Decreasing Exposure to Mechanical Ventilation in Extremely Preterm Infants

Abstract

Background and Objectives:

Provider variability in mechanical ventilation weaning practices may result in prolonged invasive ventilation exposure that may contribute to the development of BPD in extremely preterm infants. Our SMART aim was to reduce the proportion of infants born between 24 and 28 6/7 weeks’ gestation exposed to invasive ventilation for > seven days by 25% within 12 months.

Methods:

This was a single center quality improvement initiative at the University of Alabama at Birmingham conducted between January 2021 and March 2023. Interventions to reduce mechanical ventilation exposure included: 1) establishing a consensus for invasive ventilation exposure including intubation, extubation, and reintubation, 2) development of a bedside weaning readiness tool to facilitate permissive hypercapnia, 3) empowerment of RT facilitated weaning, and 4) the creation of a ventilation dashboard. All measures were analyzed using statistical process control charts.

Results:

The initiative included 340 infants with a median gestational age of 26 6/7 weeks (IQR 25–28) and birth weight of 842 grams (IQR 686–1011). The baseline proportion of infants invasively ventilated for > seven days at baseline was 44%, which decreased to 25% following implementation of the bedside weaning readiness and communication tool. Special cause variation was also observed for rates of BPD or death, which decreased from a baseline of 55% to 40%.

Conclusions:

In infants between 24 and 28 6/7 weeks’ gestation, systematic ventilator weaning was associated with a decrease in infants invasively ventilated for > seven days and a decrease in BPD or death.

Article Summary:

Systematic ventilator weaning decreased the proportion of infants invasively ventilated for > seven days, which was associated with a decrease in bronchopulmonary dysplasia or death.

Introduction

Bronchopulmonary dysplasia (BPD) commonly impacts infants born extremely preterm,¹ with rates that have increased over time.² Infants who develop BPD are not only at risk for neurodevelopmental impairment but frequently have persistence of lung disease into adulthood.^3,4 Evidence-based practices to safely prevent BPD include avoidance of intubation,⁵ less invasive surfactant administration (LISA),⁶ and pharmacotherapies including caffeine⁷ and systemic corticosteroids administered after the first postnatal week.⁸ Other commonly employed strategies such as intubation-surfactant-extubation (INSURE) to reduce exposure to mechanical ventilation may be less effective in BPD reduction.⁶ Given that rates of BPD vary by hospital,⁹ quality improvement methodology informed by evidence based practices may reduce infants’ risk for BPD development and improve long term outcomes in this vulnerable population.

In a recent systematic review of quality improvement initiatives to reduce BPD,¹⁰ optimization of non-invasive ventilation was the most commonly employed method for BPD prevention. While a reduction in delivery room intubation may be feasible and associated with a decrease in BPD,¹¹ infants born at lower gestational ages commonly require invasive support in the early postnatal period.^12,13 In infants requiring mechanical ventilation, the duration and frequency of courses of ventilation are associated with BPD.¹⁴ Therefore, systematic efforts to minimize ventilation exposure in the very infants at highest risk¹⁵ may reduce the likelihood of BPD development.

At the University of Alabama at Birmingham, the composite rate of grades 1–3 BPD³ more than doubled between 2014 and 2020 from 22% to 56% among infants born at <29 weeks’ gestation. In evaluating our center’s respiratory support practices, we noted an intubation rate in the delivery room of 39% and a surfactant exposure rate of 71%; both of these metrics were below the National Institute of Child Health and Human Development Neonatal Research Network average. We therefore concentrated efforts on the next most proximal area of respiratory care: ventilator weaning. Our SMART aim was to decrease the proportion of infants born between 24 and 28 6/7 weeks’ gestation invasively ventilated for > seven days from 44% to 33% (a decrease of 25%) by January 1, 2022 with sustainment through January 1, 2023.

