A prospective cohort study was carried out between January 2015 and August 2019 at the PICU of Hospital Víctor Ríos Ruiz (Los Ángeles, Chile). The study was approved by the Scientific – Ethics Committee of the Biobio Health Department, and informed consent was requested from the legal guardians of the patients.

The study was carried out in critical pediatric patients between 0–15 years of age presenting acute lower respiratory tract infection and requiring intubation due to failure of NIV. Patients with the following diagnoses were included: acute bronchiolitis defined as an acute respiratory infection in infants under two years of age, with rhinitis, tachypnea, wheezing, cough, crepitants, accessory muscle use and/or nasal flaring12; asthma crisis in children defined as progressive respiratory worsening due to asthma, with possible cough, wheezing, dyspnea, labored breathing and anxiety13; and pneumonia defined by the presence of fever, tachypnea, labored breathing, rhonchus, crepitants and wheezing.14 We excluded patients with cyanotic congenital heart disease, nocturnal NIV users, patients with neurological damage, and tracheostomized individuals. The sample was established based on a non-probabilistic sampling of consecutive cases.

The Unit has a protocol that was applied throughout the study (Appendix B Fig. 1 of the electronic supplementary material). The decision to start NIV was based on the presence of labored breathing (increase in respiratory frequency >2 standard deviations of the normal range for the age of the patient, and signs of moderate to severe dyspnea according to the modified Woods scale >4 points15) and/or FiO2 >50% to maintain SatO2 >94%16, despite standard medical care.

Figure 1.

Kaplan–Meier analysis of the duration of invasive mechanical ventilation (IMV) following the failure of noninvasive mechanical ventilation (NIV). A) The curves are compared according to the NIV times (≤6 h, 7–24 h and >24 h) prior to the use of IMV, yielding a log rank p-value of 0.479. B) In the subgroup with pediatric acute respiratory distress syndrome (PARDS), the log rank p-value was 0.830. C) In the subgroup without PARDS, the log rank p-value was 0.086.

(0.23MB).

We used BiPAP Synchrony (Respironics®, Pittsburg, PA, USA) and Trilogy 202 ventilators (Respironics®, Pittsburg, PA, USA) with an active humidification system (MR 850 Fisher Paykel Healthcare, Auckland, New Zealand). Performax facial and Comfort Full oronasal masks were used (Respironics®, Pittsburg, PA, USA). Appendix B of the supplementary material describes the suggested initial parameters according to the protocol of the Unit. Chloral hydrate and/or midazolam were used as sedatives to secure good adaptation. Brief disconnections (<20 min) were made every 4−6 hours for kinesiological and nursing care, and to decompress pressure points of the interface. Routine arterial puncture to assess oxygenation was not performed; instead, we used the SpO2/FiO2 (SF) index (employing SpO2 values between 88–97%)17; arterial or venous blood gas measurements were reserved for patients with partial response (unsatisfactory reduction of respiratory effort).

Failure of NIV was defined as the need for intubation – the latter being considered when labored breathing or dyspnea worsens (as evidenced by the modified Woods clinical scale) or in the presence of SpO2 <85% despite high ventilatory support. High assist was considered as an inspiratory positive airway pressure (IPAP) ≥20 cmH2O, continuous positive airway pressure or expiratory positive airway pressure (CPAP/EPAP) ≥10 cmH2O.

The following variables were recorded for all patients included in the study: age, weight, gender and Pediatric Index of Mortality II (PIM2) upon admission18; diagnosis established by the supervising pediatrician; use of antibiotic treatment, indicated by the pediatrician due to bacterial superinfection (suspected or confirmed)14; SF index17; diagnosis of PARDS, based on the criteria of the Pediatric Acute Lung Injury Consensus Conference (PALICC) for patients on NIV using the SF index19; ventilatory strategy (ventilator, interface, mode, maximum expiratory and inspiratory pressure); reason for the failure of NIV, defined as complications related to NIV (intolerance, air leakage, pressure ulcers, etc.), persistent labored breathing (reduction of respiratory frequency and insufficient or inexistent modified Woods score) and/or hypoxemia (SF <150).

For the purpose of analysis, we divided the cohort according to the duration of NIV before intubation, stratifying the patients according to Mayordomo-Colunga et al.5 into ≤6 hours, between 7−24 hours and >24 hours. In relation to monitoring once intubated, we recorded the duration of IMV (in hours), PICU and in-hospital length of stay (in days), and in-hospital mortality.

Data collection was carried out by kinesiologists of the Unit, who were in charge of recruitment and follow-up of the participants that met the screening criteria. Recording was made using an ad hoc form during follow-up and until hospital discharge or death.

