Nasal high flow therapy introduction lowers reintubation risk in a Peruvian paediatric intensive care unit

Abstract

Aim.

We examined the impact of introducing high-flow nasal oxygen therapy (HFNT) on children under five with post-extubation respiratory failure in a paediatric intensive care unit (PICU) in Peru.

Methods.

This quasi-experimental study compared clinical outcomes before and after initial HFNT deployment in the PICU at Instituto Nacional de Salud del Niño in Lima in June 2016. We compared three groups: 29 received post-extubation HFNT and 17 received continuous positive airways pressure (CPAP) from 2016-17 and 12 historical controls received CPAP from 2012-16. The primary outcome was the need for mechanical ventilation. Adjusted hazard ratios (aHR) and 95% confidence intervals (95% CI) were calculated via survival analysis.

Results.

HFNT and CPAP did not alter the need for mechanical ventilation after extubation (aHR 0.47, 95% CI 0.15-1.48 and 0.96, 95% CI 0.35-2.62, respectively) but did reduce the risk of reintubation (aHR 0.18, 95% CI 0.06-0.57 and 0.14, 95% CI 0.03-0.72, respectively). PICU length of stay was 11, 18 and 37 days for CPAP, HFNT and historical CPAP and mortality was 12%, 7% and 27%, respectively. There was no effect on the duration of sedative infusion.

Conclusion.

HFNT provided effective support for some children, but larger studies in resource-constrained settings are needed.

INTRODUCTION

Despite improvements in global child mortality, more than 650,000 children under five die of lower respiratory infections every year (1). The vast majority of these deaths occur in low- and middle-income countries, due to more restricted access to basic interventions like immunisation, antibiotics and supplemental oxygen (2–4). Middle-income countries have a high burden of paediatric respiratory disease. They are also in a unique position to reduce respiratory-related mortality, by improving access to advanced respiratory support, such as mechanical ventilation and non-invasive respiratory support. Using mechanical ventilation leads to complications like ventilator-associated pneumonia, ventilator-induced lung injuries, laryngeal trauma and the prolonged use of sedative medications. Given these complications, many high-income countries increasingly use non-invasive respiratory support, such as continuous positive airway pressure (CPAP) and high-flow nasal oxygen therapy via a cannula (HFNT), to treat and prevent respiratory failure (5,6). The availability of non-invasive support in resource-constrained areas lags behind other countries, especially in remote areas (7,8).

HFNT is a form of non-invasive respiratory support that improves ventilation and oxygenation by washing carbon dioxide from the extrathoracic dead space and generating positive airway pressure (9,10). Early retrospective studies reported reduced intubation rates for infants with bronchiolitis after HFNT was introduced (11,12). A large multicentre randomised controlled trial of HFNT use in 1472 infants with bronchiolitis demonstrated that escalation of care was lower in the HFNT group than in those treated with low-flow oxygen (13). However, another multicentre randomised controlled trial showed higher rates of treatment failure for HFNT than CPAP for infants with bronchiolitis (14). Furthermore, randomised controlled trials of neonates showed that HFNT was inferior to CPAP for neonatal respiratory distress (15,16). It is notable that all these trials were performed in resource-rich settings, where adequate staffing was available to maintain the occlusive mask interface required to provide positive pressure with CPAP. The simple nasal cannula interface of HFNT is appealing in resource-constrained settings with limited staffing and resources. Staff in these settings have reported that providing bedside HFNT is much easier than CPAP (17). To date, HFNT has been successfully implemented in a variety of resource-limited settings (18,19) and there is an ongoing trial that is examining whether its use could reduce mortality in African children (20).

Respiratory disease was the second most common cause of PICU mortality in Instituto Nacional de Salud del Niño, a large Peruvian children’s hospital, in 2016 (21). Although non-invasive respiratory support was available, using CPAP frequently resulted in treatment failure due to leaks at the interface and reintubation. This led to prolonged duration of mechanical ventilation. To address these challenges, we introduced HFNT in July 2016 as an alternative respiratory support modality. The aim of this study was to determine the effect of HFNT on the need for mechanical ventilation in children with post-extubation respiratory failure in the hospital’s PICU.

