Descriptors of Failed Extubation in Norwood Patients Using Physiologic Data Streaming

The objective of this study is to evaluate the utility of high-frequency physiologic data during the extubation process and other clinical variables for describing the physiologic profile of extubation failure in neonates with hypoplastic left heart syndrome (HLHS) post-Norwood procedure. This is a single-center, retrospective analysis. Extubation events were collected from January 2016 until July 2021. Extubation failure was defined as the need for re-intubation within 48 h of extubation. The data included streaming heart rate, respiratory rate, blood pressure, arterial oxygen saturation, and cerebral/renal near-infrared spectroscopy (NIRS). The most recent blood laboratory results before extubation were also included. These markers, demographics, clinical characteristics, and ventilatory settings were compared between successful and failed extubations. The analysis included 311 extubations. The extubation failure rate was 10%. According to univariable analyses, failed extubations were preceded by higher respiratory rates (p = 0.029), lower end-tidal CO2 (p = 0.009), lower pH (p = 0.043), lower serum bicarbonate (p = 0.030), and lower partial pressure of O2 (p = 0.022). In the first 10 min after extubation, the failed events were characterized by lower arterial (p = 0.028) and cerebral NIRS (p = 0.018) saturations. Failed events were associated with persistently lower values for cerebral NIRS 2 h post-extubation (p = 0.027). In multivariable analysis, vocal cord anomaly, cerebral NIRS at 10 min post-extubation, renal NIRS at pre-extubation and post-extubation, and end-tidal CO2 at pre-extubation remained as significant co-variables. Oximetric indices before, in the 10 min immediately after, and 2 h after extubation and vocal cords paralysis are associated with failed extubation events in patients with parallel circulation.


Introduction
The transition between positive and negative pressure ventilation affects multiple organ systems, especially the heart and lungs [1][2][3][4]. As with intubation and mechanical ventilation, positive pressure reduces ventricular afterload and preload, lowering left ventricular end-diastolic pressure [1][2][3][4]. Thus, mechanical ventilation positively affects the heart [2][3][4][5]. Furthermore, extubation or the transition from positive to negative pressure ventilation can hurt patients with vulnerable cardiac hemodynamics due to the withdrawal of the favorable effect of the positive pressure on the heart [1,3,4]. Previous studies have demonstrated that cardiac output and systemic oxygen delivery decrease during extubation which is often associated with extubation failure [1][2][3].
Children with parallel circulation represent a unique subset of patients with congenital heart disease [6,7]. They have a circulation in which the systemic and pulmonary venous blood pools mix, resulting in an equal saturation of blood going to the pulmonary and systemic circulations [8,9]. If total cardiac output remains the same, an increase in either pulmonary or systemic blood flow must be met with a decrease in blood flow to the other circulation of equal but opposite magnitude [6][7][8]. This interdependence between the pulmonary and systemic circulations leaves these patients at increased risk for inadequate systemic oxygen delivery and cardio-respiratory deterioration [6][7][8]. Thus, hemodynamic changes with intubation and extubation may be exaggerated in those with parallel circulation; therefore, patients with parallel circulation are at risk of extubation failure, an event associated with increased morbidity and mortality [10][11][12]. Limited data describe the physiologic profile associated with extubation failure, especially for children with congenital heart disease, including patients with parallel circulation.
This study aimed to determine whether the dynamic patterns of hemodynamic and oximetric indices before, during, and shortly after extubations and other clinical variables are associated with failure events in neonatal patients with hypoplastic left heart syndrome post-Norwood procedure.

Study Population
The Institutional Review Board of Baylor College of Medicine (H-40811) approved this study with a waiver of written consent. Electronic medical records were obtained for all patients diagnosed with Hypoplastic Left Heart Syndrome (HLHS) status post-Norwood procedure from January 2016 until July 2021. These records were searched for all intubation and extubation dates between the Norwood surgery date and the discharge from the cardiac intensive care unit at Texas Children's Hospital.

Demographic and Clinical Data
Demographic data included gestational age, age at the surgery, birth weight, gender, race, genetic, vocal cord, airway, or diaphragm anomalies. The clinical data include the length of stay in the cardiac intensive care unit, length of each intubation, delayed sternal closure, ECMO support, vasoactive/ inotropic support dosage (VIS score), and the most recent pH, serum bicarbonate, lactate, PO2, and PCO2 labs within 6 h before each extubation.

