Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Sustained Improvement in Tracheal Intubation

Safety Across a 15-Center Quality-Improvement


Collaborative: An Interventional Study From
the National Emergency Airway Registry for
Children Investigators*
Akira Nishisaki, MD, MSCE1
OBJECTIVES: To evaluate the effect of a tracheal intubation safety bundle Anthony Lee, MD2
on adverse tracheal intubation-associated events across 15 PICUs. Simon Li, MD, MPH3
DESIGN: Multicenter time-series study. Ronald C. Sanders Jr, MD, MS4
Calvin A. Brown III, MD5
SETTING: PICUs in the United States. Kyle J. Rehder, MD, CPPS, FCCM,
FCCP6
PATIENTS: All patients received tracheal intubations in ICUs.
Natalie Napolitano, MPH,
INTERVENTIONS: We implemented a tracheal intubation safety bundle RRT-NPS, FAARC7
as a quality-improvement intervention that includes: 1) quarterly site Vicki L. Montgomery, MD8
benchmark performance report and 2) airway safety checklists (prepro- Michelle Adu-Darko, MD, FAAP9
cedure risk factor, approach, and role planning, preprocedure bedside G. Kris Bysani, MD, FAAP, FCCM10
“time-out,” and immediate postprocedure debriefing). We define each Ilana Harwayne-Gidansky, MD,
FAAP, CHSE11
quality-improvement phase as baseline (–24 to –12 mo before checklist Joy D. Howell, MD, FAAP, FCCM12
implementation), benchmark performance reporting only (–12 to 0 mo be- Sholeen Nett, MD, PhD13
fore checklist implementation), implementation (checklist implementation Alberto Orioles, MD, FAAP14
start to time achieving > 80% bundle adherence), early bundle adherence Matthew Pinto, MD3
(0–12 mo), and sustained (late) bundle adherence (12–24 mo). Bundle Asha Shenoi, MD, Dch, FAAP15
adherence was defined a priori as greater than 80% of checklist use for David Tellez, MD16
tracheal intubations for 3 consecutive months. Serena P. Kelly, MS, CPNP-AC,
FNP-BC, CCRN, CPEN17
MEASUREMENTS AND MAIN RESULTS: The primary outcome was
Melinda Register, MS, RRT, CPFT,
the adverse tracheal intubation-associated event, and secondary outcomes RRT-NPS18
included severe tracheal intubation-associated events, multiple tracheal in- Keiko Tarquinio, MD, FAAP19
tubation attempts, and hypoxemia less than 80%. Dennis Simon, MD, PhD20
From January 2013 to December 2015, out of 19 participating PICUs, 15 Conrad Krawiec, MD21
ICUs (79%) achieved bundle adherence. Among the 15 ICUs, the adverse Justine Shults, PhD22
tracheal intubation-associated event rates were baseline phase: 217/1,241 Vinay Nadkarni, MD, MS1
(17.5%), benchmark reporting only phase: 257/1,750 (14.7%), early 0–12 for National Emergency
month complete bundle compliance phase: 247/1,591 (15.5%), and Airway Registry for Children
(NEAR4KIDS) and Pediatric
late 12–24 month complete bundle compliance phase: 137/1,002 (13.7%). Acute Lung Injury and Sepsis
After adjusting for patient characteristics and clustering by site, the adverse Investigators (PALISI)
tracheal intubation-associated event rate significantly decreased com-
pared with baseline: benchmark: odds ratio, 0.83 (0.72–0.97; p = 0.016);
early bundle: odds ratio, 0.80 (0.63–1.02; p = 0.074); and late bundle *See also p. 358.
odds ratio, 0.63 (0.47–0.83; p = 0.001). Copyright © 2020 by the Society of
CONCLUSIONS: Effective implementation of a quality-improvement Critical Care Medicine and Wolters
bundle was associated with a decrease in the adverse tracheal intubation- Kluwer Health, Inc. All Rights Reserved.
associated event that was sustained for 24 months. DOI: 10.1097/CCM.0000000000004725

250      www.ccmjournal.org February 2021 • Volume 49 • Number 2


Copyright © 2020 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Clinical Investigations

KEY WORDS: adverse event; bundle; checklist; MATERIALS AND METHODS


child; intensive care unit; intubation; pediatric;
Study Design, Setting, and Participants
safety; tracheal intubation
This study was conducted as a multicenter QI inter-

