Gupta et al Congenital Heart Disease

Risk factors for mechanical ventilation and reintubation after

pediatric heart surgery
Punkaj Gupta, MBBS,a Mallikarjuna Rettiganti, PhD,b Jeffrey M. Gossett, MS,b Justin C. Yeh, MD,c
Howard E. Jeffries, MD, MBA,d,e Tom B. Rice, MD,e,f and Randall C. Wetzel, MBBSe,g

ABSTRACT
Objective: To determine the prevalence of and risk factors associated with the
need for mechanical ventilation in children following cardiac surgery and the
need for subsequent reintubation after the initial extubation attempt.
Methods: Patients younger than 18 years who underwent cardiac operations for
congenital heart disease at one of the participating pediatric intensive care units
(ICUs) in the Virtual PICU Systems (VPS), LLC, database were included
(2009-2014). Multivariable logistic regression models were fitted to identify
factors likely associated with mechanical ventilation and reintubation. Prevalence of reintubation with increasing length of
Results: A total of 27,398 patients from 62 centers were included. Of these, 6810 ICU stay among the study patients.
patients (25%) were extubated in the operating room (OR), whereas 20,588
patients (75%) arrived intubated in the ICU. Of the patients who were extubated Central Message
in the OR, 395 patients (6%) required reintubation. In contrast, 2054 patients Although the prevalence of extubation in the
operating room is low, it is associated with
(10%) required reintubation among the patients arriving intubated postopera- low complication rate.
tively in the ICU. In adjusted models, patient characteristics, patients undergoing

CHD
high-complexity operations, and patients undergoing operations in lower-volume Perspective
centers were associated with higher likelihood for the need for postoperative Our data suggest that both odds of mechanical
mechanical ventilation and need for reintubation. Furthermore, the prevalence ventilation and odds of reintubation following
of mechanical ventilation and reintubation was lower among the centers with a pediatric cardiac surgery are a function of pa-
dedicated cardiac ICU in propensity-matched analysis among centers with and tient characteristics, surgical risk category, cen-
without a dedicated cardiac ICU. ter volume, and presence of a dedicated cardiac
ICU. We also demonstrated that approximately
Conclusions: This multicenter study suggests that proportion of patients one-quarter of patients are extubated in the
extubated in the OR after heart operation is low. These data further suggest operating room with low rate of reintubation
and low mortality.
that extubation in the OR can be done successfully with a low complication
rate. (J Thorac Cardiovasc Surg 2015;-:1-8)

From aDivision of Pediatric Cardiology, University of Arkansas for Medical Sciences,

Little Rock, Ark; bDivision of Biostatistics, Department of Pediatrics, University of Supplemental material is available online.
Arkansas for Medical Sciences, Little Rock, Ark; cDivision of Cardiology, Depart-
ment of Pediatrics, Rady Children’s Hospital, San Diego, Calif; dDivision of Pedi-
atric Critical Care, Department of Pediatrics, Seattle Children’s Hospital,
University of Washington School of Medicine, Seattle, Wash; eVirtual PICU Sys-
tems, LLC, Los Angeles, Calif; fDivision of Critical Care Medicine, Department of
Pediatrics, Medical College of Wisconsin, Milwaukee, Wis; and gDivision of Crit- Although the postoperative management of children
ical Care Medicine, Department of Pediatrics and Anesthesiology, Children’s Hos- undergoing heart surgery has changed significantly over
pital Los Angeles, USC Keck School of Medicine, Los Angeles, Calif. the years, the concept of early extubation in the operating
Data were provided by Virtual PICU Systems, LLC. No endorsement or editorial re-
striction of the interpretation of these data or opinions of the authors has been room (OR) after heart surgery is still in evolution. Some
implied or stated. institutions extubate even the most vulnerable patients in
Received for publication April 29, 2015; revisions received Sept 8, 2015; accepted for the OR after their cardiac surgery.1-5 Others continue
publication Sept 12, 2015.
Address for reprints: Punkaj Gupta, MBBS, University of Arkansas for Medical mechanical ventilation (MV) postoperatively in the
Sciences, College of Medicine, Arkansas Children’s Hospital, 1 Children’s Way, intensive care unit (ICU) due to risks associated with
Slot 512-3, Little Rock, AR 72223 (E-mail: [email protected]). untimely extubation.6-8 Identifying which children can be
0022-5223/$36.00
Copyright Ó 2015 by The American Association for Thoracic Surgery successfully removed from MV in the OR after cardiac
http://dx.doi.org/10.1016/j.jtcvs.2015.09.080 surgery is a complex decision process. Research that