Methods

Hospital Setting

This initiative was conducted within the 120 bed Level III regional neonatal intensive care unit (NICU) at the University of Alabama (UAB) between January 1, 2021 and March 14, 2023. Annually, there are between 150 and 200 infants born between 24 and 28 6/7 weeks. Upon consideration of potential drivers of BPD reduction our team generated a Driver Diagram (Figure 1), which focused on reducing invasive ventilation exposure. Key drivers were determined using a failure mode effects analysis to identify failure modes impacting the duration of invasive ventilation. Our institution uses pressure-controlled time-cycled synchronized intermittent mechanical ventilation with the Dräger Babylog^® ventilator for invasive and non-invasive positive pressure ventilation (the mode to which infants were extubated). Our center had previously trialed volume-targeted ventilation until a series of micropremies were noted to have excessive chest expansion despite appropriate volume targets¹⁶ (4 to 6 ml/kg) for which there were growing concerns of volutrauma. While LISA and INSURE were means of surfactant administration during the action phase of this initiative, their use was not systematically assessed. Both high frequency jet and oscillatory ventilation were available as rescue ventilation strategies in infants unable to maintain adequate ventilation or oxygenation using conventional ventilatory strategies.

Figure 1.

Key Driver Diagram

Our initiative was conducted after review and exemption by the Institutional Review Board. We opted to only include infants born between 24 weeks and 28 6/7 weeks’ gestation, given the need for prolonged mechanical ventilation and very high risk of BPD in infants born at < 24 weeks gestational age.¹⁵ Infants born with critical congenital anomalies, genetic disorders, or infants in which extubation was clinically restricted (e.g. spontaneous intestinal perforation requiring peritoneal drainage) were excluded. Our multidisciplinary team consisted of neonatologists, neonatal fellows, respiratory therapists (RTs), nursing leadership, nurse educators, clinical research informaticists, and data scientists. We utilized the model for improvement throughout our initiative.

Interventions

Prior to the initiative, our team noted provider variation in the criteria for which infants should either be intubated, extubated, or reintubated. We also identified opportunities for improvement involving the determination of a weaning opportunity, documentation of ventilator weaning, and the provider responsible for ventilator weaning. Interventions included (1): consensus guideline development (January 2021), (2) “Weanogram” development (April 2021), (3) bedside weaning log (April 2021), (4) RT empowerment plan (May 2021) with compliance report (August 2021), and (5) ventilator dashboard initiation (May 2022).

Development of Consensus Guidelines

We first developed guidelines for when an infant should be intubated, extubated, or reintubated. Evidence based¹³ intubation criteria included sustained bradycardia <100, a sustained fraction of inspired oxygen (FiO₂) >0.50 to maintain oxygen saturations >90% at 10 minutes following birth, a carbon dioxide (CO₂) level >65 (on either venous, capillary or arterial blood gas) refractory to corrective noninvasive ventilation maneuvers, or hemodynamic instability. Subsequently, infants were to be extubated when no longer meeting any of the intubation criteria. Reintubation criteria were the same as intubation criteria, except for a higher FiO₂ of 0.80. Intubation guideline adherence was facilitated by mandatory fields within the procedure notes regarding the intubation indication. We operationalized extubation guidelines using an extubation order requiring all criteria being met.

Standardization of Weaning Opportunities

The success of this intervention was predicated on the consistent use of transcutaneous carbon dioxide (TCO₂) monitoring, which had been made our unit’s standard of care in infants born at <29 weeks’ gestation starting at birth.¹⁷ By correlating TCO₂ values with the pH and CO₂ values from daily blood gases, providers estimated the pH at the time of an assessment using the physiologic principle of a CO₂ change of 10 mm Hg modulating the pH by increments of 0.08.¹⁸ Given inconsistency in determinations of whether ventilator weaning was indicated, we developed a bedside tool we entitled the “Weanogram” (Supplemental Figure 1) by which all providers could make a similar determination of weaning eligibility. The tool was an extension of criteria used in the SUPPORT Trial,¹³ wherein ventilator weaning would be indicated if an infant’s pH was estimated to be >7.20 at the time of assessment. A description for this tool can be found in the figure legend for Supplemental Figure 1.

Ventilator Weaning Documentation

As an initial means of monitoring ventilatory weaning compliance, we utilized a bedside ventilator monitoring log. This log was utilized to gain division wide buy-in to demonstrate (1) infants were being actively weaned from support and (2) ventilator changes were being performed appropriately. This log was attached to the ventilator wherein ventilator changes were documented both in this form and in the EMR (Cerner). Items completed in the form included the TCO₂, the ventilator change and whether the current settings were extubatable settings. These forms were employed for the first six months of the initiative with ongoing, weekly audits. The use of the powerplan, weanogram, and ventilator weaning documentation were collectively referred to as our “bundle go-live.”