Normal data distribution was checked using the Kolmogorov–Smirnov test. The variables were reported as the median and interquartile range (IQR). Categorical variables were presented as absolute values and percentages. The comparison of continuous variables between groups was carried out using the Kruskal–Wallis test with Bonferroni correction, while categorical variables were compared using the chi-square test (χ2) or Fisher’s exact test (F) in the case of fewer than 5 elements per cell.

In relation to the hours of IMV, time to the event was estimated using the Kaplan–Meier method, taking the hours of IMV as time and extubation as event. Comparison of the time to event curves was made using the log-rank test. The association of risk of increased duration of IMV was evaluated by means of the Cox proportional hazards model, reporting the hazard ratio (HR), p-value and 95% confidence interval (95% CI). The HR was interpreted as the risk of increasing the duration of IMV, and the analysis was adjusted for PARDS. Statistical significance of the log-rank test and HR was considered for p < 0.05.

Considering that there is no evidence on the duration of IMV in patients with ARF presenting failed NIV, the statistical power was calculated on a post hoc basis with an alpha error of 5% for the duration of IMV compared with the duration of IMV in the studies of Payen et al.7 and Morris et al.20

A total of 109 patients with failed NIV were included in the present study, out of a total of 470 subjects (23.2%) that received NIV during the recruitment period. The main cause of ARF was pneumonia (90%), with respiratory syncytial virus (RSV) being the most frequent etiological agent (70.6%). In 89% of the cases use was made of antibiotics in the context of bacterial superinfection (suspected or confirmed). In turn, 25.8% of the patients presented some chronic disease – the most frequent condition being asthma (66.6%). The main reason for starting NIV was increased respiratory effort, in 84.4% of the cases. The median PIM2 was 3.4% (2.8−3.8%). The diagnosis of PARDS was established in 37.6% of the cases. The Trilogy 202 ventilator with full-face mask and bi-level mode was used in 90.5% of the cases. The median maximum inspiratory (IPAP) and expiratory pressures (EPAP) were 14 cmH2O (IQR 14–15) and 7 cmH2O (IQR 6–7), respectively. In 74 cases (67.9%) the failure of NIV was attributed to persistent or increased breathing difficulty. Furthermore, in 32 subjects (28.4%) the decision to intubate was due to mixed factors (hypoxemia and respiratory effort). Regarding the duration of NIV before failure, 89.9% of the patients failed within the first 24 hours of ventilation. Table 1 shows the characteristics of the study sample.

The median duration of IMV in the global study sample was 105 hours (84–142). No association was found between the duration of IMV and the duration of NIV before intubation (log-rank test, p = 0.479). Fig. 1A shows the time to event curve for the suspension of IMV, which compares the curves of the three NIV failure times. The presence of PARDS was identified as a risk factor for a longer duration of IMV (Table 2). To avoid possible confusion in the evaluation of the main causal relationship evaluated, we adjusted a multivariate Cox proportional hazards model in which the effect of the time on NIV upon the duration of posterior IMV was controlled for by the presence of PARDS. In this model, the duration of prior NIV showed no statistical significance (Table 3).

To complement the analysis, we used the presence of PARDS as control variable, considering that it presents a strong correlation to NIV failure8 and that in our sample we identified an association to a longer duration of IMV (Table 2); as a result, the analysis was performed in the PARDS (Fig. 1B) and No PARDS subgroups (Fig. 1C). However, the results showed no association between the duration of IMV and the duration of NIV before intubation in either the PARDS subgroup (log-rank test, p = 0.830) or the No PARDS subgroup (log-rank test, p = 0.086). Nevertheless, as can be seen in Fig. 2, the duration of IMV was greater in the PARDS subgroup than in the No PARDS subgroup (log-rank test, p = 0.0386).

Figure 2.

Kaplan–Meier analysis of the duration of invasive mechanical ventilation (IMV) following the failure of noninvasive mechanical ventilation (NIV) according to the presence or not of pediatric acute respiratory distress syndrome (PARDS).

(0.12MB).

The median UCI stay was 6 days (5–8), while the median in-hospital stay was 13 days (11–16). The stay in both the UCI (p = 0.25) and in hospital (p = 0.67) was not associated to the IMV connection times.

For the post hoc statistical power estimate, we included the IMV times recorded in this study (118.3 ± 53.97 hours) and those reported in the literature (in two similar studies): 182.3 ± 556.7 hours7 and 4 days (2–7)20, for an alpha error of 5%, yielding a power of 100% and 99.1%, respectively.