T
vention to improve the safety of TIs in critically ill chil-
racheal intubation (TI) is a life-saving, yet dren in PICUs in the United States. Each participating
hazard-prone procedure for critically ill chil- ICU received an endorsement from the local ICU
dren (1). In contrast to children intubated by committee or leadership. Participating sites were the
anesthesiologists in the operating suites, critically ill part of the NEAR4KIDS Network, an international QI
children are more likely to have a rapid oxygen de- collaborative within the Pediatric Acute Lung Injury
saturation, hemodynamic instability, and difficult and Sepsis Investigators Network (8).
airway anatomy. These pediatric-specific features All consecutive TIs in the participating ICU during
increase risks and adverse events associated with TI the study period were entered into the NEAR4KIDS
procedures (1–4). registry database. We censored TIs that occurred during
To improve the TI safety for critically ill children, the “run-in implementation phase” before achievement
a multicenter quality-improvement (QI) collabora- of 80% compliance (defined as the time from check-
tive network: National Emergency Airway Registry list introduction in the ICU to time achieving > 80%
for Children (NEAR4KIDS) was developed. The bundle adherence for 3 mo) in the original analysis.
collaborative reported 20% of all TIs are associated We excluded the ICUs that did not achieve bundle ad-
with adverse tracheal intubation-associated events herence, because our goal was to evaluate the clinical
(TIAEs), and 3% are associated with severe events effectiveness of the bundled QI interventions after each
such as cardiac arrests (1). Nineteen percent of TIs ICU met and sustained the predefined QI bundle ad-
are also associated with hypoxemia with pulse ox- herence criteria. Therefore, only the TI data from the
imetry Spo2 < 80% (3). Identified risks for adverse ICUs that achieved bundle adherence were included in
TIAEs were respiratory failure, hemodynamic in- the main analyses. The TI data from the ICUs that did
stability, and nonanesthesia-resident participation not reach bundle adherence were analyzed separately.
as primary laryngoscopists (2–5). The occurrence
of either adverse TIAEs or severe hypoxemia dur- Definitions and Interventions
ing TI was associated with a longer duration of me- We defined each QI phase as baseline phase (–24 to
chanical ventilation, and the occurrence of severe –12 mo before checklist implementation), benchmark
adverse events was associated with increased ICU performance reporting only phase (–12 to 0 mo before
mortality (6). checklists introduced), run-in implementation phase
Leveraging our knowledge in the risk factors as- (time from checklist introduction to time achieving
sociated with adverse TIAEs and hypoxemia events, > 80% bundle adherence for 3 consecutive months),
we have developed a bundle of QI interventions for early bundle adherence phase (0–12 mo after > 80%
diverse PICUs (7). This bundle includes: 1) quarterly bundle adherence), and late (sustained) bundle adher-
benchmark performance data and debriefing feed- ence phase (12–24 months after the > 80% bundle ad-
back to ICUs and 2) implementation of airway safety herence). Participating ICUs received their quarterly
checklists, which consists of the following three com- benchmark performance reports after 12 months of
ponents: a) TI procedure planning with risk factor re- NEAR4KIDS participation (baseline). These bench-
view by bedside clinicians, b) time-out immediately mark reports included the number of TIs, patient and
before the TI to review team member roles and ex- provider demographics, practice characteristics, ad-
pected behaviors, and c) bedside team debriefing im- verse TIAEs, and hypoxemia rates for their ICU (site),
mediately after the TI procedure. and overall in the NEAR4KIDS collaborative. Since the
Our objective was to evaluate the clinical effective- majority of sites were already participating in the reg-
ness of the bundled QI interventions to reduce adverse istry before the study phase, some site leaders received
TIAEs after each ICU met and sustained the prede- their benchmark report during the baseline phase.
fined QI bundle adherence criteria. However, for the sites whose leaders were already

Critical Care Medicine www.ccmjournal.org      251


Copyright © 2020 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Nishisaki et al

receiving reports, an emphasis on quarterly debriefing TIAEs include cardiac arrests, esophageal intubation
with the site team on safety outcomes as well as pro- with delayed recognition, emesis with aspiration, hy-
cess variances (e.g., multiple attempts) was made dur- potension, laryngospasm, and pneumothorax/pneu-
ing the benchmark performance reporting only phase momediastinum. The nonsevere TIAEs include main
(–12 to 0 mo before checklist implementation). stem bronchial intubation, esophageal intubation with
The airway safety checklist was developed using the immediate recognition, emesis without aspiration,
observational data from the NEAR4KIDS database to hypertension requiring treatment, epistaxis, dental/
address patient, provider, and practice risk factors asso- lip trauma, medication error, dysrhythmia, and pain/
ciated with the adverse TI outcomes (7). The checklist agitation delaying the procedure. Data were initially
was designed to align with clinicians’ bedside work- collected by the bedside clinicians after each TI using
flow: preparation (risk identification and planning for a standard data collection form, followed by a data-
provider, approach, and backup plan), preprocedure verification process by each site project team following
timeout, and postprocedural debriefing. In the rare sit- a site-specific compliance plan. The compliance plan
uation of emergency, the preparation section might be required a capture rate of greater than 95% of TIs at
omitted (by indicating as an emergent procedure such each site. The data-coordinating center crosschecked
as pericardiac arrest) but the preprocedure timeout the data entry and communicated with the site leader
and postprocedural debriefing were still required. The and coordinator for clarification of the data when
checklist implementation process used the following needed. The operational definitions were used consist-
five elements: 1) strong encouragement of site lead- ently throughout the data collection.
ers to develop a multidisciplinary team for QI bundle
implementation, 2) best practice videos available to Outcomes
all sites (www.youtube.com/watch?v=YCGPOl6E1rc&
feature=youtu.be), 3) data feedback with a run chart The primary outcome was the occurrence of any ad-
with checklist compliance rate, 4) quarterly teleconfer- verse TIAE. The secondary outcomes included severe
ences to address issues related to QI implementation, TIAEs, multiple TI attempts (defined as three or more
and 5) semiannual (twice a year) face-to-face meetings attempts), and hypoxemia (3, 10). Hypoxemia (oxygen
for peer learning and exchange of barriers and facili- desaturation) was defined as the lowest oxygen pulse
tators. Detailed data regarding the QI implementation oximetry during TI procedure less than 80% in a TI en-
process, barriers, and facilitators were previously pub- counter despite initial pulse oximetry level was greater
lished (9). Full bundle adherence was a priori defined than 90% after preoxygenation (3, 6). Therefore, those
as greater than 80% compliance with a checklist use for infants who did not achieve Spo2 > 90% after preoxy-
3 consecutive months for each ICU. genation (i.e., unrepaired cyanotic heart disease) were
not considered to have hypoxemia even when they ex-
Data Collection perienced a drop in Spo2 below 80% during intubation.

