Acute hypoxemic respiratory failure is a major cause of hospital admission, intensive care unit (ICU) transfer, and mortality among critically ill patients.[1] It commonly occurs in pneumonia, acute respiratory distress syndrome (ARDS), pulmonary edema, and severe viral respiratory infections.[2]

 High-flow nasal cannula (HFNC) therapy has become an important noninvasive respiratory support modality because it delivers heated, humidified oxygen at high flow rates, improves oxygenation, reduces anatomical dead-space rebreathing, and enhances patient comfort.[3] However, HFNC does not provide adequate support for all patients, and approximately 20–40% of patients may experience treatment failure, depending on the underlying disease and clinical setting.[4] Delayed recognition of HFNC failure may result in the deterioration of the level of hypoxemia, emergency intubation, and higher mortality, making the use of reliable bedside tools that identify patients who need early escalation of respiratory support essential.[5]

The ROX index, calculated as the ratio of peripheral oxygen saturation (SpO₂) to the fraction of inspired oxygen (FiO₂), divided by the respiratory rate, is a respiratory-specific tool developed to predict HFNC failure.[6]  The National Early Warning Score 2 (NEWS2) is a standardised clinical scoring system and includes respiratory rate, oxygen saturations, need for supplemental oxygen, systolic blood pressure, pulse rate, temperature, and new onset of confusion or altered mental state.[7]

NEWS2 is a more comprehensive indicator of systemic physiological decompensation compared to the ROX index, which is used mainly to assess respiratory physiology.[8] However, NEWS2 is not a dedicated tool for forecasting HFNC failure, and there is still clinical relevance when comparing the performance of NEWS2 to ROX.[9] The NEWS2 score may also pick up other signs of deterioration that are not reflected in ROX, and that may drive a need for invasive mechanical ventilation.[8]

Although both indices are clinically relevant, there is limited data to support their relative ability to predict failure of HFNC across patient groups and healthcare settings. Identifying the index with maximum discrimination, sensitivity, and specificity may enhance bedside risk stratification, especially in resource-constrained hospitals where timely decisions for intensive monitoring and endotracheal intubation are critical. Therefore, this study aims to compare the predictive value of NEWS2 versus the ROX index for predicting HFNC failure in patients with acute hypoxemic respiratory failure, using receiver operating characteristic (ROC) curve analysis to compare their areas under the curve and determine optimal cutoffs with corresponding sensitivity and specificity. The findings may help identify a practical and reliable approach to early recognition of HFNC failure and support timely escalation of respiratory support.