The Journal of Thoracic and Cardiovascular Surgery c Volume -, Number - 1

 Congenital Heart Disease Gupta et al

Patient Population
Abbreviations and Acronyms This analysis focused on 5 years (2009-2014) of de-identified data
available in the VPS database. Patients younger than 18 years who
CPB ¼ cardiopulmonary bypass underwent operations (with or without cardiopulmonary bypass) for
ECMO ¼ extracorporeal membrane congenital heart disease at one of the participating PICUs in the VPS
oxygenation database were included. Cases were classified on the basis of the first
ICU ¼ intensive care unit cardiovascular operation of each hospital admission (the index operation).
MV ¼ mechanical ventilation The resultant patient population included 40,903 patients from 64 centers.
Patients undergoing surgical closure of an isolated patent ductus arteriosus
OR ¼ operating room (n ¼ 360), and patients with no documented index operation (n ¼ 4224)
PIM ¼ Pediatric Index of Mortality were excluded from the study population. Patients whose operation was
PRISM ¼ Pediatric Risk of Mortality not classified into one of the Society of Thoracic Surgeons–European
STS-EACTS ¼ Society of Thoracic Surgeons– Association for Cardiothoracic Surgery (STS-EACTS) Mortality
European Association for Categories (category 1, lowest mortality risk; category 5, highest mortality
risk) (n ¼ 8921) were also excluded.10 The study exclusions applied in our
Cardiothoracic Surgery study were similar to other studies in pediatric cardiac surgical population
from STS database.11,12 The final study population therefore included
27,398 patients from 62 centers (Figure 1).
unravels the factors contributing to the risk of extubation For study purposes, the study population was divided into 2 groups:
failure could affect outcomes and reduce costs. patients extubated in the OR, and patients arriving intubated in the ICU.
Previous studies evaluating early extubation (and success Reintubation was defined as need for MV after initial extubation attempt
associated with early extubation) consisted primarily of either in the OR or in the ICU during the same hospital stay. Data on
demographics, patient diagnoses, conventional MV, severity of illness
single-center reports in selected patient populations.1-5
scores, and outcomes were collected. Specific demographic and severity
Some of these studies excluded young patients, patients of illness data included age, gender, developmental disorder, failure to
undergoing complex surgical repair, and patients requiring thrive, genetic disorder, low birth weight, Pediatric Index of Mortality 2
prolonged cardiopulmonary bypass a priori from their (PIM-2) score,13 and Pediatric Risk of Mortality III (PRISM III) score.14
CHD

study population.9 For this reason, there is limited informa- Data were also collected on the use of cardiopulmonary bypass (CPB)
pump during the cardiac surgery, use of extracorporeal membrane
tion in a multi-institutional patient population involving
oxygenation (ECMO) after cardiac surgery, heart transplantation, ICU
children with varying surgical complexity and across centers length of stay, and ICU mortality. Patients were grouped into 2 surgical
with varied surgical volume. risk categories on the basis of the 5 STS-EACTS Mortality categories
To address these knowledge gaps, this study was (with categories 1, 2, and 3 classified as ‘‘low’’ risk and categories 4 and
designed to evaluate the prevalence of and risk factors 5 classified as ‘‘high’’ risk).10,11 Patients receiving orthotopic heart
transplantation with or without congenital heart surgery were classified
associated with the need for MV following cardiac surgery
in the heart transplantation category. Center characteristics were also
in a large, multi-institutional patient population using the collected, including presence of dedicated cardiovascular ICU, annual
Virtual PICU Systems, LLC (VPS) database, a pediatric discharges per center, and annual cardiac surgery cases per year.
critical care network of a diverse set of hospitals caring
for critically ill children. A secondary aim of this study Statistical Analysis
was to evaluate the prevalence of and risk factors for Descriptive statistics were expressed as median (first quartile, third
reintubation after the initial extubation attempt either in quartile) for continuous variables, and percentage (frequency) for
the OR or in the ICU. categorical variables. The distributions of continuous variables of 2 groups
(with and without MV) were compared by using the Kruskal-Wallis tests,
whereas the proportions of categorical variables were compared using a c2
MATERIALS AND METHODS test. Multivariable mixed effects logistic regression models were fitted to
Data Source identify factors likely associated with MV and reintubation (both binary
The VPS is an online pediatric critical care network formed by NACHRI outcomes) after adjustment for other risk factors. Model goodness-of-fit
(National Association of Children’s Hospitals and Related Institutions, now was assessed by using the ratio of the generalized c2 statistic to its degrees
part of Children’s Hospital Association), Children’s Hospital Los Angeles, of freedom and its discriminative ability was measured using the
and Children’s Hospital of Wisconsin to develop a Web-based database with C-statistic.
prospective data collection using standardized clinical data definitions, data We used bootstrap bagging to identify factors associated with each of
quality control, and data analysis. The VPS database is a prospective the 2 outcomes. We generated 199 bootstrap samples of the dataset with
observational cohort of consecutive pediatric ICU (PICU) admissions replacement from the original data and fit multivariable models on each
from a diverse set of hospitals caring for children in the United States. of these simulated datasets. The bootstrap estimates of the odds ratios
Data are collected and entered by trained individuals. VPS performs initial were obtained by aggregating (mean) the estimates obtained from these
and quarterly interrater reliability (IRR) testing. The IRR concordance in 199 bootstrap samples. The 95% bootstrap confidence intervals for the
the VPS database is consistently above 95%. Extensive quality validation odds ratios were obtained by obtaining the 2.5th and the 97.5th percentiles
is performed by VPS staff before release of the data for analysis. To of the bootstrap distribution.15
participate in the VPS database, the individual institutions require All predictor variables were checked for multicollinearity. PRISM III
institutional review board (IRB) approval. The University of Arkansas for was significantly correlated with PIM 2 score (r ¼ 0.42, P <.0001), and
Medical Sciences IRB for the protection of human subjects reviewed the was therefore not included in the model. Heart transplant, arrhythmia,
study protocol that involves querying de-identified patient data and deter- cardiac bypass (associated with high complexity cases, P < .0001),
mined that it does not fall under the jurisdiction of the IRB review process. developmental disorder, and failure to thrive (associated with genetic

2 The Journal of Thoracic and Cardiovascular Surgery c - 2015

Gupta et al Congenital Heart Disease

FIGURE 1. Study population. STS-EACTS, Society of Thoracic Surgeons–European Association for Cardiothoracic Surgery; PDA, patent ductus
arteriosus; ICU, intensive care unit.