The TI data collection was approved by the Institutional


Statistical Analysis
Review Board (IRB) or review was waived as a QI ac-
tivity at each NEAR4KIDS participating site. Each Sample size calculation was not performed due to the
IRB granted a waiver to obtain an individual informed nature of QI interventions and initial uncertainties in
consent from patients, guardians, or providers for data the number of participating sites and sites achieving
collection. adherence criteria. We planned to recruit at least 15
The NEAR4KIDS dataset includes patient, provider, sites with 1,200 TIs during the baseline phase to be
and practice characteristics as well as the outcomes able to detect a 4.5% reduction in adverse outcomes
(i.e., attempt success, the occurrence of predefined ad- during the checklist adherence phase. To evaluate each
verse TIAEs, and hypoxemia with Spo2 < 80%) and component of the QI bundle in reducing the adverse
process variances (i.e., multiple attempts). Consensus outcomes, we a priori decided to use the baseline phase
data definitions for data elements and outcomes are (12 mo) as a reference throughout the analyses.
previously published (1, 5). Briefly, the adverse TIAEs Categorical variables were described as number and
include both severe and nonsevere TIAEs. The severe percentages, and nonnormally distributed continuous

252      www.ccmjournal.org February 2021 • Volume 49 • Number 2


Copyright © 2020 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Clinical Investigations

variables as median and interquartile ranges (IQRs). For collection (i.e., baseline phase and benchmark perfor-
the estimated odds ratio (OR), a 95% CI was presented. mance reporting only phase) implemented QI bundles
To evaluate the effectiveness of QI bundle implemen- with checklists. Fifteen ICUs (79%) achieved greater
tation, the following steps were taken. First, a univariate than 80% QI bundle compliance for 3 consecutive
analysis was performed to evaluate each phase of the QI months over a median of 395 days (IQR, 119–752 d)
intervention on primary and secondary outcomes with (Supplemental Table B, Supplemental Digital Content
chi-square tests. Second, a multivariable analysis was per- 2, http://links.lww.com/CCM/F958). Pediatric Critical
formed with patient factors associated with the QI inter- Care Medicine Fellowship Programs were present in
vention phase as covariates. The run-in implementation 66% of the participating ICUs. There was a system for
phase (time from checklist introduction to time achiev- 24-hour in-hospital ICU attending coverage in 73%
ing > 80% bundle adherence for 3 consecutive months) of sites. Checklist compliance increased over time
was excluded from the main analysis, since the goal of (Supplemental Fig. C, Supplemental Digital Content
this project is to evaluate the TI safety outcomes after 3, http://links.lww.com/CCM/F959).
each ICU met and sustained the predefined QI bundle
adherence criteria (Supplemental Fig. A, Supplemental Patient, Provider, and Practice Characteristics
Digital Content 1, http://links.lww.com/CCM/F957).
Pediatric Index of Mortality 2 (PIM2), as a patient se- Table 1 shows patient characteristics during baseline
verity score, was missing in 14% of the TI encounters. and benchmark performance feedback phases (24 mo,
Therefore, the PIM2 score was imputed with missing at before checklist implementation) versus after greater
random assumption. This method is more robust than than 80% QI checklist adherence was achieved. Fewer
simply eliminating the subjects with missing values (11). intubations for procedural indication were reported after
Note that in our multivariable analysis, we adjusted for checklist implementation. Patient age was older and
patient factors (e.g., age, severity of illness, difficult airway PIM2 was lower in TIs after the QI checklist adherence
features, and indications) and clustering by site, but not was achieved (these differences are controlled for in the
for provider (e.g., experience level of laryngoscopist) multivariable analysis). Table 2 describes provider and
or measured practice factors (e.g., equipment use and, practice characteristics. TIs by attending physicians and
medications used), because the bundle checklists were residents decreased after achieving bundle adherence.
specifically intended to have clinicians match provider
skill sets and the approach selections to the anticipated TI Effect of the Bundled Quality-Improvement
risks and difficulties. We did adjust for clustering by site Intervention
to account for unmeasured variables in patient popula-
The rate of adverse TIAEs decreased from 17.5% (95%
tions, provider skill level, or practice patterns that might
CI, 15.4–19.7%) in the baseline phase to 13.7% (95%
contribute to the TIAEs that were site-specific.
CI, 11.6–16.0%) in the 24 months after greater than
A sensitivity analysis was conducted to include the
80% checklist adherence was achieved (Table 3). After
TIs during the run-in implementation phase, which
adjusting for patient age, PIM2 predicted mortality,
was excluded in the original analysis. Multivariable
history of difficult airway, difficult airway features, TI
analyses were repeated after the missing PIM2 values
indications, and site-level clustering, the QI phases
were imputed with this cohort. We also evaluated the
were associated with decreased prospectively de-
potential secular trend by analyzing the TI safety out-
fined adverse TIAEs: benchmark performance phase
comes across the four sites that did not meet the criteria
OR 0.84 (95% CI, 0.73–0.97; p = 0.022), early (0–12
of bundle adherence. All data analyses were conducted
mo) checklist bundle greater than 80% compliance
using STATA 14.0 (StataCorp, College Station, TX).
phase OR 0.80 (95% CI, 0.63–1.02; p = 0.074), and late
(12–24 mo) checklist bundle greater than 80% com-
RESULTS pliance phase OR 0.63 (95% CI, 0.47–0.83; p = 0.001)
(Table 4). Figure 1 shows the probability of adverse
ICU Characteristics
TIAE rates for each quarter (3 mo) before checklist
From January 2013 to December 2015, a total of implementation, and after greater than 80% checklist,
19 ICUs that completed 24 months of baseline data bundle adherence was achieved.

