cardiothoracic and Vascular Anesthesia 00 (2022) 18 5

Table 2
Impact of the ET on ICU Outcomes: A MSCM With Ponderation on the Stabi-
lized Weights Using an IPTW Estimator

Outcomes Model 1 Model 2

Primary outcomes HR (95% CI) p Value HR (95% CI) p Value

All-cause 1.82 (0.93-3.55) 0.080 1.21 (0.64-2.29) 0.563

mortality (30
days)
All-cause 0.95 (0.55-1.65) 0.854 0.78 (0.46-1.32) 0.355
mortality (90
days)
All-cause 0.94 (0.97-1.01) 0.823 0.77 (0.46-1.28) 0.316
mortality
(365 days)
Secondary
outcomes
Surgical site 1.19 (0.56-2.52) 0.649 0.93 (0.42-2.07) 0.869
infection
Fig 2. Distribution of inverse probability weight in early tracheostomy and Respiratory
late tracheostomy groups. complications
VAP 0.48 (0.32-0.72) < 0.001 0.67 (0.46-0.97) 0.035
Acute 0.62 (0.44-0.86) 0.005 0.83 (0.66-1.04) 0.109
respiratory
duration were significantly lower in the ET than the LT group failure
(HR 0.67; p = 0.035 and HR 0.39; p < 0.001, respectively) ARDS 1.25 (0.53-2.91) 0.613 1.15 (0.44-3.03) 0.770
Other
(Table 2).
complications
The E-values for the observed estimates of the association AKI 0.88 (0.58-1.33) 0.544 0.85 (0.62-1.17) 0.330
among ET, VAP, and MV duration were 2.35 and 4.57, respec- DVT 0.61 (0.36-1.04) 0.070 0.76 (0.43-1.34) 0.335
tively, and the lower limits of 95% CI closest to the null point PE 0.68 (0.33-1.40) 0.292 0.94 (0.43-2.01) 0.868
were 1.21 and 3.33, respectively (Supplemental Table S4), Severe sepsis 0.86 (0.52-1.41) 0.544 0.67 (0.44-1.02) 0.061
Shock 0.85 (0.41-1.80) 0.682 0.78 (0.34-1.78) 0.559
Because ET was defined differently in different studies, the
ICU days, median 1.42 (1.00-2.00) 0.046 1.26 (0.97-1.63) 0.078
authors performed a sensitivity analysis. Details can be seen in (IQR) (22.00)
Supplemental Table S5. This result was in congruence to the LOS, median 1.21 (0.86-1.71) 0.284 1.07 (0.83-1.39) 0.588
main findings in this passage, which may increase the robust- (IQR), d (14.00)
ness of the findings in the authors’ paper. MV duration, 0.14 (0.09-0.21) < 0.001 0.39 (0.29-0.51) < 0.001
median (IQR)
(205.76)
Sedation duration, 0.52 (0.35-0.77) 0.001 0.83 (0.65-1.05) 0.123
median (IQR)
Time-Dependent PSM Analysis (76.5)