Empowerment of RT Facilitated Weaning

Prior to the initiative, neonatal-perinatal medicine fellows and faculty were primarily responsible for ventilator weaning. Given that RTs assessed invasively ventilated preterm infants more frequently (every 4 hours), our team opted to broaden responsibilities for ventilator weaning to RTs. So as to empower RTs in making these clinical decisions while remaining within scope of practice, we extended the Weanogram into an RT power plan in the electronic medical record (EMR) in which a blood gas pH and corresponding TCO₂ range would indicate (1) the need for a ventilator change) and (2) an indication for the scale and prioritization of the ventilator change. RTs received educational sessions on Weanogram utilization from team physicians including case scenarios.

Ventilator Dashboard

To provide a more ongoing, sustainable means of tracking ventilator weaning and assessments, we created a ventilator dashboard that was generated weekly for invasively ventilated infants. The weekly dashboard was shared with the on-service attending and fellow(s), nursing leadership, and RT leadership and included (1) an assessment of overall weaning performance and (2) patient specific data (Supplemental Figure 2). This dashboard was then later automated once no longer iterative by automated data TCO₂ achievement which were then entered into a Tableau^™ derived dashboard.

Measurement Strategy and Analysis

Our primary outcome measure was the proportion of infants that died or were invasively ventilated for > seven days in infants born between 24 and 28 6/7 weeks’ gestation. We chose seven days given data suggesting an increase in risk for BPD with each additional week of ventilator exposure.¹⁴ We aimed to decrease the proportion of infants invasively ventilated for > seven days from 40% to 30%, which would be a relative 25% decrease. We chose this goal based on population data indicating a median mechanical ventilation exposure of ~2 days in a more expansive population of infants born between 25- and 31-weeks’ gestation and ~22 days in infants born at 24 weeks’ gestation.¹² This reported outcome measure slightly differed from the measure tracked and communicated to staff throughout the initiative. The measure used during the initiative was alive ventilator free days per infant, which was viewed as a more sensitive measure to more proximally relate interventions to the duration of individual patients’ ventilator exposure. Ventilator exposure was initially manually abstracted from the EMR and calculated per patient. With the creation of the automated dashboard, reports were generated weekly using data originating in Cerner.

Additional outcome measures included BPD, BPD or death, and grades of BPD over time using the Jensen definition for BPD.³ Given the anticipated variability in BPD risk across subpopulations over time (and the potential of this variable risk to impact interactions between process and outcome measures), we also analyzed estimated risk for BPD or death using the NICHD Risk Estimator on the first postnatal day.¹⁵ The social constructs of race and ethnicity were documented based on self-report as was necessary for BPD risk estimations. Our process measure was ventilation weaning compliance which was abstracted from bedside weaning logs with weaning compliance defined as implementation of three appropriate ventilator weans per patient ventilator day. Weaning compliance was reported for the first six months following RT empowerment plans. Weaning compliance beyond this time period was assessed utilizing the ventilator weaning dashboard, wherein TCO₂ distributions were shared with staff so as to identify ventilated patients with less permissive CO₂ distributions. As changes in CO₂ have been associated with development of intracranial hemorrhage (ICH),^18,19 we monitored rates of any grade ICH over the course of the initiative as a balancing measure.

All measures were analyzed using control charts within Excel^™ derived macros created at Cincinnati Children’s using standard rules of special cause variation²⁰ to establish whether interventions were associated with measurable changes over time. We longitudinally assessed outcomes using a p-chart in defined subgroups of ten infants. We chose a time period comprising 12 data points²⁰ prior to our interventions as our baseline period (December 2019-March 2021), followed by our intervention period through May 2022, and a 10 month sustainment period ending in March 2023. Analyses of differences between infants born between the baseline and intervention period as well as distributions of quarterly BPD severity were performed using chi-square analysis for categorical variables and the appropriate post test for continuous variables following tests of normality. Analyses were performed using IBM SPSS Statistics version 28.