Physiologic Data Collection
Hemodynamic data were collected using the Sickbay™ clinical platform (Medical Informatics Corp; Houston, TX, USA). Recorded physiologic data included heart rate (HR), peripheral oxygen saturation (SpO 2 ), diastolic (dBP), systolic (sBP) and mean arterial blood pressure (mBP), and respiratory rate (RR). These signals are captured at a frequency of 0.5 Hz and stored in the Sickbay system. All patients following the Norwood procedure are monitored with near-infrared spectroscopy (NIRS) oximeter. The NIRS probes were placed for cerebral and renal oximetry measurements, respectively. NIRS signals were captured at a frequency of 0.2 Hz and stored in the Sickbay system. An averaged venous oxygen saturation was determined using the cerebral (cSO 2 ) and renal (rSO 2 ) NIRS-derived oxygen saturations as follows SvO 2 = (cSO 2 + rSO 2 )/2 [13,14].

Definition of Study Time Intervals
The documented extubation time was set as the time axis's origin (0 min). Before extubating, the patient goes through a respiratory weaning process that varies from clinician to clinician and patient, lasting approximately 45-60 min. Therefore, the baseline before ventilatory weaning interval was chosen to range from − 120 to − 60 min, and the extubation preparation or weaning time ranged from − 60 to 0 min. The response interval after the extubation ranged from 0 to 180 min. Within this response interval, we focused on the first 10 min and the post-extubation steady state ranging from 120 to 180 min.

Types of Physiologic Variables
The variables of type I are the hemodynamic and oximetric indices at baseline computed as the average over the baseline interval from − 120 to − 60 min for each extubation. The variables of type II are the change (with respect to baseline values) in hemodynamic and oximetric indices at the first 10-min interval after the extubation. The variables of type III are the post-extubation steady state of the hemodynamic and oximetric index computed as the average over the steady-state window from 120 to 180 min for each extubation. This approach is designed to compare the response to each extubation event in relation to its baseline before the extubation. Variables of type IV are static variables, including demographics, clinical characteristics, mechanical ventilator settings, post-extubation non-invasive respiratory support settings, and most recent laboratory results.

Definition of Successful and Failed Extubation
Extubation was categorized as failed if there was a need for an unplanned re-intubation within 48 h. Otherwise, the extubation was deemed successful.

Statistical Analysis
Continuous demographic and clinical characteristics were expressed as medians (interquartile range). Categorical variables were described in frequencies. The variable was collected and averaged over the baseline interval to establish a baseline value for each hemodynamic or oximetric variable and each extubation.
The univariable associations between successful-vs-failed extubation and variables of type I, II, III, and IV were computed using the non-parametric Wilcoxon rank-sum test for continuous variables and Fisher's exact test for categorical variables. These tests were run using the Python statsmodels v0.12.2 library. For all tests, statistical significance was concluded at a p value < 0.05.
Multivariable analysis was based on logistic regression with mixed effects (to account for multiple extubations of the same patient) and a stepwise forward selection of significant co-variables. This selection process is based on the Akaike information criterion (AIC) and the statistical significance of the co-variables. At each step, the selection process examined all remaining variables to add to the model the variable that decreased the AIC the most while maintaining significant p values for all the included variables. The AIC penalizes models with too many variables that do not add sufficient information. Thus, this selection process excluded the least significant variables to arrive at a robust model with a reduced number of significant variables.