NOTE. Model 1 (conventional Cox model): adjusted for baseline covariates,

A total of 474 patients were included in the propensity- Model 2 (Marginal structural Cox model): adjusted for baseline and time-
matched cohort. The distribution of baseline and time-depen- varying covariates.
dent covariates in the matched cohort was well-balanced Abbreviations: ET, early tracheostomy; ICU, intensive care unit; MSCM,
between these 2 groups (Supplemental Figure S3). Tracheos- marginal structure Cox model; IPTW, inverse probability of treatment
weighting; HR, hazard ratio; VAP, ventilation-associated pneumonia; ARDS,
tomy within 7 days was not associated with reduced risk of
acute respiratory distress syndrome; AKI, acute kidney injury; DVT, deep
death (30-day mortality: HR 1.39; p = 0.354; 90-day mortality: venous thrombosis; PE, pulmonary embolism; LOS, length of stay; MV,
HR 0.82; p = 0.500; 365-day mortality: HR 0.86; p = 0.590) mechanical ventilation.
(Table 3). Notably, the subgroup analysis showed that there
were no significant interactions for the prespecified subgroup
categories, except for obesity (HR 0.55; p for interac- Discussion
tion = 0.020) (Supplemental Figure S4). Moreover, the results
of time-dependent PSM also demonstrated that the ET was To the best of the authors’ knowledge, this was the first
associated with reduced risk of VAP (HR 0.58; p = 0.015) and large-sample, longitudinal study indicating different outcomes
acute respiratory failure (HR 0.64; p = 0.016). In addition, in based on the timing of ET among trauma patients. The Euro-
this matched cohort, MV duration was shorter among the ET pean Society of Critical Care Medicine consensus did not rec-
group than among the LT group (HR 0.13; p < 0.001). Simi- ommend the optimal timing for tracheostomy in patients with
larly, the study identified a significant beneficial effect of tra- acute brain injury.14 French recommendations for the manage-
cheostomy treatment on sedation duration (HR 0.43; p < ment of patients with spinal cord injury (SCI) or at risk of SCI
0.001). recommend that ET could be performed among trauma
 ARTICLE IN PRESS
6 C. Peng et al. / Journal of Cardiothoracic and Vascular Anesthesia 00 (2022) 18

Table 3
Outcomes for Trauma Patients With an ET vs LT in the Overall and Time-Dependent Cohorts During ICU Stay

Outcomes Before Time-Dependent PSM After Time-Dependent PSM

Primary outcomes HR 95% CI P Value HR 95% CI p Value

All-cause mortality (30 days) 1.49 (0.82-2.77) 0.192 1.39 (0.69-2.80) 0.354
All-cause mortality (90 days) 0.86 (0.53-1.41) 0.547 0.82 (0.47-1.45) 0.500
All-cause mortality (365 days) 0.87 (0.54-1.39) 0.548 0.86 (0.50-1.49) 0.590
Secondary outcomes
Surgical site infection 1.08 (0.53-2.24) 0.834 1.11 (0.45-2.83) 0.825
Acute respiratory complications
VAP 0.54 (0.37-0.77) 0.001 0.58 (0.37-0.90) 0.015
Acute respiratory failure 0.59 (0.43-0.82) 0.001 0.64 (0.45-0.92) 0.016
ARDS 0.95 (0.41-2.19) 0.900 0.90 (0.36-2.26) 0.824
Other complications
AKI 0.73 (0.52-1.03) 0.071 0.76 (0.51-1.13) 0.169
DVT 0.62 (0.37-1.03) 0.064 0.69 (0.38-1.24) 0.211
PE 0.62 (0.30-1.24) 0.176 0.65 (0.28-1.45) 0.296
Severe sepsis 0.70 (0.46-1.07) 0.096 0.70 (0.43-1.13) 0.146
Shock 0.62 (0.30-1.24) 0.176 0.57 (0.25-1.24) 0.158
ICU days, median (IQR), d (22.00) 1.33 (0.97-1.82) 0.075 1.30 (0.91-1.86) 0.146
LOS, median (IQR), d (14.00) 1.17 (0.85-1.60) 0.327 1.12 (0.79-1.61) 0.521
MV duration, median (IQR) (205.76) 0.14 (0.10-0.20) < 0.001 0.13 (0.08-0.19) < 0.001
Sedation duration, median (IQR) (76.5) 0.42 (0.30-0.58) < 0.001 0.43 (0.30-0.62) < 0.001

Abbreviations: ARDS acute respiratory distress syndrome; AKI acute kidney injury; DVT deep venous thrombosis; ET early tracheostomy; HR hazard ratio; ICU
intensive care unit; LOS length of stay; LT late tracheostomy; MV mechanical ventilation; PE pulmonary embolism; PSM propensity-score matching; VAP
ventilation-associated pneumonia.