Critical Care Medicine www.ccmjournal.org      253


Copyright © 2020 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Nishisaki et al

TABLE 1.
Patient Characteristics Before Checklist Implementation and After Bundle Checklist
Adherence
Before Checklist After Checklist
Implementation Adherence
Phases (n = 2,991) (n = 2,593) p

Age
  Age, yr (median [IQR]) 1 (0–6) 1 (0–8) 0.0011
 Infant, n (%) 1,331 (44) 1,040 (40)
  Child (1–7 yr), n (%) 981 (33) 904 (35)
  Older child (8-17 yr), n (%) 563 (19) 533 (21)
  Adult (18 yr or above), n (%) 116 (4) 116 (4)
Pediatric Index of Mortality 2 (%, median [IQR])
a
2.9 (0.9–6.9) 2.1 (0.9–4.9) < 0.001
Diagnosis
  Lower respiratory, n (%) 943 (33) 864 (33) < 0.001
  Upper respiratory, n (%) 248 (9) 200 (8)
 Neurologic, n (%) 506 (18) 623 (24)
 Cardiac-surgical, n (%) 317 (11) 178 (7)
 Cardiac-medical, n (%) 135 (5) 138 (5)
 Sepsis/shock, n (%) 171 (6) 186 (7)
 Trauma, n (%) 92 (3) 65 (3)
 Other, n (%) 439 (15) 339 (13)
Indication for tracheal intubationb
 Respiratory, n (%) 1,585 (53) 1,418 (55) 0.206
 Shock, n (%) 310 (10) 304 (12) 0.105
 Procedural, n (%) 618 (21) 477 (18) 0.033
 Neurologic , n (%)
c 281 (9) 324 (13) < 0.001
Difficult airway features
  History of difficult airway, n (%) 406 (14) 425 (16) 0.003
  Any difficult airway feature, n (%) 1,092 (37) 881 (34) 0.048
IQR = interquartile range.
Pediatric Index of Mortality 2 was missing in 785 cases: 362 cases (12%) before checklist implementation and 423 cases (16%) after
a

checklist adherence.
Patients may have more than one indication for intubation.
b

Neurologic indication includes therapeutic hyperventilation, airway protection, and neuromuscular weakness.
c

The occurrence of multiple attempts and hypoxemia Tracheal Intubation Safety From the Sites That
with Spo2 < 80% did not significantly decrease after the Did Not Achieve Bundle Adherence
sites achieved bundle adherence.
Sensitivity analysis including all TIs during the The rate of adverse TIAEs, multiple attempts, and hypox-
run-in implementation phase revealed a similar re- emia did not change across QI phases among the four
sult (Supplemental Table D, Supplemental Digital ICUs that did not achieve bundle adherence: baseline
Content 4, http://links.lww.com/CCM/F960). phase (15.1%), benchmark performance reporting only

254      www.ccmjournal.org February 2021 • Volume 49 • Number 2


Copyright © 2020 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Clinical Investigations

TABLE 2.
Provider and Practice Characteristics Before and After Bundle Checklist Adherence
Before Checklist After Checklist
Implementation Adherence
Phases (n = 2,991) (n = 2,593) p

Provider, n (%)
  PCCM/EM attending 877 (30) 505 (19) < 0.001
  PCCM/EM Fellow 853 (29) 1,164 (45)
  Peds/EM resident 718 (24) 356 (14)
  Nurse practitioner 216 (7) 280 (11)
 Hospitalist 5 (0) 14 (1)
  Respiratory therapist 6 (0) 8 (0)
 Subspecialist 158 (5) 159 (6)
 Other 158 (5) 107 (4)
Devicea, n (%)
  Direct laryngoscopy 2,578 (87) 2,225 (88) 0.357b
  Video laryngoscopy 318 (11) 297 (12)
 Other 44 (1) 19 (1)
Method of intubation, n (%)
  Initial intubation 2,612 (88) 2,307 (89) 0.144
  Tube change 366 (12) 286 (11)
Medication, n (%)
 Atropine 675 (23) 600 (23) 0.612
 Glycopyrrolate 172 (6) 272 (10) < 0.001
 Fentanyl 1,617 (54) 1,660 (64) < 0.001
 Midazolam 1,555 (52) 1,334 (51) 0.685
 Ketamine 837 (28) 590 (23) < 0.001
 Propofol 492 (16) 496 (19) 0.009
 Etomidate 53 (2) 34 (1) 0.166
  Any paralytic 2,557 (85) 2,223 (86) 0.798
 Rocuronium 1,891 (63) 1,924 (74) < 0.001
 Vecuronium 403 (13) 281 (11) 0.003
 Cisatracurium 267 (9) 22 (1) < 0.001
 Succinylcholine 15 (1) 19 (1) 0.268
Endotracheal tube type, n (%)
 Cuffed 2,808 (94) 2,479 (96) 0.003
 Uncuffed 166 (6) 93 (3)
 Other ≠
15 (0) 21 (1)
EM = Emergency Medicine, PCCM = Pediatric Critical Care Medicine.
Device data were missing in eight cases (0.2%).
a