This study was designed as a prospective comparative observational study and was conducted at United Medical and Dental College, Karachi, Pakistan, over six months, from 1st December 2025 to 31st May 2026. Adult patients presenting with acute hypoxemic respiratory failure and receiving high-flow nasal cannula (HFNC) therapy were evaluated to compare the predictive performance of the National Early Warning Score 2 (NEWS2) and the respiratory rate–oxygenation (ROX) index for identifying HFNC treatment failure. The sample size was calculated using the OpenEpi online sample size calculator for a single population proportion. A previous emergency department study involving patients with acute hypoxemic respiratory failure reported an HFNC failure rate of 29.5% among patients without COVID-19. Using an anticipated HFNC failure proportion of 29.5%, a 95% confidence level, and an absolute precision of 8%, the calculated sample size was approximately 125 patients.[10] After adding 10% to account for incomplete observations or missing data, the final target sample size was 138 patients. A non-probability consecutive sampling technique was used. All eligible patients presenting during the study period were assessed for inclusion, and patients who fulfilled the eligibility criteria were recruited consecutively until the required sample size was achieved. Patients aged 18 years or older, of either sex, who presented with acute hypoxemic respiratory failure and received HFNC therapy for respiratory support were included. Patients were eligible if baseline clinical parameters required for calculating NEWS2 and the ROX index were available and treatment outcomes could be documented. Patients who required immediate endotracheal intubation before meaningful assessment following HFNC initiation, those who had already received invasive mechanical ventilation, patients receiving HFNC solely for preoxygenation, and those with incomplete essential clinical data were excluded. Only the first eligible HFNC episode per patient was considered. After obtaining approval from the institutional ethical review committee and informed consent from eligible patients or their legally authorized representatives, demographic and clinical information was recorded using a structured data collection proforma. Demographic variables included age and sex, while clinical variables included the underlying cause of respiratory failure, comorbidities, respiratory rate, heart rate, blood pressure, temperature, peripheral oxygen saturation (SpO₂), fraction of inspired oxygen (FiO₂), oxygen requirements, and level of consciousness. The NEWS2 score was calculated using respiratory rate, oxygen saturation, supplemental oxygen requirement, systolic blood pressure, pulse rate, temperature, and consciousness level. The ROX index was calculated using the formula: ROX index = (SpO₂/FiO₂)/respiratory rate, with SpO₂ expressed as a percentage and FiO₂ expressed as a fraction. Both indices were recorded at predefined intervals, including baseline and 2, 6, and 12 hours after HFNC initiation, whenever the patient remained under observation. The primary outcome was HFNC failure, defined as the need for endotracheal intubation and invasive mechanical ventilation because of persistent or worsening respiratory failure. Patients who improved clinically and were successfully weaned from HFNC without intubation were classified as treatment successes. Decisions regarding intubation were made by the treating clinical team according to the patient's overall condition and established clinical indications, rather than solely on the basis of either study index. The collected data were entered and analyzed using IBM SPSS Statistics version 26. Continuous variables, including age, respiratory rate, oxygen saturation, FiO₂, NEWS2 scores, and ROX index values, were assessed for normality using the Shapiro–Wilk test. Normally distributed variables were presented as mean ± standard deviation, while non-normally distributed variables were expressed as median and interquartile range. Categorical variables, including sex, underlying diagnosis, comorbidities, and HFNC treatment outcome, were summarized as frequencies and percentages. Patients with HFNC success and failure were compared using the independent-samples t-test for continuous variables and the chi-square test for categorical variables. Receiver operating characteristic (ROC) curve analysis was performed to assess the predictive performance of NEWS2 and the ROX index for HFNC failure. The area under the curve (AUC), with 95% confidence intervals, was calculated for each index, and the areas under the two correlated ROC curves were compared using DeLong's test. Optimal cutoff values were determined using the Youden index, and sensitivity, specificity, positive predictive value, and negative predictive value were calculated at the selected cutoffs. A two-sided p-value of less than 0.05 was considered statistically significant.

A total of 138 patients with acute hypoxemic respiratory failure receiving high-flow nasal cannula (HFNC) therapy were included. The study population predominantly comprised older adults and males, with pneumonia being the most common underlying cause of respiratory failure. The most common comorbid conditions were hypertension and diabetes mellitus. HFNC therapy was successful in most patients, while approximately one-third required endotracheal intubation (Table 1).

Patients who failed the HFNC treatment showed a much higher degree of physiological deterioration compared to successful ones. Higher respiratory and heart rates, increased oxygen requirements, lower oxygen saturation, and reduced blood pressure were associated with treatment failure. In addition, patients who needed intubation had significantly higher NEWS2 scores and lower ROX indices, which reflect higher clinical severity (Table 2).

There were significant trends identified across the predictive indices for each of the treatment outcome groups in the serial assessment. NEWS2 scores progressively decreased among patients with successful HFNC therapy but remained elevated in those experiencing treatment failure. In contrast, ROX indices worsened in the intubation group, but showed improvement in the success group. These differences were statistically significant at all assessment intervals (Table 3).

Receiver operating characteristic analysis demonstrated good discriminatory performance for both NEWS2 and the ROX index in predicting HFNC failure. There was a gain in predictive accuracy as time went on, with the greatest AUC indicating at 12 hours. NEWS2 always showed a higher discriminatory power than the ROX index in all stages of measurements (Table 4).

The sensitivity, specificity, positive and negative predictive values of both indices were found to be progressively better with time in the diagnostic performance analysis. NEWS2 was numerically higher than the ROX index, especially at later assessment time points. Both scoring systems showed high negative predictive values, suggesting potential usefulness in identifying patients at lower risk of HFNC failure (Table 5).

