Introduction

Bremains a pivotal concern in pediatric healthcare, affecting children below the age of two [1]. Bronchiolitis stands as the leading cause of hospitalization among infants younger than one year old, constituting approximately 100,000 hospitalizations annually in the United States. The economic impact of this burden is substantial, with estimated costs totaling $734 million. This statistic underscores the significant strain bronchiolitis places on pediatric healthcare resources and highlights the urgent need for effective interventions to alleviate its impact [2].Despite evolving guidelines advocating streamlined care and resource optimization, the burden persists, with approximately 1-3% of cases necessitating hospital admission [3]. Efforts to curtail unnecessary healthcare services, such as chest radiography, have demonstrated limited success [4]. Furthermore, the escalating trends in intensive care unit (ICU) utilization and associated costs underscore the pressing need for novel interventions [5].

The emergence of High-Flow Nasal Cannula (HFNC) as a respiratory support modality injected renewed hope into the landscape of bronchiolitis management [6-8]. Initial expectations centered around reduced pediatric ICU (PICU) admission rates and improved cost-effectiveness [6]. However, pivotal randomized trials, including studies by Kepreotes et al., Franklin et al., and Durand et al., yielded inconclusive evidence regarding HFNC's superiority over standard oxygen therapy [9-11]. Despite the lack of clear-cut superiority, HFNC gained popularity in pediatrics due to its ease of use and excellent tolerability by children, contrasting with other non-invasive ventilation methods such as Continuous Positive Airway Pressure (CPAP) [11].

In the study conducted by Granda et al., which evaluated nine clinical scales, including three newly introduced ones, to predict relevant outcomes in pediatric bronchiolitis patients in an emergency department setting, it is relevant to highlight that no significant differences in performance were identified among the scales. The results indicated an overall uniformity in the ability to predict relevant outcomes, with no single scale statistically outperforming the others. These innovative findings provide important insights for clinical practice, suggesting that, in the context of bronchiolitis, various scales, including the Wood Downes Ferres score, demonstrate comparable performance in predicting relevant clinical outcomes [12].

In light of this context, our study posits a compelling hypothesis: given time for learning and adaptation, HFNC could unveil previously unrealized benefits, particularly in terms of reducing hospital or PICU length of stay. Additionally, we delve into the immediate effects of HFNC on respiratory work within the initial six hours of therapy. Our investigation scrutinizes pre and postHFNC heart and respiratory rates, as well as the impact on the Wood-Downes-Ferres score, aiming to uncover nuanced insights into the broader implications of HFNC in pediatric bronchiolitis care.

Methods

We conducted an analysis of data from patients needing to be hospitalized in Santa Catarina Hospital, a prominent private tertiary general hospital located in the heart of São Paulo, Brazil. Santa Catarina Hospital boasts a total of 352 hospital beds, including 58 pediatric ward beds and 26 Pediatric Intensive Care Unit (PICU) beds, making it a vital healthcare institution in the region during the seasonal period. The study included information on children aged 24 months or less diagnosed with bronchiolitis according to the criteria of the American Academy of Pediatrics [1,13]. Exclusion criteria comprised congenital heart disease, initial diagnosis of bacterial pneumonia, or incomplete data in medical records. The study received ethics approval (National Research Ethics Commission certificate number 41744620.5.0000.0068), and a waiver of the consent form was granted due to the retrospective nature of the study.

Data were retrospectively collected by reviewing the MV software (an electronic health medical record system) and the admission books of both the PICU and pediatric ward. In the primary data analysis, interrupted time series were constructed for key outcomes, including the rate of patients needing to be hospitalized in the ward or PICU, ventilation time, PICU, and hospital stays.

"Exposure" was defined as the commencement of High Flow Nasal Cannula (HFNC) use. Prior to exposure, protocol training for the initiation and weaning of HFNC with the healthcare team was conducted in the beginning of 2017 (the first three months of the year, before exposure to HFNC), including the use of the WoodDownes-Ferres score for objective stratification of disease severity before using HFNC. The pre-exposure period encompassed the years 2015 and 2016, preceding the implementation of HFNC. March 2017 marked the initiation of exposure. The analysis extended until the end of the seasonality of each year from 2017 to 2019. Time series were constructed considering the months of seasonality as time points (from the beginning of March to the end of August). We did not study the years 2020 to 2022 due to the coronavirus pandemic. Autoregressive Integrated Moving Average (ARIMA) models were built, taking into account trends and autocorrelations, to estimate the effect of the intervention [14].

