Our study showed that S/F, a noninvasive continuous monitoring variable, might be a good predictor for HFNC outcomes in children with AHRF. We created a nomogram to serve as a shortcut prediction tool for HFNC failure using S/F as a variable at initiation and 2 h after HFNC implementation, as well as the presence of hemato-oncologic disease.
Multiple studies have shown that S/F has a good correlation with P/F in patients with respiratory failure [10, 16]. Our study showed similarly consistent results with a good correlation between S/F and P/F. Furthermore, we showed that S/F had a better predictive power for HFNC failure than P/F. The best predictive S/F cutoff at HFNC initiation was 230 in our study; this was higher than that reported in a previous study in which a S/F cutoff of < 195 during the first hour of treatment was associated with HFNC failure . While the aforementioned study included patients with cardiac comorbidities, we excluded children with congenital heart diseases because they have distinct S/F levels. We acknowledge that our inclusion criteria might have led to different S/F cutoff levels for the prediction of HFNC failure. Fine-tuning of the S/F cutoff is essential to achieve an excellent prediction power for HFNC failure. Accordingly, we used a previously reported therapeutic goal for S/F and combined it with our initial S/F cutoff to create a categorical variable . Finally, a S/F cutoff of < 230 at initiation and < 200 at 2 h was observed to have a remarkable prediction power (OR, 13.067; 95% CI 5.06−35.84).
An emerging issue for HFNC implementation in patients with AHRF is the concern of delayed intubation, which might worsen the clinical deterioration [18, 19]. Therefore, timely and appropriate identification of HFNC failure is crucial. Several indices, such as P/F and S/F, have been reported to be predictors for HFNC outcomes [14, 20]. The RR oxygenation index (the ratio of SpO2/FiO2 to RR) has recently been proposed to be a better predictor for HFNC failure than S/F alone in adults [3, 21]. However, the RR oxygenation index is difficult to apply in children with AHRF due to the variability of RR with age in children. Our categorical S/F variable may help clinicians decide whether endotracheal intubation should be performed within 2 h, which, in turn, would prevent delayed intubation.
Our study showed that the presence of an underlying hemato-oncologic disease was independently associated with HFNC failure, suggesting the deleterious effect of such a disease on HFNC outcomes. Our findings support those of a previous study which reported that HFNC neither improved discomfort nor decreased the need for intubation in patients with hemato-oncologic diseases . In our study, 70% of patients with hemato-oncologic diseases in the HFNC failure group had a severe AHRF with a P/F of 150 mm Hg at HFNC initiation, and pneumonia was the cause of AHRF in all patients with hemato-oncologic diseases. This result parallels that of a previous study, which showed that the etiology of AHRF (pneumonia, OR 11.2) was a significant risk factor for HFNC failure . HFNC failure in children with hemato-oncologic diseases might lead to various clinical conditions, complications, and problems unrelated to AHRF . Further, the conditions associated with the hemato-oncologic diseases might not be influenced by the mode of oxygen delivery . Moreover, supporting evidence has shown that the time needed to improve oxygenation during AHRF might be longer in patients with hemato-oncologic diseases than in other patients . These findings may explain why the presence of underlying hemato-oncologic disease was identified as an independent parameter for HFNC failure in our study. As such, HFNC in patients with hemato-oncologic diseases and AHRF should be monitored with more caution.
Our study is the first to build a nomogram that predicts HFNC failure in children with AHRF. With the help of our nomogram, which was constructed using a combination of S/F and hemato-oncologic disease as predictors, clinicians may estimate the individual probability of an HFNC outcome in a patient without the need for an invasive examination. This, in turn, may help clinicians make a timely decision for intubation. Furthermore, we included both internal and external validation procedures, which demonstrated strong discrimination and calibration. With the ability to estimate individual risk in an easy to use and straight forward manner, we believe that our nomogram has an advantage over simple predictors.
Our results should be interpreted with caution, as six patients who required escalation to NIV were not assessed. NIV was actively implemented during the middle of the study period; consequently, patients receiving NIV were excluded to maintain the homogeneity of the study. We also acknowledge the inclusion of measurements in the analysis that were performed with > 97% SpO2, where the oxyhemoglobin dissociation curves might have been unchanged . However, children with AHRF who receive appropriate oxygen therapy have been shown to have an SpO2 > 97% . While additional real-world clinical evidence in children with AHRF is necessary, it was reasonable to include patients with > 97% SpO2 in the present study to reflect current practice. A good correlation between S/F and P/F using data with S/F > 97% has also been demonstrated, which is consistent with our results .