Objective: To assess a comparative analysis of the ROX index, Wood-Downes-Ferrés score (WDF), p-ROXI, and the SpO₂/FiO₂ ratio as predictors of high-flow nasal cannula (HFNC) failure in children hospitalized for bronchiolitis.

Methods: Data were extracted from the clinical trial “Comparison between HFNC and NIV in children with acute respiratory failure caused by bronchiolitis” conducted at a tertiary Brazilian hospital (Emergency Department and PICU). The inclusion criteria were children under 2 years of age admitted for bronchiolitis who developed mild to moderate respiratory distress and were eligible for HFNC therapy. Performance was determined by ROC and AUC metrics to define the best sensitivity and specificity for each variable. Children were evaluated at 0 h, 2 h, 6 h, 12 h, 24 h, 48 h, 72 h and 96 h after HFNC therapy initiation.

Results: A total of 126 patients were recruited for this analysis. The median age was 3 months. Ninety-one percent of the patients had an identified viral agent, with RSV being the most common (65%). Twenty-three percent (29/126) of patients experienced failed HFNC therapy and required mechanical ventilation. The best cutoff points at 12 hours were 4.5 for WDF (AUC=0.83, 0.74-0.92), 8.8 for ROX (AUC=0.7, 0.54-0.84), 1.45 for p-ROXI (AUC=0.56, 0.38-0-74), and 269 for SpO₂/FiO₂ (AUC=0.64, 0.48-0.74). The scores and indices were also correlated with the PICU and hospital LOS.

Conclusions: The ROX index and WDF were the most accurate scores for assessing HFNC failure considering 12-hour cutoff points. Trial registration number: U1111-1262-1740; RBR-104z966s. Date of registration: 03/01/2023.

The following data were extracted from the clinical trial titled “Comparison between high-flow nasal cannula and noninvasive ventilation (NIV) in children with acute respiratory failure caused by bronchiolitis”. The primary analysis is under structuring for publication. In the present study, patients with a diagnosis of mild to moderate acute respiratory failure due to bronchiolitis were randomized into two groups according to sealed envelopes (HFNC or NIV). Those who experienced therapy failure were intubated. In an interim analysis, the failure rate was 32% (17 of the 52 randomized patients). Therefore, we estimated that a sample of at least 87 patients would be representative of this population for the current study, with a margin of error of 10% and a 95% confidence interval. This study was approved by the Institutional Review Board (IRB) of the hospital (Hospital Municipal Infantil Menino Jesus Research Ethics Committee; approval number 39509820.0.0000.5639, 11/24/2020; trial registration number: U1111-1262-1740; RBR-104z966s, 03/01/2023). Despite clinical trial submission had been occurred before patient randomization, its registration was approved later because some formatting and data adjustments were required. All procedures were performed in accordance with the ethical standards of the responsible committee and the Declaration of Helsinki. Written informed consent was obtained from the children’s parents or guardians before the data were collected and was applied by the physician responsible for starting the protocol. The STROBE guidelines were applied for text structuring considering the score comparison.

Local training was carried out with physicians and physiotherapists before and during the research for device usage and protocol application. Training efficacy evaluation was not performed.

The HFNC system used was an Airvo 2 (Fisher & Paykel), and the initial flow was 2 L.kg.min. FiO₂ was titrated to maintain a peripheral oxygen saturation between 94 and 99%. All eligible patients who developed mild to moderate respiratory distress (classified by WDF score) were connected to a nonrebreathing mask until the device was properly installed.

The inclusion criteria were children under 2 years of age admitted for bronchiolitis that progressed to mild to moderate respiratory distress (WDF < 8) during hospitalization [27,28]. The exclusion criteria were patients with severe respiratory distress, bronchopulmonary dysplasia, cyanotic congenital heart disease or hemodynamic repercussion; liver disease; neuromuscular disease; or tracheostomy. Infants with a viral panel positive for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and whose parents refused to participate in this study were also excluded.

The ROX index, WDF score, SpO₂, FiO₂ and vital signs were registered at 0 h, 2 h, 6 h, 12 h, 24 h, 48 h, 72 h and 96 h after HFNC therapy initiation. The variables assessed included age, weight, sex, comorbidities, viral panel identified by reverse transcription polymerase chain reaction (RT‒qPCR), time on HFNC, mechanical ventilation duration, PICU length of stay (LOS), hospital LOS, and mortality. Data in S1 Table and S1 Appendix provide a description of the variables applied in each score.

The criteria for indicating HFNC failure and mechanical ventilation requirements were signs of severe respiratory distress (WDF 8 to 14), a respiratory rate > 60 bpm (in children up to 1 year old) or > 40 bpm (in children 1 to 2 years old), or a heart rate > 160 bpm. For patients who experienced respiratory improvement, weaning therapy was administered according to the institutional protocol.

As a primary outcome, the ROX index, WDF, SpO₂/FiO₂ and p-ROXI, which are predictors of HFNC failure,were evaluated. The secondary outcomes analyzed were mechanical ventilation duration, PICU LOS, hospital LOS, and mortality.

The descriptive statistics included measures of central dispersion (means and standard deviations, medians, and interquartile ranges), in addition to absolute and relative frequencies.

The Mann‒Whitney U test was used to compare distributions. To evaluate correlations, the Spearman test was used, and to compare frequencies, the chi-square test was used. Receiver operating characteristic (ROC) curves and the Youden index were used to identify the best cutoff points for predicting HFNC therapy failure and determining thesensitivity and specificity of the cutoff points. The ROC curves were compared through the DeLong test. Generalized estimating equations (GEEs) were applied to analyze repeated measures of the ROX index and WDF. Statistical analysis was performed with R (Vienna, Austria).

The p-ROXI was calculated according to Yildizdas (SpO₂/FiO₂/respiratory rate z score for age). Respiratory rate z scores for age were obtained from Sepanski et al [16,30].

The interrater reliability of the WDF was not evaluated. The standard method was applied for missing data, which is the analysis of complete cases, not using any technique of replacement or imputation [31]. The percentages of missing data for each period were 0 h (0), 2 h (0.2%), 6 h (0.8%), 12 h (0.9%), 24 h (0.1%), 48 h (0.7%), 72 h (0) and 96 h (0). Missing data occurred as a result of incomplete forms.