Challenging Convention: Daytime Versus Nighttime Extubation in the Pediatric ICU

Setting

The PICU at Children's of Alabama is a 24-bed, tertiary care, mixed medical and surgical unit within a free-standing, academic children's hospital. There is a separate 26-bed stepdown unit staffed by nonintensivist prescribers and a separate 20-bed cardiac ICU. There are approximately 1,600 admissions/y to the PICU, with 44.8% of children requiring mechanical ventilation for a mean duration of 4.3 d in 2019.

There are 4 respiratory therapists (RTs) dedicated solely to the PICU on each shift. Except in cases of extracorporeal membrane oxygenation, continuous renal replacement therapy, organ transplantation, septic shock, and critical airways, the patient-to-nurse ratio is maintained at 2:1. The nursing and RT staffing levels are the same during both the day shift and the night shift. During day shifts, the prescriber team consists of 2 attending intensivists, 1–2 nurse practitioners, 1–2 fellows of varying levels of training, and 4–5 residents. During night shifts, there is at least 1 upper-level PICU fellow (ie, in second or third year of training) or an attending intensivist directing care in the PICU. In cases where an upper-level PICU fellow is present at night, an attending generally takes at-home call. At night, 1–2 residents are on call. Other than a few sporadic events, nurse practitioners did not work night shifts during the study period. There is round-the-clock in-house anesthesia coverage with residents or certified registered nurse anesthetists. The anesthesia attending is available in-house until 23:00 (ie, 11:00 pm) on most days. There is no round-the-clock coverage by in-house otolaryngologists.

Study Design and Subject Selection

This was a secondary, retrospective cohort analysis of the local ventilation liberation quality improvement initiative data. The requirement for informed consent was waived by the institutional review board. The ventilation liberation initiative is an ongoing, multi-phase project aimed at decreasing the average duration of mechanical ventilation for all children admitted to this PICU. Data from the first 4 phases were included. Regardless of phase, inclusion criteria were age < 19 y and the presence of an oral or nasal endotracheal tube. Phases 1–3 focused on the development and implementation of an RT-driven extubation readiness test pathway and included only patients intubated for a primary lower respiratory tract etiology (see the supplementary materials at http://www.rcjournal.com). We have previously published exclusion criteria and methodology for these phases.²⁴ Phase 1 included pre-intervention data collected prior to interventions, which occurred in phases 2 and 3. Phase 2 was the initial introduction of the pathway, and phase 3 was the introduction of a daily rounding tool accompanying the same pathway. The fourth and largest enrollment phase included all patients meeting the aforementioned inclusion criteria with no exclusion criteria (ie, patients intubated for any reason). This fourth phase included both a baseline and an intervention data set. Throughout all phases, extubation decisions were not mandated based on extubation readiness tests results and were left to the discretion of the prescriber team.

Mechanical ventilation encounters were extracted from the data sets for these phases. The following encounters were excluded: palliative extubations, unplanned extubations, and encounters that did not include an extubation attempt (ie, tracheostomy placement or mortality). An individual subject could have multiple ventilator encounters included from the same hospital admission or from different hospital admissions. However, ventilator encounters that immediately followed an extubation failure (ie, within 48 h) were not included.

Data Collection and Definitions

Throughout all 4 phases, data were either collected by or confirmed for accuracy by the principal investigator. The data for phase 1 and pre-intervention data for phase 4 were obtained through retrospective chart review of ventilator encounters identified through local database review. Data for all other phases were collected prospectively.

The following demographic data were collected: age, race, and gender. The following clinical factors were collected: Pediatric Index of Mortality-3 (PIM-3) score, primary indication for intubation, date/time of intubation and extubation, re-intubation events, duration of mechanical ventilation at first extubation for that encounter, length of PICU stay, PICU mortality, cause of extubation failure, and presence or absence of comorbid extubation failure risk factors. Reviewer bias was mitigated through the use of strict data definitions. While there was no standard adjudication process, there were informal consensus discussions between the principal and senior investigator in the case of ambiguity regarding any data elements.

Day shift was defined as 07:00 to 18:59, and night shift as 19:00 to 06:59. The timing of intubation and extubation was abstracted from the respiratory and nursing flowsheets in the electronic medical record (EMR). Extubation failure was defined as replacement of the endotracheal tube ≤ 48 h after extubation. The cause of extubation failure was obtained through direct communication with the prescriber present at the re-intubation. Where this was not possible, the cause of extubation failure was obtained by reviewing the fellow or attending physician documentation in the EMR. Noninvasive ventilation (NIV) was defined as any use of CPAP or bi-level positive airway pressure. High-flow nasal cannula was not classified as NIV, and data on its use were not collected. Subjects on baseline NIV for any length of time during their routine home care were not considered in the NIV use data analysis. Length of PICU stay was defined as the difference between physical arrival to the PICU and the time a transfer order was entered. Comorbid extubation failure risk factors were evidence-based and included age ≤ 24 months, syndromic or dysgenetic comorbidity, chronic neurologic comorbidity, chronic respiratory comorbidity, acute or chronic airway problem, acute surgical airway problem, or chronic NIV use.²² The presence or absence of these risk factors was determined by the principal investigator and adjudicated with the senior investigator when necessary.

In cases of extubation failure, re-intubation events were abstracted from the EMR and, where possible, direct communication with prescribers present during the event. The number of laryngoscopy attempts and adverse events were collected from the procedure notes, physician dictations, and direct communication when possible. Airway experts were defined as anesthesia providers or otolaryngologists.

Statistical Methods

No sample size calculation was performed given the retrospective study design evaluating an existing data set. Categorical variables are presented as sums and percentages. Statistical comparisons for these variables were performed using the chi-square test. All continuous variables were nonparametric and are presented as medians with interquartile range. For comparison of these variables, the Mann-Whitney U test was performed. These statistical analyses were prepared using SPSS 25 (IBM, Armonk, New York). Logistic regression analysis was also performed to identify the covariates associated with the outcome of extubation failure. Covariates included extubation shift, age, PIM-3 score, primary indication for mechanical ventilation, total extubation failure risk factors present, and duration of mechanical ventilation at the first extubation attempt. To quantify the association, the odds ratios were estimated with the corresponding 95% CIs. This analysis was performed with only the first mechanical ventilation course for an individual subject (no repeat subjects permitted), as well as at the level of the mechanical ventilation encounter (repeat subjects permitted). For the mechanical ventilation encounter analysis, logistic regression analysis was performed with generalized linear mixed models to account for the correlation of outcomes within repeat subjects. All hypothesis tests were 2-tailed with a P value < .05 to indicate statistical significance. Linear regression analysis was performed using SAS 9.4 (SAS Institute, Cary, North Carolina).