CHD
syndrome, P<.0001) were thus not included in either of the multivariable ICU after their heart operation. The median duration of
models. The final model for MV included the following patient-specific MV among children arriving intubated in the ICU was
variables: age at surgery (months), gender (male/female), weight-for-age
z-score, PIM-2 score, low birth weight, genetic syndrome, high-
2 days (interquartile range [IQR] 1-5). Overall mortality
complexity cases, acute lung injury, pulmonary hypertension, and more for the study population was 743 patients (3%). Most study
than one surgery. In addition to these variables, the model for reintubation patients were located in a dedicated cardiac ICU (61%,
included duration of first MV (days), high-frequency oscillatory ventila- 16,767 patients). The annual center volume for the study
tion, ECMO, cardiac arrest, chylothorax, diaphragm, renal failure, population was 152 cardiac surgery cases per center (IQR
seizures, sepsis, and brain hemorrhage. Presence of a dedicated cardiac
ICU and average annual cardiac surgery cases per center were the 2
47-299).
center-specific variables included in both the models. Restricted cubic Of the study patients, 2449 (9%) required reintubation
splines were used to model all continuous variables to allow for nonlinear after the initial extubation attempt either in the OR or in
association of the variable on the log odds of the outcome. A generalized the ICU during the same hospital stay. The median time
estimating equations type analysis was performed to take into account to reintubation was 1 day (IQR 0-3) after the initial
the clustering of patients into different centers.
We compared incidence of MV and reintubation among patients with
extubation attempt. Of the 6810 patients extubated in the
and without a dedicated cardiac ICU by using propensity score matching. OR, 395 (6%) required reintubation in the ICU. In contrast,
First the probability of a patient being seen in a center with a dedicated 2054 patients (10%) required reintubation among the
cardiac ICU was obtained by using a logistic regression model as a function remaining 20,588 patients requiring MV postoperatively
of patient- and center-level risk factors used in the multivariable models in the ICU. The incidence of reintubation was higher among
discussed. Then, each patient in the dedicated cardiac ICU group was
1:1 matched with a patient in the no cardiac ICU group based on the logit
patients undergoing high-complexity operations compared
of the propensity score. The caliper used in the matching was chosen such with patients undergoing low-complexity operations
that all variables used in the matching process were balanced between the 2 (18% vs 6%, P<.0001) (Figure 1). The median ICU length
groups (standardized difference of 0.10 or less).16 Incidence of MV and of stay in children requiring reintubation was 22 days (IQR
reintubation were compared between the 2 matched groups using the 11-47) compared with 4 days (IQR 3-8) among patients not
McNemar test for paired dichotomous outcomes. Statistical software
used for analysis included R 3.0.2 (R Foundation for Statistical Computing,
requiring reintubation during their hospital stay. The
Vienna, Austria) and SAS 9.4 (SAS Institute, Inc, Cary, NC). All tests were mortality was higher among patients requiring reintubation
2-sided assuming a significance level of 5%. compared with patients who were not reintubated during
their hospital stay (11% vs 2%). The proportion of patients
reintubated increased with increasing ICU length of stay
RESULTS (Figure 2).
A total of 27,398 patients from 62 centers were included. Patient characteristics, center data, and unadjusted
Of these, 6810 patients (25%) were extubated in the OR, outcomes among the patients extubated in the OR and
whereas 20,588 patients (75%) arrived intubated in the patients arriving intubated in the ICU are displayed in

The Journal of Thoracic and Cardiovascular Surgery c Volume -, Number - 3

 Congenital Heart Disease Gupta et al
CHD

FIGURE 2. Prevalence of reintubation with increasing length of ICU stay among the study patients. ICU, Intensive care unit; OR, operating room.

Table 1. Patients who arrived intubated in the ICU were cardiac ICU (Table 3). Independent risk factors associated
younger, smaller, more likely to have preoperative risk with the need for reintubation after initial extubation
factors, and had a higher PIM-2 score and a higher PRISM included younger age, lower weight-for-age z-score,
III score. Patients arriving intubated in the ICU underwent presence of genetic disorder, pulmonary hypertension,
more complex operations, and were associated with higher having received operations of higher complexity, need for
need for reoperation and increased use of ECMO after reoperation, acute lung injury, higher PIM-2 score, presence
cardiac surgery. Patients needing intubation after cardiac of postoperative complications (such as cardiac arrest,
surgery were associated with higher incidence of acute chylothorax, diaphragm paralysis, seizures, sepsis), use of
lung injury, pulmonary hypertension, cardiac arrest, high-frequency oscillatory ventilation, use of ECMO,
chylothorax, diaphragm paralysis, seizures, and sepsis. extubation in the OR, longer duration of MV before initial
The unadjusted mortality was higher among patients extubation, and presence of a dedicated cardiac ICU
arriving intubated in the ICU compared with patients (Table 4). The ratio of the generalized c2 statistic and its
extubated in the OR. degrees of freedom were 1.28 and 0.90 respectively for the
Table 2 demonstrates the prevalence of MV and 2 models predicting MV and reintubation. The C-statistic
reintubation among the patients extubated in the OR and pa- for the 2 generalized linear mixed effects regression models
tients arriving intubated in the ICU for 8 STS benchmark were 0.89 and 0.84, respectively, indicating very good
operations. The highest prevalence of extubation in the discriminative and predictive ability for the model.
OR was among the patients undergoing Fontan operation Furthermore, we evaluated the association of a dedicated
and ventricular septal defect (VSD) closure. The highest cardiac ICU with need for MV and need for reintubation.
prevalence of reintubation was noted among the patients Propensity models for the need for MV and need for
undergoing the Norwood operation and truncus arteriosus reintubation are depicted in Tables E1 and E2. In matched
repair. sample, the prevalence of MV and reintubation was lower
After adjusting for patient characteristics, surgical among the centers with a dedicated cardiac ICU (cardiac
complexity, and center volume, independent risk factors ICU vs other ICU, MV: 78% vs 88%, reintubation: 7%
associated with need for MV included younger age, lower vs 10%) (Table E3).
weight-for-age z-score, low birth weight, presence of
genetic disorder, pulmonary hypertension, having received DISCUSSION
operations of higher complexity, need for reoperation, acute Data from this large, multicenter study establish that
lung injury, higher PIM-2 score, and presence of a dedicated approximately one-quarter of patients undergoing heart