patients with upper-level SCI (C2-C5).6 Guidelines from a traumatic SCI demonstrated no differences in VAP between
French expert panel show that the question of the optimal tim- the ET and LT groups.37 The literature review regarding ET
ing of tracheostomy in ICU is hard to analyze, for most studies in different groups of trauma and ICU patients is shown in
comparing ET and LT include non-tracheotomized patients in Supplemental Table S6. Another important point to note was
the “late” group.28 However, based on contradictory, low-qual- that ET placement was not associated with improved short-
ity evidence, these suggestions cannot reach a consensus. The term or long-term all-cause mortality (eg, first 30 days,
authors’ study here aimed to provide supportive evidence that 60 days, and 365 days in the ICU). Timing of tracheostomy
ET is potentially beneficial for trauma patients in terms of placement on overall mortality is, in fact, a controversial topic.
reducing VAP and MV duration. The conclusion of a recent meta-analysis37 aligned with a ret-
PaO2, PaCO2, and pH were the most important predictors of an rospective study conducted by Devarajan J et al.,38 in which
ET, which was probably attributable to the fact that PaO2, PaCO2, their result of in-hospital mortality (ET: 21.1% v LT: 40.4%;
and GCS were highly predictive of prolonged MV support.29 In p = 0.002) favored the ET. A study regarding acute brain
fact, estimation of the ongoing need for MV may necessitate an injury suggested that the 28-day mortality rate was signifi-
ET.30 Furthermore, LT was more often performed in patients cantly lower with an ET.39 Nevertheless, great caution regard-
with lower GCS in case extubation became possible.31 ing the interpretation of beneficial outcomes of an ET is
The authors’ findings indicated that an ET may reduce the needed in these studies due to the small sample size and het-
incidence of VAP (33% relative reduction) and the duration of erogeneous population. The authors’ finding was consistent
MV (61% relative reduction). This outcome is clinically with prior reports in trauma patients. To be specific, a propen-
important because VAP is the most common nosocomial infec- sity-matched study conducted by McLaughlin C et al found
tion in the ICU setting, especially among trauma patients.32 that there were no significant differences in mortality (1.26
Additionally, a reduction of ICU days on MV also means cost [0.46-3.49]) between the ET and LT groups among children
savings. Previous studies had similar findings. Samiei Nasr D with traumatic brain injury.40 Similarly, a retrospective data-
et al. conducted a retrospective study and found that ET base analysis showed that no association was found between
decreased the duration of MV (ET: 8.11 § 4.9 days v LT: 16.3 the time to tracheostomy and in-hospital mortality among
§ 6.01 days, p < 0.05).33 At the same time, 2 meta-analyses patients with SCI patients.41 A recent study concerning chest
conducted by Chorath K et al.34 and Liu CC et al.,35 respec- trauma patients reported that mortality benefits were not corre-
tively, also indicated that ET was associated with reduced MV lated with tracheostomy timing.42
duration. In addition, shorter MV duration and lower probabil- Nevertheless, these investigations into the clinical outcomes of
ity of VAP also were found in studies regarding severe trau- an ET must be interpreted with caution, given the high heteroge-
matic brain injury population, conducted by de Franca SA et neity with respect to the inclusion and exclusion criteria, clinical
al.36 Nevertheless, a recent meta-analysis in patients with acute characteristics, surgical techniques, and the definition of timing.
 ARTICLE IN PRESS
C. Peng et al. / Journal of Cardiothoracic and Vascular Anesthesia 00 (2022) 18 7