Results

During the course of the initiative, 350 infants were included, of whom 120 infants were born during the baseline period and 230 infants were born during the intervention period. Infants born during the baseline period did not differ to those born during the intervention period with regard to gestational age (median 26.6; IQR 25.4–27.6 vs median 27.0; IQR 25.5–27.5 weeks; p=0.19), birth weight (median 850; IQR 674–995 vs median 837; IQR 696–1027 grams; p=0.49), and male sex (54% vs. 49%; p=0.33). Other clinical factors related to BPD development including postnatal corticosteroid exposure (23% vs 17%; p=0.24) and patent ductus arteriosus exposure (29% vs 38%; p=0.11) did not differ between study periods, however there was a lower rate of antenatal corticosteroid exposure in the baseline period compared to intervention period (91% vs 96%; p=0.04). The baseline risk for BPD or death as estimated on postnatal day one did not differ between baseline and intervention periods (62% vs. 48%; p=0.35) ((Table 1).

Table 1.

Demographic and Clinical Characteristics of Included Patients

	Baseline (n=120)	Intervention Period (n=230)	P value
Demographic Characteristics n (%)
Gestational age (weeks)a	26.6 (25.4–27.6)	27.0 (25.5–27.5)	0.19
24w – 25w 6 days	38 (32)	66 (29)	0.56
26w – 28w 6 days	82 (68)	164 (71)
Birth weight (g)a	850 (674–995)	837 (696–1027)	0.49
Male sex	65 (54)	111 (48)	0.29
Raceb			0.45
Black	61 (51)	115 (50)
White	48 (40)	90 (39)
Otherc	3 (3)	14 (6)
Unknown	8 (7)	11 (5)
Hispanic or Latino ethnic groupb	0 (0)	5 (2)	0.25
Antenatal corticosteroids	109 (91)	221 (96)	0.04
Baseline risk for BPD or death (%)a,d	62 (21–80)	48 (28–74)	0.35
Clinical Exposures and Outcomes n (%)
Invasive ventilation exposure	85 (71)	168 (73)	0.66
Surfactant	84 (70)	170 (74)	0.44
Inhaled nitric oxide	16 (13)	32 (14)	0.88
Corticosteroids for BPD	27 (23)	38 (17)	0.17
Patent ductus arteriosus	35 (29)	87 (38)	0.11
Early onset sepsis	3 (3)	2 (1)	0.23
Late onset sepsis	15 (13)	28 (12)	0.93
Severe retinopathy of prematurity	7 (6)	13 (6)	0.95
Severe intracranial hemorrhage	16 (14)e	20 (9)	0.14
Necrotizing enterocolitis stage ≥ 2	13 (11)	12 (5)	0.05
Death	13 (11)	20 (9)	0.52

^amedian (IQR);

^bgroup based on self report from parent or guardian

^cother designation used in the context of a parent or guardian not identifying with the designations of non-Hispanic Black, Hispanic, or non-Hispanic White (the designations used in the NICHD BPD Risk Estimator¹⁵)

^das estimated using the NICHD BPD Risk Estimator¹⁵

^edata not available in 5 patients

BPD: bronchopulmonary dysplasia

The baseline proportion of infants that died or were invasively ventilated for > seven days in the first postnatal month was 44% between December 2019 and March 2021. A shift was present during May of 2021 with a decrease to 25% following the nursing/RT education (Figure 2). Rates of BPD or death at 36 weeks’ PMA decreased from 55% during the baseline period to 40% during the intervention period with a shift noted in November 2021 subsequent to our compliance report (Figure 3). All special cause variation resulted from eight points either above or below the center line. In further reviewing quarterly rates of BPD by severity, the rates of grade 1, 2, and 3 BPD were 38%, 10%, and 4% during the baseline period and were 10%, 10%, and 0% during the final monitoring quarter, respectively (p=0.002; Figure 4).

Figure 2.

P-chart indicating the proportion of infants that died or were invasively ventilated for > seven days between December of 2019 and March 2023. Special cause variation was noted in May of 2021 at which time the percentage of infants decreased from 44% to 25%. Numbers in parentheses along the x-axis correspond to the number of infants included per monitoring period. UCL: upper control limit; LCL: lower control limit.