p value was calculated for tracheal intubations with direct versus video laryngoscopy.
b

Critical Care Medicine www.ccmjournal.org      255


Copyright © 2020 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Nishisaki et al

TABLE 3.
Univariate Analysis: Primary and Secondary Outcomes Before and After Bundle Checklist
Adherence
Checklist Checklist
Baseline Benchmark Adherence Adherence
(Year 1), (Year 2), (Year 1), (Year 2), Total,
Phase n = 1,241 n = 1,750 n = 1,591 n = 1,002 n = 5,584 p

Primary outcome, n (%)


  Any adverse TIAEs 217 (17.5) 257 (14.7) 247 (15.5) 137 (13.7) 858 (15.4) 0.067
Secondary outcome, n (%)
  Severe TIAEs 67 (5.4) 87 (5.0) 76 (4.8) 54 (5.4) 284 (5.1) 0.848
  Nonsevere TIAEs 167 (13.5) 193 (11.0) 187 (11.8) 99 (9.9) 646 (11.6) 0.053
  Multiple attempts 130 (10.5) 222 (12.7) 195 (12.3) 133 (13.3) 680 (12.2) 0.178
  Oxygen desaturation 209 (16.8) 301 (17.2) 282 (17.7) 188 (18.8) 980 (17.6) 0.652
TIAEs = tracheal intubation-associated events.
Baseline phase consists of –24 to –12 mo before the checklist implementation. During this baseline phase, some sites received benchmark


feedback data.
Benchmark phase consists of –12 mo to the date when the bundle checklist was officially implemented. See the Materials and Methods


section for details.


Checklist adherence phase year 1 is defined as the first 12 months from the month when the adherence criteria were first met. For the
adherence criteria, refer to the Materials and Methods section of the article.
p value is calculated by χ2 analysis.
This table does not include the data during bundle implementation (from implementation date to checklist adherence > 80%).

phase (13.7%), and run-in implementation phase (13.4%) ICU leadership buy-in, and involving multidisciplinary
(p = 0.672) (Supplemental Table E, Supplemental Digital clinicians (12). The checklists thoroughly addressed
Content 5, http://links.lww.com/CCM/F961). three critical components of safe TI procedure: initial
patient risk factor assessment and planning, preproce-
dural time out, and scripted postprocedure debriefs.
DISCUSSION
The effect of benchmark performance review and
Our study demonstrated that TI procedural safety checklist implementation was significant, and the im-
was enhanced across 15 PICUs after successful imple- pact was most evident in the sustaining maintenance
mentation and sustained maintenance of QI bundles phase. Specifically, the adverse TIAE rate was signif-
with checklists, performance reports, and individual icantly reduced initially with the implementation of
postprocedural peer debriefing. The vast majority of the benchmark performance review dashboards. It
the ICUs were able to achieve QI bundle adherence continued to decrease during the first 12 months of
defined a priori as greater than 80% use of QI bundle compliant checklist implementation (early), achiev-
checklists. The rate of adverse TIAEs decreased signif- ing statistical significance during the 12–24-month
icantly over time across the successfully implementing sustaining maintenance (late) phase. This suggests that
sites, after adjusting for important patient-level factors impact may also be influenced by the duration of im-
(e.g., age, severity of illness, difficult airway features, plementation (sustained penetration). Once bundle
and indication). These effects were sustained for at adherence of greater than 80% was achieved, adverse
least 24 months. TI event rates steadily decreased over time. This finding
Our QI intervention incorporated best practice QI contrasts with the unchanged rates of adverse TIAEs in
approaches, including benchmarking, evidence-based the four ICUs that did not achieve bundle adherence.
checklist development and implementation, leveraging Substantial variance was observed in the reduction of

256      www.ccmjournal.org February 2021 • Volume 49 • Number 2


Copyright © 2020 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Clinical Investigations

TABLE 4.
Multivariable Analysis for Any Adverse Tracheal Intubation-Associated Events and Quality-
Improvement Implementation Phase, Adjusting for Patient Factors and Site-Level Clustering
Variable OR (95% CI) p