4 The Journal of Thoracic and Cardiovascular Surgery c - 2015

Gupta et al Congenital Heart Disease

TABLE 1. Patient characteristics

Variable Number of observations Extubated in OR (n ¼ 6810) Arrived intubated in ICU (n ¼ 20,588) P value
Baseline characteristics
Age at surgery (mo) 27,398 41.3 (9.0, 96.4) 5.3 (0.7, 29.7) <.001
Male gender 27,398 3697 (54.3%) 11,399 (55.4%) .12
Weight-for-age Z-score 27,046 0.6 (1.5, 0.3) 1.1 (2.0, 0.1) <.001
Developmental disorder 27,396 258 (3.8%) 528 (2.6%) <.001
Failure to thrive 27,396 469 (6.9%) 1971 (9.6%) <.001
Genetic disorder 27,396 905 (13.3%) 4040 (19.6%) <.001
Low birth weight 27,396 155 (2.3%) 1281 (6.2%) <.001
PIM-2 score 27,398 5.3 (5.7, 4.4) 3.7 (4.2, 2.9) <.001
PRISM III score 25,106 5 (2, 8) 7 (3, 10) <.001
Use of HFOV 27,398 10 (0.1%) 178 (0.9%) <.001
Operation details
CPB case 27,398 4908 (72.1%) 13,869 (67.4%) <.001
Need for reoperation 27,398 259 (3.8%) 3536 (17.2%) <.001
Use of ECMO 27,398 49 (0.7%) 919 (4.5%) <.001
Heart transplantation 27,398 96 (1.4%) 404 (2%) <.001
STS-EACTS surgical risk category 27,398
1 3395 (49.9%) 5986 (29.1%)
2 2308 (33.9%) 5270 (25.6%)
3 419 (6.2%) 2990 (14.5%)
4 629 (9.2%) 5082 (24.7%)
5 59 (0.9%) 1260 (6.1%)

CHD
Diagnoses
Arrhythmias 27,396 1335 (19.6%) 3268 (15.9%) <.001
Acute lung injury 27,396 859 (12.6%) 6813 (33.1%) <.001
Pulmonary hypertension 27,396 131 (1.9%) 1265 (6.1%) <.001
Cardiac arrest 27,396 55 (0.8%) 631 (3.1%) <.001
Chylothorax 27,396 42 (0.6%) 301 (1.5%) <.001
Diaphragm paralysis 27,396 95 (1.4%) 495 (2.4%) <.001
Use of hemodialysis catheter 27,398 10 (0.1%) 84 (0.4%) <.001
Renal failure 27,396 57 (0.8%) 748 (3.6%) <.001
Seizures 27,396 160 (2.4%) 586 (2.8%) .02
Sepsis 27,396 26 (0.4%) 256 (1.2%) <.001
Brain hemorrhage 27,396 13 (0.2%) 116 (0.6%) <.001
Unadjusted outcomes
ICU mortality 27,398 44 (0.6%) 699 (3.4%) <.001
ICU length of stay 27,398 3 (2, 5) 6 (3, 12) <.001
Center data <.001
Dedicated cardiac ICU 27,398 4706 (69.1%) 12,061 (58.6%) <.001
Annual cardiac surgery per center 27,397 309 (211, 571) 295 (185, 525) <.001
Continuous variables are summarized by the triplet of quartiles 50th (25th, 75th). Categorical variables are summarized as n (%). OR, Operating room; ICU, intensive care unit;
PIM, Pediatric Index of Mortality; PRISM, Pediatric Risk of Mortality; HFOV, high-frequency oscillatory ventilation; CPB, cardiopulmonary bypass; ECMO, extracorporeal
membrane oxygenation; STS-EACTS, Society of Thoracic Surgeons–European Association for Cardiothoracic Surgery.

operations are extubated in the OR, whereas approximately Although the ideal approach for patients undergoing car-
three-quarters of patients arrive intubated postoperatively in diac surgery should be extubation in the OR, this approach
the ICU. In addition, approximately 9% of patients require may be associated with some grave consequences.1,2 Some
reintubation after their initial extubation attempt. These of these consequences include significant hemodynamic
data suggest that both odds of MV, and odds of reintubation instability, sudden cardiac arrest, massive catecholamine
after cardiac surgery are a function of patient characteris- surge, and pulmonary hypertensive crisis, unnecessary
tics, surgical risk category, and center volume. Furthermore, airway trauma, increased risk for nosocomial infections, and
we demonstrated that patients arriving intubated in the ICU prolonged duration of MV and ICU length of stay.6-8,17,18
after cardiac surgery are associated with increased ICU Our study, therefore, identified risk factors associated with
mortality and increased ICU length of stay. the need of MV outside the OR. In our multi-institutional

The Journal of Thoracic and Cardiovascular Surgery c Volume -, Number - 5

 Congenital Heart Disease Gupta et al

TABLE 2. Prevalence of mechanical ventilation and reintubation in 8 STS benchmark operations