Figure 3.

P-chart indicating the proportion of infants born between 24.0 weeks and 28 6/7 weeks’ gestation that developed either grade 1–3 bronchopulmonary dysplasia or died during the initiative. Rates of BPD or death decreased from 55% to 40% in November 2021 following special cause variation. Numbers in parentheses along the x-axis correspond to the number of included infants per monitoring period. Estimated BPD Risk reflects the median per monitoring period as estimated using the 2011 NICHD BPD Outcome Estimator on postnatal day 1. UCL: upper control limit; LCL: lower control limit.

Figure 4.

Stacked bar chart indicating the proportion of infants that developed no BPD, grade 1 BPD, grade 2 BPD, grade 3 BPD, or death. The presence and distribution of BPD severity during each epoch of the intervention period were compared to the cumulative distribution of outcomes during the baseline period. Grade 2 and 3 BPD were combined for analyses to satisfy conditions necessary for chi square analysis. *p<0.05

Regarding the process measure of ventilator assessment compliance, between the months of May 2021 and October 2021, the rate of ventilator assessment compliance was sustained at ~87% (Supplemental Figure 3). For the balancing measure of ICH, the rate remained at ~24% throughout both the baseline and action periods of the initiative without evidence of special cause variation (Supplemental Figure 4).

Discussion

In this single center quality improvement initiative, systematic ventilator weaning was associated with a decrease in the proportion of infants that died or were invasively ventilated for > seven days. Moreover, this improvement was not only associated with a decline in rates of BPD but an increase in the proportion of infants with milder BPD phenotypes.

In considering interventions to reduce BPD at our institution, several drivers were considered. Rates of delivery room intubation and surfactant delivery were at or below reported national benchmarks, for which reducing cumulative invasive ventilation exposure was our primary focus. Permissive hypercapnia, which may reduce lung injury,²¹ operationalized through systematic weaning practices was central to our initiative. While fluctuations in CO₂¹⁸ have been associated with ICH, we did not observe a change in the proportion of infants with ICH during the course of the initiative. Few previous improvement initiatives have reported reductions in ventilation exposure in a comparable population of preterm infants. Prior initiatives implementing potential better practices to minimize invasive ventilation exposure including infants born at < 32 week’s gestation²² or <1500g^23,24 have reported a decrease in invasive ventilation exposure to ~2–3 days, which is comparable to what was achieved in our initiative.

Invasive ventilation remains necessary in some extremely preterm infants, with a high correlation between the duration of mechanical ventilation exposure and gestational age. In a prior cohort study, nearly 50% of infants born at 24 weeks’ gestation remained intubated at the end of the first postnatal month as compared to 30% of infants born at 25 weeks’ gestation.¹² So as to both minimize the number of infants exposed to invasive ventilation and maximize the number of infants extubated, standard intubation and extubation criteria were implemented in this initiative. These criteria were derived from the SUPPORT Trial,¹³ in which infants in the continuous positive airway pressure group were to be extubated upon reaching a PaCO₂ below 65 mm Hg with a pH >7.20.

An imposing challenge for quality improvement initiatives targeting BPD reduction, is defining the patient population to include and the means of analyzing data over time to quantify improvement. As ~90% of infants born at 22 to 23 weeks’ gestation ultimately develop BPD,¹ BPD prevention may not be feasible for which infants born at these gestational ages were excluded from our and prior¹⁰ analyses. While we restricted our population to infants born between 24 and 28 6/7 weeks’ gestation, infants’ baseline BPD risk varied throughout the intervention period, which may have impacted the temporality between decreases in ventilation exposure and the outcome of BPD or death. A prior initiative with a similar patient population (median of 27 weeks’ gestation at birth), intubation rate in the delivery room (34%) and duration of mechanical ventilation exposure (4 days) reported a rate of BPD or death of 30%¹¹ compared to 40% observed in our initiative. Past initiatives integrating a BPD prevention bundle included infants born at <33 weeks’ gestation with an average gestational age of 29 weeks’ have reported BPD rates of 9%-32% following interventions.^25,26