Quality-improvement phase
 Baseline Reference
  Benchmark performance 0.84 (0.73–0.97) 0.022
  Early checklist adherence (year 1) 0.80 (0.63–1.02) 0.074
  Late checklist adherence (year 2) 0.63 (0.47–0.83) 0.001
Patient factors
 Age
   Infant (< 12 mo) Reference
   Young child (1–7 yr) 0.95 (0.79–1.14) 0.591
   Older child (8–17 yr) 0.94 (0.72–1.21) 0.618
   Adult (18 yr or above) 1.14 (0.83–1.57) 0.424
Pediatric Index of Mortality 2 scorea 1.00 (1.00–1.01) 0.331
Indication for tracheal intubation
 Respiratory 1.33 (1.07–1.65) 0.011
 Shock 1.58 (1.33–1.88) < 0.005
 Procedural 1.06 (0.85–1.32) 0.612
 Neurologic 1.10 (0.91–1.32) 0.334
Difficult airway features
  History of difficult airway 1.23 (0.98–1.53) 0.072
  Any difficult airway feature 1.10 (0.93–1.31) 0.276
OR = odds ratio.
Tracheal intubation-associated events denotes tracheal intubation-associated events.
Baseline phase consists of –24 to –12 mo before the checklist implementation.
Benchmark performance consists of –12 mo to the date when the bundle checklist was officially implemented.
Multiple imputation was used to impute Pediatric Index of Mortality 2 in 785 cases. No other variables were imputed.
a

Multivariable logistic regression using generalized estimate equation with an independent covariate structure and binomial linkage.
This table does not include the data during bundle implementation (from implementation date to checklist adherence > 80%).

adverse TIAEs across the ICUs (13). Future studies are expect several components of the bundle checklist (risk
needed to identify the organizational-level factors that identification and planning in provider, approach, and
may be responsible for the variance. backup plan) would reduce the multiple attempts and
It is notable that none of our secondary outcomes, oxygen desaturation rates. This lack of decrease in mul-
including severe TIAE, multiple TI attempts, and hy- tiple attempts and hypoxemia may be a limitation of
poxemia with Spo2 < 80%, were substantially decreased our checklist-based intervention, which was designed
during the QI implementation despite the lower PIM2 to reduce TIAEs primarily focused on the skills and
predicted mortality during this period. One would behaviors during laryngoscopy and endotracheal tube

Critical Care Medicine www.ccmjournal.org      257


Copyright © 2020 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Nishisaki et al

the main analysis, pre-


suming the impact of the
full QI bundle would not
be achieved unless check-
list compliance was greater
than 80% for 3 months at
each site. We addressed
this limitation with the
sensitivity analysis, which
demonstrated a similar
clinical effectiveness of QI
bundle implementation
with the run-in implemen-
tation phase included. We
excluded four ICUs that
did not achieve the bundle
adherence in our analyses.
This limits the generaliza-
bility of our study findings.
Figure 1. Estimated probability for adverse tracheal intubation-associated event. Quarter indicates In other words, those ICUs
3 mo on x-axis. The vertical line indicates the time at which each ICU met the criteria for bundle that achieved the adher-
compliance (> 80% checklist use). TI = tracheal intubation. ence may be different from
those that did not in their
placement, and did not focus on bag-mask ventilation ICU structure, safety culture, and TI processes. These
effectiveness, specific devices, or alternative airway four ICUs that did not achieve adherence were large
management skills. Our QI bundle incorporated the academic ICUs with residency and fellowship pro-
current best practice with specific attention to a shared grams. The more complex system in a large academic
mental model, anticipation, and planning. However, ICU may require a well-structured interdisciplinary
this might not be sufficient to reduce the rate of severe QI implementation leadership and a robust commu-
TIAEs, multiple attempt rates, or the occurrence of hy- nication with frontline clinicians as identified in our
poxemia during the TI procedure, which may be more previous mixed-method study (9). Although reporting
related to the severity of illness and trajectory of di- bias could occur, each site developed and followed a
sease than procedural intervention quality. Specific QI rigorous site-specific compliance plan to ensure sys-
interventions, such as: 1) optimizing hemodynamics tematic capture of all TI events and accurate data entry.
to prevent cardiac arrest during TI by early utilization The time-series study design could be prone to
of epinephrine infusion and correction of metabolic changes in secular trends and other unmeasured factors
acidosis (2), 2) use of passive oxygen flow during lar- affecting safety practices. Since each site implemented
yngoscopy (i.e., apneic oxygenation) (14–18), and 3) the QI bundle at various times of the year and achieved
routine use of video laryngoscopy to reduce multiple greater than 80 % adherence at different times, making
attempts and esophageal intubations (19), may poten- seasonal trend biases from patient risk factors (e.g., more
tially result in safer TI practice with lower severe event respiratory failure patients during winter time) or pro-
rates, fewer multiple attempts, and fewer hypoxemia vider expertise (e.g., trainee expertise) less likely. Our
events during TI. Several PICUs have recently imple- analyses accounted for clustering by sites to mitigate po-
mented apneic oxygenation and/or video laryngos- tential unmeasured factors specific to a site. Our analysis
copy as the next set of QI interventions (14). of TI data from four ICUs that did not achieve bundle
Our study has limitations inherent to time-series adherence did not show significant improvement in TI
study design and the nature of QI activity. First, we a safety outcomes despite their baseline lower TIAE rates.
priori eliminated the QI implementation phase from This result suggests that the significant and sustained