Extubated in operating room Arrived intubated in ICU
Variable Total patients No. of patients Reintubation No. of patients Reintubation
VSD repair 1330 303 (22.8) 4 (1.3) 1027 (77.2) 41 (4.0)
TOF repair 463 66 (14.2) 6 (9.1) 397 (85.8) 26 (6.6)
CAVC repair 313 25 (8.0) 6 (24.0) 288 (92.0) 38 (13.2)
ASO 692 22 (3.2) 8 (36.4) 670 (96.8) 44 (6.6)
ASO þ VSD repair 288 9 (3.1) 2 (22.2) 279 (96.9) 33 (11.8)
Fontan 1494 691 (46.2) 31 (4.5) 803 (53.8) 33 (4.1)
Truncus arteriosus repair 212 10 (4.7) 3 (30.0) 202 (95.3) 37 (18.3)
Norwood operation 1078 26 (2.4) 13 (50.0) 1052 (97.6) 287 (27.3)
Categorical variables are summarized as n (%). ICU, Intensive care unit; VSD, ventricular septal defect; TOF, tetralogy of Fallot; CAVC, complete atrioventricular canal;
ASO, arterial switch operation.

cohort, younger, smaller patients with increased preoperative than the other ICU models. Better outcomes in the cardiac
risk factors, patients associated with higher severity of illness ICU also may be related to other factors, such as training
scores, and patients undergoing high-complexity operations and availability of personnel and use of standardized
were associated with higher likelihood for MV outside the management protocols.19,20
OR. Most of these factors are consistent with other single- There are both advantages and disadvantages of
institution studies in which younger and smaller patients continuing MV outside the OR after pediatric cardiac
with complex repairs required MV after cardiac surgery.1,2
A primary determinant of extubation success is the need TABLE 4. Risk factors associated with need for reintubation after
for reintubation. Reintubation rates after pediatric cardiac pediatric heart surgery in multivariable models
CHD

surgery vary from 6% to 27% in the existing literature1,6-8; Odds ratio P

however, these studies were conducted among select Variable Comparison (95% CI) value
patient populations in single institutions. In contrast, our Male gender Yes vs No 1.04 (0.93-1.16) .55
study demonstrated a reintubation rate of 9% involving Age 12-mo 0.89 (0.86-0.92) <.0001
patients with varying surgical complexity and across increase
centers of varied surgical volume. In our study, centers Weight-for-age Z-score 1 SD increase 0.92 (0.88-0.96) <.0001
with a dedicated cardiac ICU were associated with lower Genetic disorder Yes vs No 1.45 (1.30-1.62) <.0001
odds of MV and lower odds of reintubation. It is possible Low birth weight Yes vs No 0.98 (0.78-1.21) .91
that the multidisciplinary approach in the cardiac ICU Pulmonary hypertension Yes vs No 1.31 (1.08-1.63) .005
caters to the postoperative needs of these patients better High complexity operations Yes vs No 1.64 (1.44-1.85) <.0001
Need for reoperation Yes vs No 4.54 (3.94-5.28) <.0001
PIM-2 score 0.5-unit 1.17 (1.14-1.21) <.0001
TABLE 3. Risk factors associated with need for mechanical increase
ventilation after pediatric heart surgery in multivariable models Cardiac arrest Yes vs No 1.70 (1.36-2.19) <.0001
Acute lung injury Yes vs No 1.64 (1.49-1.79) <.0001
Odds ratio P
Chylothorax Yes vs No 2.22 (1.64-3.05) <.0001
Variable Comparison (95% CI) value
Diaphragm paralysis Yes vs No 4.69 (3.88-5.72) <.0001
Male gender Yes vs No 1.01 (0.93-1.10) .87
Brain hemorrhage Yes vs No 0.88 (0.42-1.45) .54
Age 12-mo increase 0.84 (0.82-0.85) <.0001 Renal failure Yes vs No 1.06 (0.82-1.32) .79
Weight-for-age Z-score 1 SD increase 0.95 (0.93-0.98) <.0001 Seizures Yes vs No 1.83 (1.41-2.32) <.0001
Genetic disorder Yes vs No 1.26 (1.13-1.41) <.0001 Sepsis Yes vs No 2.35 (1.47-3.34) <.0001
Low birth weight Yes vs No 2.11 (1.59-2.75) <.0001
Use of HFOV Yes vs No 27.17 (17.73-48.18) <.0001
Pulmonary hypertension Yes vs No 1.77 (1.30-2.40) <.0001 Use of ECMO Yes vs No 0.96 (0.75-1.17) .65
High-complexity operations Yes vs No 1.96 (1.73-2.24) <.0001 Extubated in operating Yes vs No 2.76 (2.31-3.23) <.0001
Need for reoperation Yes vs No 1.55 (1.24-1.91) <.0001
room
Acute lung injury Yes vs No 2.54 (2.26-2.85) <.0001 Duration of MV 2-d increase 1.04 (1.02-1.06) <.0001
PIM-2 score 0.6-unit 1.67 (1.62-1.81) <.0001 Annual cardiac surgery 100-case 1.03 (0.99-1.08) .16
increase
per center increase
Annual cardiac surgery 100-case 1.02 (0.99-1.05) .09 Cardiac ICU Yes vs No 0.78 (0.66-0.90) .005
per center increase
High-complexity operations: Society of Thoracic Surgeons–European Association
Cardiac ICU Yes vs No 0.47 (0.41-0.53) <.0001 for Cardiothoracic Surgery surgical risk categories 4 and 5. CI, Confidence interval;
High-complexity operations: Society of Thoracic Surgeons–European Association PIM-2, Pediatric Index of Mortality; HFOV, high-frequency oscillatory ventilation;
for Cardiothoracic Surgery surgical risk categories 4 and 5. CI, Confidence interval; ECMO, extracorporeal membrane oxygenation; MV, mechanical ventilation; ICU,
SD, standard deviation; PIM-2, Pediatric Index of Mortality; ICU, intensive care unit. intensive care unit.