Monitoring weaning practices through compliance reports and our ventilator dashboard also likely contributed to the success of our initiative. As guideline adherence may diminish with time,²⁷ these efforts evaluated the consistency of systematic ventilatory weaning efforts. The ventilator dashboard allowed dissemination of performance metrics to on service care providers including identification of weaning opportunities in those infants with lower CO₂ distributions. Other efforts to standardize respiratory management through the use of an audit and feedback dashboard have similarly reduced invasive ventilation exposure in patients receiving general anesthesia.²⁸ However, the generalizability and feasibility of dashboard integration into clinical care may be limited by hospital resources.²⁹

This initiative’s strengths include a reduction in the heterogeneity of clinical practice as it relates to several aspects of respiratory care in extremely preterm infants, including intubation, identification and facilitation of ventilatory weaning, and extubation. Through the constellation of these practices, we observed a reduction in ventilation exposure in a large tertiary care NICU with a large staff of bedside nurses, respiratory therapists, neonatologists, and trainees. Whereas we achieved our SMART aim of reducing ventilation exposure, it remains unknown whether the magnitude of these changes is sufficient to explain associated declines in BPD. While less systematic, efforts to standardize postnatal corticosteroid utilization in infants on invasive support on postnatal day 14 also occurred during the intervention period. Moreover, introduction of evidence-based care practices that may impact BPD rates in one NICU may not impact rates in another unit.³⁰ More specifically, other evidence practices such as volume targeted ventilation may obviate the need for systematic ventilator weaning.³¹ Given the multifactorial origins of BPD, impactful interventions are likely institutional specific.

Conclusion

In this quality improvement initiative systematic ventilator weaning practices were associated both with an increase in ventilator freedom as well as a decrease in the rate of BPD or death. Moreover, this observed increase in ventilator freedom was sustained for over a year. Given the use of invasive ventilation to support patients more broadly, such data may be relevant to other pediatric and potentially even adult clinical contexts. However, given the many mechanisms of lung injury in the preterm infant, broader, multi-institutional efforts to reduce longitudinal ventilator exposure are needed to better characterize the benefits of such efforts to long term pulmonary outcomes in preterm infants.

Supplementary Material

Supplemental Figure 1

Supplemental Figure 1. “Weanogram” for the systematic identification of weaning opportunities in invasively ventilated preterm infants. To use the Weanogram, a provider would (1) review the infant’s most recent blood gas, (2) find the relevant row on the Weanogram relating the transcutaneous CO₂ (TCO₂) to the pH from the blood gas, (3) assess the current TCO₂, and (4) determine whether a ventilator change was indicated. As an example, if the pH was 7.28 on the most recent blood gas with a corresponding TCO₂ was 58, and the current TCO₂ was 54, the provider would wean the ventilator.

Supplemental Figure 3

Supplemental Figure 3. P-chart for the process measure of ventilator assessment compliance over the course of the initiative. The proportion of ventilator weaning compliance remained at ~90% during the period during which it was monitored. Numbers in parentheses along the x-axis correspond to the number of infants assessed per monitoring period. UCL: upper control limit; LCL: lower control limit.

Supplemental Figure 2

Supplemental Figure 2. Representation of the ventilator weaning dashboard. The overall performance section included the number of invasively ventilated infants, average ventilator-free days and TCO₂ distribution in invasively ventilated infants. For patient specific data, the proportion of time an infant remained within, below, and above their TCO₂ targets were visualized along with when dexamethasone would be indicated. TCO₂ values utilized in the dashboard were abstracted manually from the EMR and imputed into an Excel^™ derived macro.

Supplemental Figure 4

Supplemental Figure 4. P-chart for the balancing measured of any grade intracranial hemorrhage. The proportion of infants with hemorrhage remained at 24% for the entirety of the initiative. Numbers in parentheses along the x-axis correspond to the number of infants included per monitoring period. UCL: upper control limit; LCL: lower control limit.

Abbreviations:

BPD

bronchopulmonary dysplasia

RT

respiratory therapist

FiO₂

fraction of inspired oxygen

CO₂

carbon dioxide

TCO₂

transcutaneous CO₂

TV

tidal volume

ICH

intracranial hemorrhage