The strength of this study was that it was drawn from a rela- careful methodology involving MSM and time-dependent
tively large-sample, population-based longitudinal cohort in PSM analyses to eliminate the potential impact of confound-
the United States and adjusted for time-dependent covariates, ers, the results of their study require further validation through
ensuring the robustness, reliability, and generality of the find- well-designed, prospective, multicenter randomized controlled
ings. Apart from this, to estimate the effect of tracheostomy trials.
treatment on in-hospital outcomes, the authors employed mul-
tivariate analyses, such as time-dependent PSM and MSCM.
Conflict of Interest
Notably, using tracheostomy as a time-varying variable in the
MSCM, Tsuchiya A et al.43 explored the effect of early trache-
None.
ostomy on clinical outcomes in adult patients with serve burns.
The MSCM model also has been employed successfully in
other situations to address time-dependent interventions.44 In Supplementary materials
addition, another sensitivity analysis was conducted. The defi-
nition set by physicians regarding the time to tracheostomy Supplementary material associated with this article can be
(ranging between 3 and 14 days) might be one reason that pre- found in the online version at doi:10.1053/j.jvca.2022.12.022.
vious studies drew disparate conclusions as to the effect of ET
on clinical outcomes.45 Therefore, referring to Hyde GA et
References
al.,9 a different ET (
5 days) was used, and similar results
were found.
1 Durbin CG Jr.. Tracheostomy: Why, when, and how? Respiratory Care
Moreover, by implementing multiple imputations to missing 2010;55:1056–68.
data, the analysis gave a relatively robust conclusion to reduce 2 Goettler CE, Fugo JR, Bard MR, et al. Predicting the need for early trache-
the estimation bias and improve validity. Unmeasured con- ostomy: A Multifactorial analysis of 992 intubated trauma patients. J
founders also were treated by E-Value. It is also important to Trauma 2006;60:991–6.
note that patients from the authors’ study belonged to a spe- 3 Combes A, Luyt CE, Nieszkowska A, Trouillet JL, Gibert C, Chastre J. Is
tracheostomy associated with better outcomes for patients requiring long-
cific trauma population, whereas most literature that was term mechanical ventilation? Crit Care Med 2007;35:802–7.
reviewed belonged to trauma services and several other health- 4 Terragni PP, Antonelli M, Fumagalli R, et al. Early vs late tracheotomy for
care areas, such as general ICU, neurosurgery, cardiology, and prevention of pneumonia in mechanically ventilated adult ICU patients: A
medicine emergency. When compelling randomized trials are randomized controlled trial. JAMA 2010;303:1483–9.
not yet forthcoming, it is incrementally valuable that credible 5 Young D, Harrison DA, Cuthbertson BH, Rowan K. Effect of early vs late
tracheostomy placement on survival in patients receiving mechanical ven-
evidence for the effect of the tracheostomy in trauma patients tilation: The TracMan randomized trial. JAMA 2013;309:2121–9.
was supplied in this study. 6 Roquilly A, Vigue B, Boutonnet M, et al. French recommendations for the
This study had several limitations, consistent with those management of patients with spinal cord injury or at risk of spinal cord
inherent to many large administrative database studies. First, injury. Anaesth Crit Care Pain Med 2020;39:279–89.
7 Carney N, Totten AM, O’Reilly C, et al. Guidelines for the management of
based on electronic records of routine clinical practice, missing
severe traumatic brain injuryfourth edition Neurosurgery 2017;80:6–15.
data and outliers were common. To retain statistical power, the 8 Kang BH, Cho J, Lee JC, Jung K. Early versus late tracheostomy in trauma
authors used multiple imputations to reduce the risk of bias patients: A propensity-matched cohort study of 5 years’ data at a single
from missing data. Second, as a retrospective study, unmea- institution in Korea. World J Surg 2018;42:1742–7.
sured confounding was inevitable; thus, to quantify the poten- 9 Hyde GA, Savage SA, Zarzaur BL, et al. Early tracheostomy in trauma
tial implications of it, E-Value sensitivity analysis was used in patients saves time and money. Injury 2015;46:110–4.
10 Alali AS, Scales DC, Fowler RA, et al. Tracheostomy timing in traumatic
this study. The authors found that an unmeasured confounder brain injury: A propensity-matched cohort study. J Trauama Acute Care
was unlikely to explain the entirety of the treatment effect. In Surg 2014;76:70–6;discussion 6-8.
addition, both time-dependent PSM and MSCM were 11 Keenan JE, Gulack BC, Nussbaum DP, et al. Optimal timing of tracheos-
employed; however, residual confounding cannot be excluded tomy after trauma without associated head injury. J Surg Res
fully. Third, instead of outcome adjudication, the authors’ out- 2015;198:475–81.
12 Barquist ES, Amortegui J, Hallal A, et al. Tracheostomy in ventilator
comes were defined by the International Classification of Dis- dependent trauma patients: A prospective, randomized intention-to-treat
eases, Ninth Revision (ICD-9) and ICD-10 diagnosis codes. study. J Trauma 2006;60:91–7.
Incorrect codes inevitably exist, so the codes used by the study 13 Scales DC, Thiruchelvam D, Kiss A, Redelmeier DA. The effect of trache-
authors were all verified by previous articles.46,47 Fourth, lim- ostomy timing during critical illness on long-term survival. Crit Care Med
ited by ICD codes, the authors could identify all complications 2008;36:2547–57.
14 Robba C, Poole D, McNett M, et al. Mechanical ventilation in patients
or tell whether the complications were due to the disease itself with acute brain injury: Recommendations of the European Society of
or the ET. Intensive Care Medicine consensus. Intensive Care Med 2020;46:2397–
410.
Conclusions 15 Robba C, Galimberti S, Graziano F, et al. Tracheostomy practice and tim-
ing in traumatic brain-injured patients: A CENTER-TBI study. Intensive
Care Med 2020;46:983–94.
In this cohort study, the authors found that the ET was asso- 16 Goldberger AL, Amaral LA, Glass L, et al. PhysioBank, PhysioToolkit,
ciated with lower VAP risk and shorter MV duration, but not and PhysioNet: Components of a new research resource for complex physi-
with a reduction in mortality among trauma patients. Despite ologic signals. Circulation 2000;101:E215–20.
 ARTICLE IN PRESS
8 C. Peng et al. / Journal of Cardiothoracic and Vascular Anesthesia 00 (2022) 18