6 The Journal of Thoracic and Cardiovascular Surgery c - 2015

Gupta et al Congenital Heart Disease

surgery. Sudden removal of positive pressure ventilation on CONCLUSIONS

extubation immediately after surgery in children with Large variations were demonstrated in clinical practices
compromised hemodynamics can dramatically worsen the of MV following pediatric cardiac surgery among patients
ongoing low cardiac output state and precipitate sudden with varying surgical complexity and across centers with
cardiac arrest. Furthermore, extubation-related stress varied surgical volume. Our study establishes that
response causes a massive catecholamine surge that can approximately one-quarter of patients undergoing heart
initiate life-threatening arrhythmias or precipitate a operations are extubated in the OR, whereas approximately
pulmonary hypertensive crisis, thereby leading to sudden three-quarters of patients are arriving intubated postopera-
cardiovascular collapse. On the other hand, potential tively in the ICU. In addition, approximately 9% of patients
advantages of early extubation include less ventilator- require reintubation after their initial extubation attempt.
associated complications, such as pulmonary infections Both odds of MV and odds of reintubation after cardiac
and atelectasis, less airway irritation, less barotrauma surgery are a function of patient characteristics, surgical
secondary to positive airway pressure ventilation, reduced risk category, and center volume. These data further suggest
parental stress, reduced requirements for sedatives, more that extubation in the OR can be done successfully with a
rapid patient mobilization, and less resource utilization, low complication rate. Using databases such as the VPS
decreased length of ICU and hospital stay, and reduced for clinical outcomes research may in the future decrease
costs.1-6 the cost of discovery and guide us in improving outcomes
Our study has several potential limitations. This study is for critically ill patients.
subject to the limitations of all observational analyses,
including selection bias, residual confounding, and mea- Conflict of Interest Statement
surement error. Other limitations of the present study are Authors have nothing to disclose with regard to commercial
related to the nature of the VPS dataset. Not all US pediatric support.
hospitals participate in the VPS database. Nonetheless, the

CHD
present report represents the most inclusive evaluation, with The authors thank the Office of Grants and Scientific Publica-
data from 62 pediatric heart centers. We were also limited to tions for their editorial assistance.
consideration of variables collected in the VPS database.
Although we attempted to adjust for important patient References
confounders, other unmeasured confounders could have 1. Harris KC, Holowachuk S, Pitfield S, Sanatani S, Froese N, Potts JE, et al. Should
early extubation be the goal for children after congenital cardiac surgery? J
affected our analysis. As such, we could not evaluate or Thorac Cardiovasc Surg. 2014;148:2642-7.
account for the potential impact of other variables (such 2. Howard F, Brown KL, Garside V, Walker I, Elliott MJ. Fast-track paediatric
as hospital structure and process of care measures, training cardiac surgery: the feasibility and benefits of a protocol for uncomplicated
cases. Eur J Cardiothorac Surg. 2010;37:193-6.
or availability of ICU personnel, or nursing factors) on our 3. Davis S, Worley S, Mee RBB, Harrison AM. Factors associated with early
evaluation of odds for MV. Due to database limitation, our extubation after cardiac surgery in young children. Pediatr Crit Care Med.
study also lacked data on number of reintubations 2004;5:63-8.
4. Vricella LA, Dearani JA, Gundry SR, Razzouk AJ, Brauer SD, Bailey LL. Ultra
performed in the OR. Our database also lacks reasons for fast track in elective congenital cardiac surgery. Ann Thorac Surg. 2000;69:
reintubations for the study patients. 865-71.
Our study lacked data on certain key patient and treat- 5. Laussen PC, Roth SJ. Fast tracking: efficiently and safely moving patients
through the intensive care unit. Prog Pediatr Cardiol. 2003;18:149-58.
ment variables, such as PaO2 (partial pressure of oxygen 6. Gupta P, McDonald R, Gossett JM, Butt W, Shinkawa T, Imamura M, et al. A
in arterial blood), PaCO2 (partial pressure of carbon dioxide single center experience of extubation failure in infants undergoing the Norwood
in arterial blood), FiO2 (fraction of inspired oxygen), PaO2/ operation. Ann Thorac Surg. 2012;94:1262-8.
7. Gupta P, McDonald R, Goyal S, Gossett JM, Imamura M, Agarwal A, et al.
FiO2 ratio, presence of focal versus diffuse lung disease, use Extubation failure in infants with shunt-dependent pulmonary blood flow and
of nitric oxide, and presence of air leak that could have been univentricular physiology. Cardiol Young. 2013;8:1-9.
used in the multivariable modeling. Our study also lacked 8. Gupta P, Chow V, Gossett JM, Yeh JC, Roth SJ. Incidence, predictors and
outcomes of extubation failure in children after orthotopic heart transplantation:
data on MV, such as plateau, mean, and end expiratory pres- a single-center experience. Pediatr Cardiol. 2015;36:300-7.
sures, data on inotropes, and sedatives and neuromuscular 9. Mittnacht AJC, Thanjan M, Srivastava S, Joashi U, Bodian C, Hossain S, et al.
blocking agents that could have potentially affected Extubation in the operating room after congenital heart surgery in children.
J Thorac Cardiovasc Surg. 2008;136:88-93.
both need of MV and need for reintubation. Due to the 10. O’Brien SM, Clarke DR, Jacobs JP, Jacobs ML, Lacour-Gayet FG, Pizarro C,
retrospective, multi-institutional nature of this study, we et al. An empirically based tool for analyzing mortality associated with
did not consider the role of difficulties from mechanical congenital heart surgery. J Thorac Cardiovasc Surg. 2009;138:1139-53.
11. Gupta P, Jacobs JP, Pasquali SK, Hill KD, Gaynor JW, O’Brien SM, et al.
ventilator weaning in the group with failed extubation. Epidemiology and outcomes following in-hospital cardiac arrest after pediatric
Due to the large sample size, some differences that are cardiac surgery. Ann Thorac Surg. 2014;98:2138-44.
not clinically significant may appear to be statistically 12. Welke KF, O’Brien SM, Peterson ED, Ungerleider RM, Jacobs ML, Jacobs JP.
The complex relationship between pediatric cardiac surgical case volumes and
significant in univariable analysis; however, the effects of mortality rates in a national clinical database. J Thorac Cardiovasc Surg.
these variables were addressed in the multivariable analysis. 2009;137:1133-40.