258      www.ccmjournal.org February 2021 • Volume 49 • Number 2


Copyright © 2020 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Clinical Investigations

improvement requires the bundle adherence, which was 9 Division of Critical Care, Children’s Hospital of the University
most visible after adjusting for patient factors. It is also of Virginia, Charlottesville, VA.
possible that our intervention was only able to reduce the
10 Pediatric Critical Care Medicine, Pediatric Acute Care
adverse TIAE rates to a certain level; therefore, the non- Associates of North Texas PLLC, Medical City Children’s
Hospital, Dallas, TX.
adherence sites failed to show an improvement due to
their baseline lower TIAE rates. However, we are not able 11 Division of Critical Care, Stony Brook Children’s Hospital,
Stony Brook, NY.
to eliminate or adjust for the indirect effect of other spe-

12 Department of Pediatrics, New York-Presbyterian Weill
cific concurrent QI activities, the Hawthorne effect, and
Cornell Medical Center, New York, NY.
general changes in medical practice affecting multiple
13 Division of Pediatric Critical Care, Department of Pediatrics,
sites. Finally, our study did not evaluate the long-term
Dartmouth-Hitchcock Medical Center, Lebanon, NH.
outcomes such as the duration of mechanical ventilation
14 Division of Critical Care, Children’s Hospital and Clinics of
or ICU length of stay, since our QI intervention focused
Minnesota, Minneapolis, MN.
on the intubation procedure, that is, the beginning of in-
15 Department of Pediatrics, Division of Pediatric Critical Care,
vasive mechanical ventilation. Future QI interventions Kentucky Children’s Hospital, University of Kentucky School
should address the ventilation-weaning process with the of Medicine, Lexington, KY.
length of mechanical ventilation as the targeted outcome. 16 Pediatric Critical Care Medicine, Department of Pediatrics,
Phoenix Children’s Hospital, Phoenix, AZ.
CONCLUSIONS 17 Division of Pediatric Critical Care Medicine, Doernbecher
Effective implementation of a QI bundle in 15 collab- Children’s Hospital, Portland, OR.
orative sites was associated with a significant decrease 18 Department of Respiratory Therapy, Children’s Healthcare of
Atlanta, Atlanta, GA.
in the adverse event rate that was sustained for 24
months. These effects persisted after adjusting for age, 19 Division of Pediatric Critical Care Medicine, Department of
Pediatrics, Emory University School of Medicine, Atlanta
patient factors, mortality risk, and clustering by site.
GA.
20 Department of Anesthesiology and Critical Care Medicine,
ACKNOWLEDGMENT University of Pittsburgh School of Medicine, Pittsburgh, PA.
We thank all National Emergency Airway Registry for 21 Division of Pediatric Critical Care Medicine, Department of
Children sites participating in airway safety quality- Pediatrics, Penn State Hershey Children’s Hospital, Pennsylvania
State University College of Medicine, Hershey, PA.
improvement bundles for their dedication and hard
work. We also thank Hayley Buffman, MPH, and Cassie 22 Department of Biostatistics, Perelman School of Medicine at
the University of Pennsylvania, Philadelphia, PA.
Simpson-Dukes for their administrative support.
Supplemental digital content is available for this article. Direct
URL citations appear in the printed text and are provided in the
1 Department of Anesthesiology and Critical Care Medicine, HTML and PDF versions of this article on the journal’s website
Children’s Hospital of Philadelphia, Philadelphia, PA. (http://journals.lww.com/ccmjournal).
2 Division of Critical Care Medicine, Department of Pediatrics, Drs. Nishisaki, Napolitano, Shults, and Nadkarni are supported
Nationwide Children’s Hospital, Ohio State University, by the Agency for Healthcare Research and Quality (AHRQ
Columbus, OH. R03HS021583, R18HS022464, and R18HS024511). Dr.
3 Department of Pediatrics, Division of Pediatric Critical Care Nadkarni is supported by the Endowed Chair in Critical Care
Medicine, Maria Fareri Children’s Hospital, Valhalla, NY. Medicine at Children’s Hospital of Philadelphia. Dr. Napolitano’s
4 Section of Critical Care, Department of Pediatrics, Arkansas institution received funding from AHRQ, Draeger, Aerogen,
Children’s Hospital, Little Rock, AR. Philips/Respironics, Smiths Medical, and VERO-Biotech. Dr.
Howell received funding from UptoDate. Dr. Nadkarni’s institu-
5 Department of Emergency Medicine, Brigham and Women’s
tion received funding from AHRQ R18. The remaining authors
Hospital, Harvard Medical School, Boston, MA.
have disclosed that they do not have any conflicts of interest.
6 Division of Pediatric Critical Care, Department of Pediatrics,
Successful implementation of a tracheal intubation patient safety
Duke Children’s Hospital, Durham, NC.
bundle intervention improves outcomes and is sustained in across
7 Department of Respiratory Therapy, Children’s Hospital of PICUs
Philadelphia, Philadelphia, PA.
For information regarding this article, E-mail: nishisaki@email.
8 Division of Pediatric Critical Care, University of Louisville chop.edu
and Norton Children’s Hospital, Louisville, KY.