17 Pollard TJ, Johnson AEW, Raffa JD, Celi LA, Mark RG, Badawi O. The pathogens in a public tertiary teaching hospital of Eastern India. Indian J
eICU Collaborative Research Database, a freely available multi-center Crit Care Med 2015;19:14–20.
database for critical care research. Sci Data 2018;5:180178. 33 Samiei Nasr D, Khoundabi B, Monshizadeh Azar G, Malekmohammad M,
18 Benchimol EI, Smeeth L, Guttmann A, et al. The Reporting of studies Con- Jamaati H, Hashemian SM. Beneficial outcomes of early tracheostomy in
ducted using Observational Routinely-collected health Data (RECORD) patients requiring prolonged mechanical ventilation. Tanaffos 2020;19:350–5.
statement. PloS Med 2015;12:e1001885. 34 Chorath K, Hoang A, Rajasekaran K, Moreira A. Association of early vs
19 Siempos II, Ntaidou TK, Filippidis FT, Choi AMK. Effect of early versus late tracheostomy placement with pneumonia and ventilator days in criti-
late or no tracheostomy on mortality and pneumonia of critically ill cally ill patients: A meta-analysis. JAMA Otolaryngol Head Neck Surg
patients receiving mechanical ventilation: A systematic review and meta- 2021;147:450–9.
analysis. Lancet Respir Med 2015;3:150–8. 35 Liu CC, Livingstone D, Dixon E, Dort JC. Early versus late tracheostomy:
20 Gomes Silva BN, Andriolo RB, Saconato H, Atallah AN, Valente O. Early A systematic review and meta-analysis. JAMA Otolaryngol Head Neck
versus late tracheostomy for critically ill patients. Cochrane Database Syst Surg 2015;152:219–27.
Rev 2012:CD007271. 36 Cai SQ, Hu JW, Liu D, et al. The influence of tracheostomy timing on out-
21 Robins JM, Hernan MA, Brumback B. Marginal structural models and comes in trauma patients: A meta-analysis. Injury 2017;48:866–73.
causal inference in epidemiology. Epidemiology 2000;11:550–60. 37 Mubashir T, Arif AA, Ernest P, et al. Early versus late tracheostomy in
22 Andersen LW, Raymond TT, Berg RA, et al. Association between tracheal patients with acute traumatic spinal cord injury: A systematic review and
intubation during pediatric in-hospital cardiac arrest and survival. JAMA meta-analysis. Anesth Analg 2021;132:384–94.
2016;316:1786–97. 38 Devarajan J, Vydyanathan A, Xu M, et al. Early tracheostomy is associated
23 Nakahara S, Tomio J, Takahashi H, et al. Evaluation of pre-hospital admin- with improved outcomes in patients who require prolonged mechanical
istration of adrenaline (epinephrine) by emergency medical services for ventilation after cardiac surgery. J Am Coll Surg 2012;214:1008–16;e4.
patients with out of hospital cardiac arrest in Japan: Controlled propensity 39 Pinheiro Bdo V, Tostes Rde O, Brum CI, Carvalho EV, Pinto SP, Oliveira
matched retrospective cohort study. BMJ 2013;347:f6829. JC. Early versus late tracheostomy in patients with acute severe brain
24 Lu B. Propensity score matching with time-dependent covariates. Biomet- injury. J Bras Pneumol 2010;36:84–91.