The Journal of Thoracic and Cardiovascular Surgery c Volume -, Number - 7

 Congenital Heart Disease Gupta et al

13. Slater A, Shann F, Pearson G, for the PIM Study Group. PIM2: a 18. Harrison AM, Cox AC, Davis S, Piedmonte M, Drummond-Webb JJ, Mee RB.
revised version of the paediatric index of mortality. Intensive Care Med. 2003; Failed extubation after cardiac surgery in young children: prevalence, pathogen-
29:278-85. esis, and risk factors. Pediatr Crit Care Med. 2002;3:148-52.
14. Pollack MM, Patel KM, Ruttimann UE. PRISM III: an updated pediatric risk of 19. Burstein DS, Jacobs JP, Li JS, Sheng S, O’Brien SM, Rossi AF, et al. Care models
mortality score. Crit Care Med. 1996;24:743-52. and associated outcomes in congenital heart surgery. Pediatrics. 2011;127:
15. Carpenter J, Bithell J. Bootstrap confidence intervals: when, which, what? A e1482-9.
practical guide for medical statisticians. Stat Med. 2000;19:1141-64. 20. Gupta P, Beam BW, Noel TR, Dvorchik I, Yin H, Simsic JM, et al. Impact of
16. Austin PC. Balance diagnostics for comparing the distribution of baseline pre-operative location on outcomes in congenital heart surgery. Ann Thorac
covariates between treatment groups in propensity-score matches sample. Stat Surg. 2014;98:896-903.
Med. 2009;28:3083-107.
17. Kanter RK, Bove EL, Tobin JR, Zimmerman JJ. Prolonged mechanical
ventilation of infants after open heart surgery. Crit Care Med. 1986;14: Key Words: mechanical ventilation, pediatric cardiac sur-
211-4. gery, reintubation, surgical complexity, center volume
CHD

8 The Journal of Thoracic and Cardiovascular Surgery c - 2015

Gupta et al Congenital Heart Disease

APPENDIX 1. SUPPLEMENTARY ONLINE Table E1. Propensity model for risk factors associated
CONTENT with need for mechanical ventilation based on loca-
Gupta P, Rettiganti M, Gossett JM, Yeh JC, Jeffries HE, tion of the patient ...............8.e2
Rice TB, Wetzel RC. Risk factors for mechanical Table E2. Propensity model for risk factors associated
ventilation and reintubation after pediatric heart surgery. with need for reintubation based on location of
the patient................8.e3
TABLE OF CONTENTS Table E3. Study outcomes based on location of patients
Figure E1. Center Volume and proportion of patients after cardiac surgery...............8.e3
receiving mechanical ventilation for the VPS study
center.................... 8.e2

CHD

The Journal of Thoracic and Cardiovascular Surgery c Volume -, Number - 8.e1

 Congenital Heart Disease Gupta et al

FIGURE E1. Center volume and proportion of patients receiving mechanical ventilation for the Virtual PICU Systems, LLC, study centers.
CHD

TABLE E1. Propensity model for risk factors associated with need for mechanical ventilation based on location of the patient
Unmatched Matched
Cardiac ICU Other ICU Cardiac Other ICU
Variable (n ¼ 16,767) (n ¼ 10,631) Std Diff ICU (n ¼ 3634) (n ¼ 3634) Std Diff
Age at surgery (mo) 36.6 (54.8) 38.8 (56.6) 0.04 37.3 (56.7) 35.1 (53.7) 0.03
Male gender 9205 (54.9%) 5891 (55.4%) 0.01 2022 (55.6%) 2038 (56.1%) 0.008
Weight-for-age Z-score 0.9 (1.4) 0.9 (1.4) 0.001 0.9 (1.4) 1.0 (1.4) 0.01
Developmental disorder 471 (2.8%) 315 (3%) 0.01 109 (3%) 90 (2.5%) 0.03
Failure to thrive 1601 (9.5%) 839 (7.9%) 0.05 210 (5.8%) 232 (6.4%) 0.02
Genetic disorder 3039 (18.1%) 1906 (17.9%) 0.005 659 (18.1%) 656 (18.1%) 0.002
Low birth weight 1008 (6%) 428 (4%) 0.09 148 (4.1%) 145 (4%) 0.004
Acute Lung Injury 3951 (23.6%) 3721 (35%) 0.25 1407 (38.7%) 1467 (40.4%) 0.03
Pulmonary hypertension 686 (4.1%) 710 (6.7%) 0.11 189 (5.2%) 210 (5.8%) 0.02
PIM-2 score 3.8 (1.3) 3.8 (1.3) 0.05 3.7 (1.3) 3.6 (1.1) 0.02
High-complexity operations 4413 (26.3%) 2617 (24.6%) 0.03 941 (25.9%) 1027 (28.3%) 0.05
Heart transplantation 443 (2.6%) 57 (0.5%) 0.16 69 (1.9%) 54 (1.5%) 0.03
Need for reoperation 2242 (13.4%) 1553 (14.6%) 0.03 601 (16.5%) 693 (19.1%) 0.06
Annual cardiac surgery per center 442 (146) 179 (110) 2.02 277 (30) 273 (48) 0.09
ICU, Intensive care unit; PIM-2, Pediatric Index of Mortality; Std Diff, standardized difference.