METHODS

Setting and study population

Instituto Nacional de Salud del Niño is a major tertiary referral centre for children with medical and surgical diseases throughout the country. It has 15 PICU beds, and critically ill children are admitted on a case-by-case basis. More than 80% of patients receive mechanical ventilation in the PICU, and approximately 70% are intubated prior to PICU admission. This study considered all children under the age of five years who received mechanical ventilation within 24 hours of PICU admission between July 2016 and November 2017 (Figure 1). Informed consent was obtained from their parents or guardians while the patients were intubated. Children with craniofacial and airway malformations were excluded. Patients received the local standard of care, including the standardised extubation protocol that was introduced in February 2016 (Appendix S1). The intervention group included children who developed post-extubation respiratory failure, defined by initiation of respiratory support with HFNT or CPAP. The treating physicians selected the type of respiratory support using their clinical judgement rather than pre-specified criteria. Children who received no respiratory support following extubation were excluded. The historical controls were children under five years of age who received post-extubation CPAP during the four years preceding HFNT implementation, from January 2012 to June 2016. They were identified by retrospective chart review.

Figure 1.

Flow diagram of study participants. The intervention group included all children less than five years of age in the PICU who developed post-extubation respiratory failure from July 2016 to November 2017. The historical CPAP control group included all children less than five years of age in the PICU who developed post-extubation respiratory failure and received support with CPAP from January 2012 to June 2016.

Study design

This quasi-experimental study compared clinical outcomes in children from birth to five years of age with post-extubation respiratory failure in the hospital’s PICU before and after HFNT implementation. The primary outcome of interest was the need for mechanical ventilation following initial extubation. The secondary outcomes included reintubation, PICU length of stay, days of continuous sedative infusions, mortality and complications, such as pneumothorax and skin breakdown. All PICU providers received comprehensive HFNT training prior to its deployment. The study was approved by the Seattle Children’s Institutional Review Board and the Ethics Committee at Instituto Nacional de Salud del Niño.

HFNT and CPAP management

At the time of this study, HFNT was not approved or available for paediatric use within Peru, but CPAP was readily available for use in the hospital’s PICU. Patients in the HFNT group received support with the Optiflow Junior HFNT system (Fisher & Paykel Healthcare, Inc., Auckland, New Zealand) according to the previously published research protocol (22). Additional details are available in Appendix S2. The patients remained in the PICU for the duration of the HFNT support. Patients in the historical and prospective CPAP groups received support with the 850 System (Fisher & Paykel Healthcare, Inc., Auckland, New Zealand) according to usual care under the management of the treating physician. Patients receiving CPAP support could transfer to the inpatient ward at the discretion of the treating physician in accordance with the local standard of care. The majority of the patients, 12 in the historical cohort and 15 in the prospective cohort, remained in the PICU for the duration of CPAP support.

Data collection and statistical analysis

All clinical data for enrolled patients were collected from the medical records and entered into a secure Research Electronic Data Capture database (University of Washington Institute of Translational Health Sciences, Washington, USA). The baseline characteristics of the three groups were described with descriptive statistics. We used means and standard deviations for normally distributed data, medians and interquartile ranges for non-normally distributed data and frequency and percentages for categorical variables. The effect of HFNT on the need for mechanical ventilation, reintubation and PICU length of stay was evaluated with survival analysis using Kaplan-Meier curves and Cox proportional hazard regression. These were adjusted for age, sex and paediatric risk of mortality (PRISM) score because of known associations between these variables and increased mortality (23–25). The log-rank test was used to compare survival curves between the groups. Days of continuous sedative infusions were compared among the three groups using multiple linear regression because no subjects were censored. These were adjusted for PRISM score to account for differences in severity of illness. The chi-square test of independence was used to compare differences in respiratory support complications between the three groups. All analyses were performed using STATA 14.1 (Stata Corp, Texas, USA).