rics 2005;61:721–8. 40 McLaughlin C, Darcy D, Park C, et al. Timing of tracheostomy placement
25 Pedersen AB, Mikkelsen EM, Cronin-Fenton D, et al. Missing data and among children with severe traumatic brain injury: A propensity-matched
multiple imputation in clinical epidemiological research. Clin Epidemiol analysis. J Trauma Acute Care Surg 2019;87:818–26.
2017;9:157–66. 41 Anand T, Hanna K, Kulvatunyou N, et al. Time to tracheostomy impacts
26 VanderWeele TJ, Ding P. Sensitivity analysis in observational research: overall outcomes in patients with cervical spinal cord injury. J Trauma
Introducing the e-value. Ann Intern Med 2017;167:268–74. Acute Care Surg 2020;89:358–64.
27 Haneuse S, VanderWeele TJ, Arterburn D. Using the e-value to assess the 42 Fokin AA, Wycech J, Chin Shue K, et al. Early versus late tracheostomy in
potential effect of unmeasured confounding in observational studies. trauma patients with rib fractures. J Surg Res 2020;245:72–80.
JAMA 2019;321:602–3. 43 Tsuchiya A, Yamana H, Kawahara T, et al. Tracheostomy and mortality in
28 Trouillet JL, Collange O, Belafia F, et al. Tracheotomy in the intensive patients with severe burns: A nationwide observational study. Burns
care unit: Guidelines from a French expert panel. Ann Intern Care 2018;44:1954–61.
2018;8:37. 44 Ohbe H, Jo T, Yamana H, Matsui H, Fushimi K, Yasunaga H. Early enteral
29 Scantling D, Granche J, Williamson J, Gracely E, Thosani D, McCracken nutrition for cardiogenic or obstructive shock requiring venoarterial extra-
B. Development of clinical tracheostomy score to identify cervical spinal corporeal membrane oxygenation: A Nationwide inpatient database study.
cord injury patients requiring prolonged ventilator support. J Trauma Intensive Care Med 2018;44:1258–65.
Acute Care Surg 2019;87:195–9. 45 Mehta AB, Syeda SN, Bajpayee L, Cooke CR, Walkey AJ, Wiener RS.
30 Nam IC, Shin YS, Jeong WJ, et al. Guidelines for tracheostomy from the Trends in tracheostomy for mechanically ventilated patients in the United
Korean Bronchoesophagological Society. Clin Exp Otohinolaryngol States, 1993-2012. Am J Respir Crit Care Med 2015;192:446–54.
2020;13:361–75. 46 Chang DW, Shapiro MF. Association between intensive care unit utiliza-
31 Arabi Y, Haddad S, Shirawi N, Al Shimemeri A. Early tracheostomy in tion during hospitalization and costs, use of invasive procedures, and mor-
intensive care trauma patients improves resource utilization: A cohort tality. JAMA Int Med 2016;176:1492–9.
study and literature review. Crit Care 2004;8:R347–52. 47 Wu MH, Tsou PY, Wang YH, et al. Impact of post-sepsis cardiovascular
32 Dasgupta S, Das S, Chawan NS, Hazra A. Nosocomial infections in the complications on mortality in sepsis survivors: A population-based study.
intensive care unit: Incidence, risk factors, outcome and associated Crit Care 2019;23:293.