Critical Care Medicine www.ccmjournal.org      259


Copyright © 2020 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Nishisaki et al

REFERENCES bundle: A mixed-method analysis. Pediatr Crit Care Med 2017;


18:965–972
1. Nishisaki A, Turner DA, Brown CA 3rd, et al; National 10. Lee JH, Turner DA, Kamat P, et al; Pediatric Acute Lung Injury
Emergency Airway Registry for Children (NEAR4KIDS); and Sepsis Investigators (PALISI); National Emergency Airway
Pediatric Acute Lung Injury and Sepsis Investigators (PALISI) Registry for Children (NEAR4KIDS): The number of tracheal
Network: A National emergency airway registry for children:
intubation attempts matters! A prospective multi-institutional
Landscape of tracheal intubation in 15 PICUs. Crit Care Med
pediatric observational study. BMC Pediatr 2016; 16:58
2013; 41:874–885
11. Janssen KJ, Donders AR, Harrell FE Jr, et al: Missing covariate
2. Shiima Y, Berg RA, Bogner HR, et al; National Emergency
data in medical research: To impute is better than to ignore. J
Airway Registry for Children Investigators: Cardiac arrests as-
Clin Epidemiol 2010; 63:721–727
sociated with tracheal intubations in PICUs: A multicenter co-
12. Klugman D, Melton K, Maynord PO, et al: Assessment of

hort study. Crit Care Med 2016; 44:1675–1682
an unplanned extubation bundle to reduce unplanned extu-
3. Li S, Hsieh TC, Rehder KJ, et al; for National Emergency
bations in critically ill neonates, infants, and children. JAMA
Airway Registry for Children (NEAR4KIDS) and Pediatric
Pediatr 2020; 174:e200268
Acute Lung Injury and Sepsis Investigators (PALISI) Network:
Frequency of desaturation and association with hemodynamic 13. Nett S, Emeriaud G, Jarvis JD, et al; NEAR4KIDS Investigators
adverse events during tracheal intubations in PICUs. Pediatr and Pediatric Acute Lung Injury and Sepsis Investigators
Crit Care Med 2018; 19:e41–e50 (PALISI) Network: Site-level variance for adverse tracheal
intubation-associated events across 15 North American
4. Graciano AL, Tamburro R, Thompson AE, et al: Incidence
PICUs: A report from the national emergency airway registry
and associated factors of difficult tracheal intubations in
for children. Pediatr Crit Care Med 2014; 15:306–313
pediatric ICUs: A report from National Emergency Airway
Registry for Children: NEAR4KIDS. Intensive Care Med 2014; 14. Napolitano N, Laverriere EK, Craig N, et al; National Emergency
40:1659–1669 Airway Registry for Children (NEAR4KIDS) and Pediatric
Acute Lung Injury and Sepsis Investigators (PALISI): Apneic
5. Sanders RC Jr, Giuliano JS Jr, Sullivan JE, et al; National
oxygenation as a quality improvement intervention in an aca-
Emergency Airway Registry for Children Investigators and
demic PICU. Pediatr Crit Care Med 2019; 20:e531–e537
Pediatric Acute Lung Injury and Sepsis Investigators Network:
Level of trainee and tracheal intubation outcomes. Pediatrics 15. Chaudhuri D, Granton D, Wang DX, et al: Moderate certainty
2013; 131:e821–e828 evidence suggests the use of high-flow nasal cannula does not
6. Parker MM, Nuthall G, Brown C 3rd, et al; Pediatric Acute Lung decrease hypoxia when compared with conventional oxygen
Injury and Sepsis Investigators (PALISI) Network: Relationship therapy in the peri-intubation period: Results of a systematic
between adverse tracheal intubation associated events and review and meta-analysis. Crit Care Med 2020; 48:571–578
PICU outcomes. Pediatr Crit Care Med 2017; 18:310–318 16. Binks MJ, Holyoak RS, Melhuish TM, et al: Apneic oxygenation
7. Li S, Rehder K, Giuliano J, et al; for the National Emergency during intubation in the emergency department and during re-
Airway Registry for Children (NEAR4KIDS) and Pediatric trieval: A systematic review and meta-analysis. Am J Emerg
Acute Lung Injury and Sepsis Investigators (PALISI) Network: Med 2017; 35:1542–1546
Development of a quality improvement bundle to reduce tra- 17. Binks MJ, Holyoak RS, Melhuish TM, et al: Apnoeic oxygena-
cheal intubation-associated event in pediatric ICUs. Am J Med tion during intubation in the intensive care unit: A systematic
Qual 2016; 31:47–55 review and meta-analysis. Heart Lung 2017; 46:452–457
8. Randolph AG, Meert KL, O’Neil ME, et al; Pediatric Acute Lung 18. Soneru CN, Hurt HF, Petersen TR, et al: Apneic nasal oxy-
Injury and Sepsis Investigators Network: The feasibility of con- genation and safe apnea time during pediatric intubations by
ducting clinical trials in infants and children with acute respira- learners. Paediatr Anaesth 2019; 29:628–634
tory failure. Am J Respir Crit Care Med 2003; 167:1334–1340 19. Grunwell JR, Kamat PP, Miksa M, et al; National Emergency
9. Davis KF, Napolitano N, Li S, et al; National Airway Emergency Airway Registry for Children (NEAR4KIDS) and the Pediatric
for Children (NEAR4KIDS) and Pediatric Acute Lung Injury Acute Lung Injury and Sepsis (PALISI) Network: Trend and
and Sepsis Investigators (PALISI) Network: Promoters and outcomes of video laryngoscope use across PICUs. Pediatr
barriers to implementation of tracheal intubation airway safety Crit Care Med 2017; 18:741–749

260      www.ccmjournal.org February 2021 • Volume 49 • Number 2


Copyright © 2020 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.

You might also like