RESULTS

Of the 54 patients who were enrolled, 29 received HFNT, 17 received CPAP and eight received no respiratory support after extubation. Those who did not receive any respiratory support were excluded from the analysis and the 46 who received HFNT and CPAP were compared with 12 historical controls (Figure 1). There were no differences in age, sex, weight, PRISM score or median duration of mechanical ventilation prior to initial extubation, but more patients in the HFNT group had a primary diagnosis of bronchiolitis (Table 1). Mortality was highest in the historical CPAP group and lowest in the HFNT group. There were very few complications with respiratory support during the study, with no pneumothoraces and three reports of skin breakdown.

Table 1.

Patient characteristics

	Historical CPAP controls (n=12)	CPAP intervention (n=17)	HFNT intervention (n=29)	Total population (n=66)
Age in months, mean (SD)	10.0 (8.2)	7.7 (5.2)	5.9 (5.9)	7.9 (6.6)
Age in months, median (IQR)	8.4 (3.0-14.4)	6.6 (4.6-9.3)	3.4 (1.7-8.0)	6.2 (2.3-10.9)
Male sex, n (%)	6 (50)	11 (65)	16 (55)	37 (56)
Weight in kg, mean (SD)†	5.9 (2.9)	5.2 (1.4)	5.4 (2.5)	5.8 (2.5)
PRISM score, mean (SD)†	12.2 (5.0)	9.8 (8.7)	10.6 (4.8)	11.1 (6.1)
Primary diagnosis, n (%)
Bronchiolitis	1 (8)	2 (12)	10 (34)	13 (20)
Pneumonia	5 (42)	7 (41)	12 (41)	25 (38)
Other	6 (50)	8 (47)	7 (24)	38 (58)
Ventilator days before initial extubation, median (IQR)	3 (1-14)	5 (1-5)	6.5 (2-11)	5 (1-11)
Mortality, n (%)	3 (27)	2 (12)	2 (7)	10 (15)
Pneumothorax, n (%)	0 (0)	0 (0)	0 (0)	0 (0)
Skin breakdown, n (%)	0 (0)	2 (12)	1 (3)	3 (5)

Historical CPAP controls received post-extubation CPAP in 2012-2016. CPAP intervention received post-extubation CPAP in 2016-2017. NHFT intervention received post-extubation NHFT in 2016-2017. Total population includes all patients enrolled in the study, including those that received no post-extubation respiratory support. Standard deviation (SD) and interquartile range (IQR)

Following initial extubation, there was no difference in the need for mechanical ventilation between the children in the prospective CPAP (aHR 0.96, 95% CI 0.35-2.62) and HFNT (aHR 0.47, 95% CI 0.15-1.48) groups and the historical CPAP controls (Table 2, Figure 2A). The median time to remaining ventilator free was 11 days (95% CI 8-12) in the HFNT group compared with six days in the CPAP group (95% CI 3-10) and three days (95% CI 1-8) in the historical CPAP control group. Reintubation within 48 hours occurred more frequently in the historical CPAP controls (37.5%) than in the CPAP (8.3%) and HFNT (7.0%) groups. Patients had a lower risk of reintubation after HFNT implementation, regardless of the mode of respiratory support (HFNT: aHR 0.18, 95% CI 0.06-0.57 and CPAP: aHR 0.14, 95% CI 0.03-0.72; Table 2, Figure 2B). Patients who received CPAP were discharged sooner from the PICU than historical CPAP controls (aHR 3.42, 95% CI 1.43-8.19; Table 2, Figure 2C). There was no statistically significant difference in time to PICU discharge between patients who received HFNT and the historical CPAP controls (1.89, 95% CI 0.98-3.66). The median time to PICU discharge was 11 days (95% CI 7-18) in the CPAP group, 18 days (95% CI 14-24) in the HFNT group and 37 days (95% CI 15-101) in the historical control group. Patients who received CPAP had fewer days of intravenous sedation (3.0 days, 95% CI 2.0-5.5) than those who received HFNT (9.5 days, 95% CI 7.5-14.5) and the historical CPAP controls (10.0 days, 95% CI 4.0-25.0). When these data were adjusted for the PRISM score, there was no significant difference in the number of intravenous sedation days in the CPAP or HFNT group and the historical CPAP controls (p=0.06 and p=0.07).