8.e2 The Journal of Thoracic and Cardiovascular Surgery c - 2015

Gupta et al Congenital Heart Disease

TABLE E2. Propensity model for risk factors associated with need for reintubation based on location of the patient
Unmatched Matched
Cardiac ICU Other ICU Cardiac Other ICU
Variable (n ¼ 16,767) (n ¼ 10,631) Std Diff ICU (n ¼ 3638) (n ¼ 3638) Std Diff
Age at surgery (mo) 36.6 (54.8) 38.8 (56.6) 0.03 36.5 (55.4) 36.7 (54.8) 0.004
Male gender 9205 (54.9%) 5891 (55.4%) 0.01 2021 (55.1%) 2045 (55.8%) 0.01
Weight-for-age Z-score 0.9 (1.4) 0.96 (1.5) 0.001 0.99 (1.4) 1.00 (1.5) 0.006
Development disorder 471 (2.8%) 315 (3%) 0.009 101 (2.8%) 89 (2.4%) 0.02
Failure to thrive 1601 (9.5%) 839 (7.9%) 0.05 237 (6.5%) 232 (6.3%) 0.01
Genetic disorder 3039 (18.1%) 1906 (17.9%) 0.005 659 (18%) 662 (18.1%) 0.003
Low birth weight 1008 (6%) 428 (4%) 0.09 142 (3.9%) 134 (3.7%) 0.01
Pulmonary hypertension 686 (4.1%) 710 (6.7%) 0.11 177 (4.8%) 192 (5.2%) 0.02
PIM-2 score 3.8 (1.3) 3.88 (1.4) 0.06 3.74 (1.2) 3.66 (1.5) 0.06
Arrhythmia 3056 (18.2%) 1547 (14.6%) 0.09 633 (17.3%) 613 (16.7%) 0.009
Acute lung injury 3951 (23.6%) 3721 (35%) 0.25 1418 (38.7%) 1473 (40.2%) 0.03
Cardiac arrest 435 (2.6%) 251 (2.4%) 0.01 68 (1.9%) 75 (2%) 0.01
Chylothorax 229 (1.4%) 114 (1.1%) 0.02 16 (0.4%) 18 (0.5%) 0.004
Diaphragm paralysis 407 (2.4%) 183 (1.7%) 0.04 67 (1.8%) 59 (1.6%) 0.01
Renal failure 584 (3.5%) 221 (2.1%) 0.08 69 (1.9%) 80 (2.2%) 0.02
Seizures 499 (3%) 247 (2.3%) 0.04 76 (2.1%) 83 (2.3%) 0.01
Sepsis 177 (1.1%) 105 (1%) 0.007 21 (0.6%) 27 (0.7%) 0.01
Brain hemorrhage 91 (0.5%) 38 (0.4%) 0.02 10 (0.3%) 15 (0.4%) 0.02
High complexity operations 4413 (26.3%) 2617 (24.6%) 0.04 940 (25.6%) 1006 (27.4%) 0.03
Need for reoperation 2242 (13.4%) 1553 (14.6%) 0.03 568 (15.5%) 660 (18%) 0.07

CHD
Heart transplantation 443 (2.6%) 57 (0.5%) 0.17 60 (1.6%) 50 (1.4%) 0.02
Use of ECMO 588 (3.5%) 380 (3.6%) 0.003 150 (4.1%) 146 (4%) 0.002
Use of HFOV 79 (0.5%) 109 (1%) 0.06 20 (0.5%) 23 (0.6%) 0.007
Duration of MV before reintubation 3.73 (9.5) 3.67 (9.1) 0.007 3.89 (7.2) 3.92 (8.7) 0.003
Extubated in OR 4706 (28.1%) 2104 (19.8%) 0.19 667 (18.2%) 568 (15.5%) 0.07
Annual cardiac surgery per center 442 (146) 179 (110) 2.02 276 (30) 272 (50) 0.09
ICU, Intensive care unit; PIM-2, Pediatric Index of Mortality; ECMO, extracorporeal membrane oxygenation; HFOV, high-frequency oscillatory ventilation; MV, mechanical
ventilation; OR, operating room; Std Diff, standardized difference.

TABLE E3. Study outcomes based on location of patients after cardiac surgery
Unmatched Matched*
Cardiac ICU Other ICU Cardiac ICU Other ICU
Variable (n ¼ 16,767) (n ¼ 10,631) P value (n ¼ 3634) (n ¼ 3634) P value
Need for mechanical 12,061 (71.9%) 8527 (80.2%) <.0001 2846 (78.3%) 3199 (88.0%) <.0001
ventilation

Cardiac ICU Other ICU Cardiac ICU Other ICU

(n ¼ 16,767) (n ¼ 10,631) P value (n ¼ 3638) (n ¼ 3638) P value
Need for reintubation 1409 (8.4%) 1040 (9.8%) <.0001 263 (7.2%) 361 (9.9%) <.0001
ICU, Intensive care unit. *Separate matching was performed for need for mechanical ventilation and reintubation. Please refer to Online supplement for variables used in match-
ing for both models.

The Journal of Thoracic and Cardiovascular Surgery c Volume -, Number - 8.e3

 Congenital Heart Disease Gupta et al

000 Risk factors for mechanical ventilation and reintubation after pediatric heart
surgery
Punkaj Gupta, MBBS, Mallikarjuna Rettiganti, PhD, Jeffrey M. Gossett, MS, Justin C. Yeh, MD,
Howard E. Jeffries, MD, MBA, Tom B. Rice, MD, and Randall C. Wetzel, MBBS, Little Rock, Ark,
San Diego and Los Angeles, Calif, Seattle, Wash, Milwaukee, Wis

Although the prevalence of extubation in the operating room is low, it is associated with low
complication rate.
CHD

The Journal of Thoracic and Cardiovascular Surgery c - 2015