Results
In a cohort of 134 patients, there were 311 extubations during the interstage period with complete physiological data in the Sickbay system. Among these 134 patients, 34% were Hispanic, 55% were non-Hispanic White, 10% were non-Hispanic Black, and 1% were non-Hispanic Asian. The median age at the time of Norwood surgery was 7 days (IQR 4-10 days), and the median weight was 3.4 kg (IQR 3.0-3.7 kg). The median gestational age was 39 weeks (IQR 37-39 weeks). In this cohort, 42% had a genetic disorder, 10% had an airway anomaly, 2% had diaphragm anomaly, and 28% had a vocal cord anomaly. Fifty-seven percent had a right-ventricular to pulmonary artery shunt (RVPAS) and 43% a Blalock-Taussig shunt (BTS). For the Norwood surgery, the median cardio-pulmonary bypass time was 200 min (IQR 178-218 min), the median aortic cross-clamp time was 105 min (IQR 88-117 min), and the median circulatory arrest time was 11 min (IQR 8-15 min).
During the interstage period under study, the patients experienced a median number of extubation events equal to 2 (IQR 1-3). The extubation failure rate was 10% (31 of 311 events). The median length of intubation for the 311 events was 95 h (IQR 40-190 h). Out of the 31 extubation failures, 24 were followed by a re-intubation with a successful extubation. The median intubation length for these 24 re-intubations was 6.1 days. The other 5 extubations from the 31 failures were followed by a re-intubation whose extubation failed again. The median intubation length for these 5 reintubations was 4.3 days. Among those 31 failed events, the median duration from extubation to unplanned re-intubation was 16 h (IQR 3-27 h).
For the set of 311 extubation events, the median preextubation pH was 7.38 (IQR 7.35-7.42). The median pre-extubation serum bicarbonate was 27 mEq/L (IQR 25-30 mEq/L). The median pre-extubation lactate was 1.0 mmol/L (IQR 0.8-1.3 mmol/L). The median pre-extubation PO 2 was 45 mmHg (IQR 42-48 mmHg). The median pre-extubation PCO 2 was 43 mmHg (IQR 40-48 mmHg). The characteristics of pre-and post-extubation respiratory support and pre-extubation laboratory values are shown in Tables 1 and 2. The hemodynamic indices at baseline for the entire cohort and the successful and failed extubation groups are displayed in Table 3.

Variable Type I: Physiologic State at Baseline
Hemodynamic and oximetric indices at baseline (− 120 min to − 60 min) were analyzed, demonstrating that failed extubations were preceded by higher RR than the successful extubations (p = 0.029). Additionally, the ETCO 2 at baseline before failed extubations was lower than for the successful extubations (p = 0.009). Details are shown in Table 3.

Variable Type II: Physiologic Response at 10 min After Extubation
We analyzed the change (with respect to baseline values) in hemodynamic and oximetric indices (HR, mBP, sBP, dBP, RR, SpO 2 , SvO 2 ) at the first 10-min interval after the extubation. These results are shown in Table 3. We found that failed extubations experienced lower values for SpO 2 (p = 0.028), SvO 2 (p = 0.015), and cNIRS (p = 0.018) than the successful extubations during those first 10 min.

Variable Type III: Physiologic Response at Steady State After Extubation
We also analyzed whether there was a persistent association over time in the hemodynamic and oximetric indices with the successfulness of the extubations. See Table 3, which shows that the steady-state cNIRS is lower for failed extubations (p = 0.027).

Variable Type IV: Demographics, Clinical Characteristics, Non-invasive Respiratory Support, and Labs
Failed extubations were associated with lower pH (p = 0.043), lower serum bicarbonate (p = 0.030), and lower PO 2 (p = 0.022). The demographic characteristics and the presence of genetic, airway, or diaphragm anomalies were not significantly associated with failed extubations. However, vocal cord anomalies are strongly associated with failed events (p = 0.005). See details in Table 3.   showing the dynamics of oximetric variables before, during, and after the extubations.