Table 2.

Survival analyses comparing the effects of HFNT and CPAP with historical CPAP controls

	CPAP	HFNT
Cox proportional hazard regression, aHR (95% CI)*
Need for mechanical ventilation	0.96 (0.35, 2.62)	0.47 (0.15, 1.48)
Reintubation	0.14 (0.03, 0.72)	0.18 (0.06, 0.57)
PICU discharge	3.42 (1.43, 8.19)	1.89 (0.98, 3.66)
Multiple linear regression, Coefficient (95% CI)	8.35 (−42.77, 59.47)	14.88 (−29.91, 59.67)

^*Reference group included historical CPAP controls. Adjusted hazard ratio (aHR) adjusted for age, sex, PRISM score. No additional effect observed when adjusting for weight.

Figure 2.

Survival analyses comparing the effects of HFNT and CPAP with historical CPAP controls. Kaplan-Meier curves generated from Cox proportional hazard regression, adjusted for age, sex and PRISM score, are depicted in each panel. The log-rank test compared survival curves between groups. (A) instantaneous risk of needing mechanical ventilation. (B) instantaneous risk of requiring reintubation. (C) instantaneous risk of PICU discharge.

DISCUSSION

In this pilot study, we report the impact of post-extubation HFNT use on patient outcomes in a tertiary Peruvian PICU after HFNT was introduced in June 2016. Our findings represent the first report of HFNT use in paediatric patients in Peru and add to the growing body of literature that has evaluated HFNT use in resource-constrained settings. Given the small sample size, it was challenging to identify differences in outcomes between the groups, but this study demonstrates that HFNT implementation was feasible in a resource-constrained facility.

Although we found no difference in the need for mechanical ventilation following initial extubation, the HFNT group was ventilator-free for a longer median time. Multiple studies have shown reduced intubation rates in paediatric patients after HFNT implementation (11,12,26), but few have explored the effect of HFNT on ventilator-free days. A univariate analysis by Kawaguchi et al reported fewer mechanical ventilation days in the post-HFNT implementation group compared with pre-HFNT historical controls (5.9 vs 4.0; p<0.001), but this difference was not significant in the adjusted matched-pair analysis (27). These data suggest that the greatest effect of HFNT is in preventing intubation, which could explain the lack of effect in our post-extubation population.

In our study, all patients had a reduced risk of reintubation after HFNT implementation, regardless of their respiratory support modality. This is consistent with a study by Shioji et al that demonstrated a lower reintubation rate in infants after cardiac surgery receiving HFNT than historical controls receiving non-invasive ventilation (3% versus 17%, p=0.06) (28). The pre- and post-implementation study designs of the quoted studies suggest that temporal changes in care delivery, such as implementing a standardised extubation process in our study, may explain differences in reintubation rates. Other paediatric studies have shown no difference in extubation failure rates for preterm infants (29) or infants after cardiac surgery (30) who received HFNT or CPAP for post-extubation respiratory failure. Together, these data suggest that HFNT and CPAP may provide comparable support for post-extubation respiratory failure.

HFNT support did not change PICU length of stay compared with historical CPAP controls. Multiple studies have demonstrated no effect of HFNT on PICU length of stay (11,26,27). Shioji et al observed a decreased PICU length of stay for patients who received HFNT compared with historical controls who received noninvasive ventilation (10 vs 17, p<0.01). However, the significance of this finding was difficult to interpret due to the pre- and post-intervention study design (28). The PICU length of stay was reduced in the prospective CPAP group compared with historical CPAP controls in our study. This finding was interesting; however, this difference may be due differences in our research protocol for management of patients with CPAP outside the PICU. Surprisingly, we saw no effect on the duration of sedative infusions with the use of either respiratory support modality. This could be explained by the lack of difference in ventilator days or differences in sedation practices at the hospital. Given that apnoea has been observed with continuous sedatives and the lack of availability of dexmedetomidine, our physicians generally wean continuous sedative infusions while they are weaning mechanical ventilation support.