In the other arm of the study, we analyzed measurements of respiratory and heart rates, as well as Wood-Downes-Ferres scores, for patients who received HFNC immediately before and 6 hours after the start of therapy. The Wood- Downes-Ferres score categorizes bronchiolitis severity as mild (1–3 points), moderate (4–7 points), or severe (8–14 points) [15]. To describe the standardized mean differences of the effects, we used the Hedges' g test, where "g" between 0.2 and 0.5 indicates a small effect, values between 0.5 and 0.8 denote a moderate effect, and values above 0.8 suggest a large effect. These values are negative when the direction of the effect is a reduction. To compare measurements before and after HFNC, we utilized the paired T-test. Statistical analyses were performed using SPSS software (IBM Corp., Armonk, NY), and R (R Foundation for Statistical Computing, Vienna, Austria)

Results

Patient Demographics

A retrospective analysis encompassing 626 hospitalized bronchiolitis patients from 2015 to 2019 reveals distinctive patient profiles, with annual case distributions as follows: 106 (2015), 76 (2016), 157 (2017), 144 (2018), and 143 (2019). Notable characteristics include a 9.4% prevalence of prematurity history, 1.5% Broncho dysplasia incidence, and 13.8% recurrent wheezing. Respiratory syncytial virus (RSV) predominated, with consistently high positive test rates (Table 1).

Table 1: Demographic data and other patient characteristics IQR– interquartile range 27-75. Data are in numbers/percentages, except where indicated.

ARIMA Models and HFNC Impact

Utilizing ARIMA models, no discernible effect was noted on the rates of non- invasive and invasive mechanical ventilation post the introduction of high-flow nasal cannula (HFNC). An initial increment in ICU length of stay in 2017 by 2.18 days subsided in subsequent years. Conversely, a reduction in hospital median stay emerged, estimated at 1.4 days (2017), 1.8 days (2018), and 2.6 days (2019). The introduction of HFNC correlated with an initial 13.3% increase in PICU admissions in 2017, followed by a notable 28% reduction in 2019 (Table 2A, 2B) (Figure 1 and 2).

Table 2A: ARIMA model for NIV Phase percentages: level effect.

Table 2B: ARIMA model for MV phase percentages.

Figure 1: percentage of patients undergoing mechanical ventilation and HFNC in the years. Panel A shows comparisons between the frequencies of the years in relation to the year 2019, the last year of the series, using the chi-square test.

Figure 2: percentage of cases that required HFNC and progressed to mechanical ventilation in the years 2017 to 2019, with a trend line showing a reduction in the percentage (p = 0.03 by chi-square, comparing the frequencies of 2019 and 2017)

Respiratory and Heart Rate Changes with HFNC

Analysis of respiratory and heart rates, pre and post 6 hours of HFNC initiation in 98 patients (2017-2019), yielded significant findings. Pre-HFNC mean respiratory rates of 47.3 (±12) decreased significantly to 37.8 (±10.3) post- HFNC (p < 0.0001, paired T-test), exhibiting a substantial effect size (Hedges’ g = -0.83). Similarly, pre-HFNC mean heart rates of 145 (±20) decreased to 138 (±20) post-HFNC (p = 0.004, paired T-test) with a smaller effect size (Hedges’ g = -0.34) (Figures 3,4,), (Tables 3, 4).

Figure 3: Violin graph highlighting the averages (red dots) of respiratory frequencies of 98 patients, before and after 6 hours of HFNC, with a significant reduction (p <0.0001 by paired T test).

Figure 4: Violin graph highlighting the averages (red dots) of the heart rates of 98 patients, before and after 6 hours of HFNC, with a significant reduction (p = 0.004 by paired T test).

Table 3: ARIMA model for median ICU length of stay. Phase: level effect.

Table 4: ARIMA model for median hospital stay times.

Wood-Downes-Ferres Score

Pre-HFNC scores (Mean: 5.3, SD: 1.24) significantly differed from post-HFNC scores (Mean: 4.3, SD: 1.32) (p < 0.0001, T-test). The effect size, as determined by Hedges’ g, was substantial at -0.79, signifying a moderate to large effect. (Figure 5)

Figure 5: Score Wood-Downes- Ferris.