Discussion
This current study demonstrates factors associated with extubation failure in neonatal patients with hypoplastic left heart syndrome post-Norwood procedure. Vocal cord anomaly (p = 0.022), higher respiratory rates at pre-extubation baseline (p = 0.029), lower partial pressure of O 2 at 10 min post-extubation (p = 0.022), lower cerebral NIRS at 10 min post-extubation (p < 0.001), lower renal NIRS at pre-extubation baseline (p = 0.001), lower renal NIRS at post-extubation steady state (p = 0.017), and lower end-tidal CO2 at pre-extubation baseline (p < 0.001) were all associated with increased risk of extubation failure. Mechanical ventilation parameters and the presence or absence of CPAP trial were not associated with the failed extubation events. Failed extubation is associated with increased length of stay and increased risk of inpatient mortality [10][11][12]15]. Extubation failure is most frequently defined in published data as needing re-intubation within 24 or 48 h [16][17][18]. Using these definitions, extubation failure ranged from 5 to Table 3 Vitals signs at baseline, 10-min after extubation and post-extubation steady state HR heart rate; mBP mean arterial blood pressure; sBP systolic arterial blood pressure; dBP diastolic arterial blood pressure; RR respiratory rate; SpO 2 peripheral oxygen saturation; SvO 2 venous oxygen saturation; cNIRS cerebral NIRS oxygen saturation; rNIRS renal NIRS oxygen saturation; ETCO 2 end-tidal carbon dioxide  24% [16,17,19]. Previous studies identified many factors associated with extubation failure, including the presence of an airway and genetic anomalies in pediatric patients with and without congenital heart disease [18,[20][21][22][23]. Similarly, we found vocal cord anomaly as a factor associated with failed extubation in the present study. When studies looked at children after the Norwood operation, specifically, the following have been demonstrated to be associated with extubation failure: the need for nitric oxide, longer duration of mechanical ventilation, the presence of chylothorax, and higher cumulative midazolam dose [10][11][12]. The decrease in NIRS demonstrates an insufficiency in meeting the oxygen demand in the brain and kidneys [1-3, 6, 7, 13, 14, 24-28]. When applied to parallel circulation physiology, these findings seem consistent with lower oxygen delivery possibly triggered by lower cardiac output or complex hemodynamic interaction due to the systemic-topulmonary interplay in single ventricle neonates status post-Norwood procedure [1-3, 6, 7, 14, 24-26, 29-31]. The same cardiovascular construct can cause lower arterial oxygenation 10 min post-extubation [1-3, 6, 7, 14, 24-26, 30, 31]. Other studies also found a reduction in renal NIRS during an extubation readiness trial and an increase in cerebral NIRS from baseline to extubation in congenital heart patients at risk of extubation failure [24,25]. However, those studies did not assess specifically the Norwood population.
Our institution does not have a standardized extubation readiness trial for this patient population [24,32]. Some clinicians perform the CPAP trial as an extubation readiness assessment, but it is not done universally. The , and renal NIRS (right) before, during, and after the extubation events. The plots display the moving averages plus/minus standard error for the successful and failed extubations subgroups presence or absence of a CPAP trial was not associated with the incidence of failed extubation events.
Also, a lower ETCO 2 at baseline could be caused by increased respiratory rate due to abnormal cardiac and pulmonary interaction, reflecting vulnerable cardiac output and alteration in the balance of pulmonary-to-systemic blood flow ratio in single ventricle neonates with parallel circulation [1-3, 6, 7, 14, 24-26]. Previous studies also found a rapid shallow breathing pattern before extubation in pediatric cardiac patients as a factor associated with failed extubation [32,33].
The present study utilizes high-fidelity continuous monitoring data to determine hemodynamic indices before, immediately after, and a few hours following extubation. This is important as such data can help guide clinical decision-making. Specifically, identifying those patients at risk for extubation failure shortly after the procedure allows the clinician to adjust the inotropic support and provide selectively non-invasive respiratory support to avoid the failed extubation events [34,35]. As failed extubations have been demonstrated to be associated with increased length of stay and increased mortality, preventing failed extubations can help avoid these outcomes [10, 15-20, 23, 36]. After extubation, the hemodynamic state must be monitored closely immediately after extubation to help minimize morbidity and mortality [34,35].
This study is novel because it is one of the few studies to use high-fidelity streaming data from continuous monitoring to complement other clinical information about the patient to characterize extubation failure. Other pediatric studies have investigated factors associated with extubation failure but have not done so with such a high-fidelity focus on hemodynamics and oximetry. Additionally, this is one of few studies to focus on extubation failure in those with parallel circulation. Our findings can be directly applied in the clinical care of children with parallel circulation.
While this study has its strengths and is additive to the existing literature, it is not without limitations. One of the variables associated with extubation failure is the presence of vocal cord anomaly. This variable was identified based on clinical suspicion and corroborated using ultrasound or flexible laryngoscopy. We acknowledge that there is a possibility that some vocal cord anomalies could have been missed if there was no clinical suspicion. As a single-center study, there may be institution-based practices that are not captured here that may impact extubation success. Thus, some findings may not be reproducible at other centers. As this study focused on characterizing the patient's hemodynamics and oximetric dynamics rather than specific interventions, it should help maximize generalizability.

Conclusion
Hemodynamic and oximetric indices before extubation, in the 10 min immediately after extubation, in the 3 h following extubation, and vocal cords paralysis are associated with failed extubation events in patients with parallel circulation. These findings may serve as a steppingstone for future studies to evaluate the potential of dynamic physiologic markers to complement other clinical information to detect extubation failure early and guide clinical interventions.

Author Contributions
The authors take responsibility for all aspects of the reliability and freedom from bias of the data presented and their discussed interpretation.

Declarations
Conflict of interest Dr. Rusin is a co-founder of Medical Informatics Corp. No funding was provided by the company to support this work. All other authors report no conflicts of interest. No additional sources of funding to report.