There were several limitations to our study. First, our small sample size limited the statistical power of the study, making it difficult to detect differences among the groups. The initial sample size calculations showed that 50 subjects in each group would be required in order to detect a difference in outcomes. We closed enrolment before we reached this goal because we were keen to expand the indications for HFNT use within the hospital to children with acute respiratory failure prior to intubation. An alternative study design could have evaluated the effect of HFNT use on intubation rates, but we would have had to train all staff in the emergency department and general paediatric wards, in addition to the PICU. This was not feasible.

The second limitation was that our primary outcome, the need for mechanical ventilation, could have been influenced by the pre-intervention period, given that the intervention occurred after the initial extubation. We had the same thought about the two secondary outcomes, which were the PICU length of stay and days of continuous sedative infusions. We chose our primary outcome because we hypothesised that HFNT implementation could reduce total ventilator days by encouraging providers to extubate sooner and reducing reintubation rates. These could then decrease PICU length of stay and the duration of continuous sedative infusions. However, if the majority of the ventilator days had occurred prior to our intervention, we would have been much less likely to detect an effect. In this respect, choosing the reintubation rate may have been a more appropriate primary outcome.

In addition, the respiratory support modality in the intervention group, HFNT or CPAP, was at the discretion of the treating physician. This increased selection bias in the treatment groups. This could explain the higher proportion of patients with bronchiolitis in the HFNT group and may have biased our results. Our local collaborators advocated for this design because some physicians were hesitant to adopt a new respiratory support modality, despite completing a comprehensive HFNT training programme before its introduction. Furthermore, subjects could transition between respiratory support modalities at the discretion of the treating physician. We performed our analysis based on the initial type of respiratory support, which could have influenced the results.

Notably, the mean age in all groups was less than one year, even though the inclusion criteria allowed us to enrol children up to five years of age. This discrepancy may be due to selection bias by the treating physicians, who assumed greater efficacy of non-invasive respiratory support for younger children. This makes our results less generalizable for older children. Finally, the pre- and post-implementation study design made it difficult to distinguish temporal differences in the standard of care from any changes related to HFNT implementation. The 15% difference in observed mortality between the pre- and post-intervention CPAP groups, as well as the four year historical control period, raise concerns that factors other than the availability of HFNT contributed to the observed outcomes. These need to be considered when interpreting our results.

CONCLUSION

This study showed that introducing HFNT into a resource-constrained facility was feasible and suggests that HFNT could provide effective post-extubation respiratory support for some children. As more facilities throughout the world consider introducing non-invasive respiratory support, it is essential that data on the clinical impact are collected and reported. These data will help the individual facility to improve its clinical processes. They will also enable the global community to learn from the experiences of others when implementing new technologies or processes. By sharing our data and experiences, we can all work together to develop sustainable strategies that provide effective interventions that reduce global child mortality.

Supplementary Material

sup FigS1

Figure S1. Survival analyses comparing the effects of HFNT and CPAP with historical CPAP controls. Kaplan-Meier curves generated from Cox proportional hazard regression, adjusted for age, sex and PRISM score, are depicted in each panel. Below the curves, the number of observations at each time point is listed to understand the number of censored observations. The log-rank test compared survival curves between groups. (A) instantaneous risk of needing mechanical ventilation. (B) instantaneous risk of requiring reintubation. (C) instantaneous risk of PICU discharge.

Key Notes.

• This study examined the impact of high-flow nasal oxygen therapy (HFNT) on clinical outcomes for children under five with post-extubation respiratory failure in a Peruvian paediatric intensive care unit (PICU).

• It compared patients who received initial HFNT in 2016-2017 with contemporary and historical continuous positive airway pressure groups.

• HFNT did not change the need for mechanical ventilation, but it did reduce the reintubation rate in children with post-extubation respiratory failure.

Abbreviations

CPAP

continuous positive airway pressure

HFNT

high-flow nasal oxygen therapy

PICU

paediatric intensive care unit

PRISM

paediatric risk of mortality