The AUC values of the NEWS2 were always greater than the ROX index (p < 0.05, DeLong's test for all comparisons). These differences were not statistically significant at any assessment interval, however, and neither scoring system showed a statistically superior discriminatory performance for predicting HFNC failure (Table 6).

Table 1. Baseline Demographic and Clinical Characteristics of Study Participants (n = 138)

Table 2. Baseline Physiological Parameters and Clinical Scores According to HFNC Treatment Outcome

Table 3. Comparison of NEWS2 and ROX Index at Different Time Intervals According to HFNC Outcome

Table 4. Receiver Operating Characteristic (ROC) Curve Analysis of NEWS2 and ROX Index

for Predicting HFNC Failure

Table 5. Diagnostic Performance of NEWS2 and ROX Index at Selected Cutoff Values

Table 6. Pairwise Comparison of NEWS2 and ROX Index Predictive Performance Using DeLong's Test

The present study compared the accuracy of NEWS2 and the ROX index in predicting HFNC failure in patients with acute hypoxemic respiratory failure. These results indicated an overall HFNC failure rate of 29.7%, with patients who needed intubation having significantly higher NEWS2 scores but lower ROX indices. Both indices showed good discriminant properties, especially at later assessment intervals. Although NEWS2 consistently showed higher AUC values, the differences between the two indices were not statistically significant. The observed association between lower ROX values and HFNC failure aligns with Sancho et al. (2021), who conducted a prospective study of 85 patients with SARS-CoV-2-related respiratory failure. They reported an HFNC failure rate of 55.3% and identified the 24-hour ROX index as a useful predictor of treatment success, with an AUC of 0.826. The higher failure rate compared with the present study may reflect differences in patient characteristics and disease severity.[11] Likewise, a meta-analysis by Prakash et al. (2021) showed a pooled ROX AUC of 0.81 and sensitivity of 70% with a specificity of 79%, in a total of 1,301 patients. The results of the present study support the discriminatory ability of the ROX index seen in these cases, although the present results indicate that the discriminatory ability of the ROX index is better at later assessment times.[12] In a systematic review and meta-analysis published in Respiratory Research (2022) including 1,933 patients, a pooled ROX AUC of 0.75, with a pooled sensitivity of 67% and a pooled specificity of 72%, was reported. These values were lower than the later values observed in the present study, which may indicate that the predictive performance will vary depending on the timing of the assessment and the study population.[13] Lee et al. (2022) investigated the ROX index and SpO₂/FiO₂ ratio among patients with COVID-19 receiving HFNC across seven Korean hospitals. Their multicenter study confirmed the usefulness of respiratory oxygenation measurements to detect treatment failure in patients. This is consistent with the present finding that lower ROX indices were associated with subsequent intubation.[14] A 2023 retrospective study of COVID-19-associated ARDS showed the ROX index had an AUC of 0.89 for the failure of HFNC and was superior to each respiratory parameter. This is similar to the 12-hour ROX AUC seen in the present study and reinforces the importance of using a combined oxygenation and respiratory rate index.[15] In a meta-analysis of 24 articles and 4,790 patients, Yau et al. found a pooled ROX AUC of 0.771 with an optimal cutoff of 5.23. They also found that thresholds had varying optimal values between measurement intervals. This confirms the current finding that serial ROX may be more informative than using only a baseline measurement.[16] Kwon et al. (2024) compared the performance of the ROX index in patients with and without COVID-19 who presented with acute hypoxemic respiratory failure. They also noted the importance of the underlying disease etiology in the interpretation of ROX-based predictions. This is pertinent to the present study, which included patients with different causes of respiratory failure rather than a single disease category.[10] Zarama et al. (2024) externally validated NEWS2, the SpO₂/FiO₂ ratio, and the ROX index for predicting adverse outcomes among emergency department patients with COVID-19. They offer a comparison between systemic and respiratory-specific assessment tools that is relevant to their investigation. However, their outcome measures differed from the present study's primary endpoint of HFNC failure requiring intubation.[17] Yu et al. (2025) compared machine-learning models with established clinical indices for predicting HFNC outcomes. In their external validation cohort, the ROX index achieved an AUC of 0.74, while a support vector machine model achieved 0.79. The results show the moderate predictive power of the traditional indices and serve as a background for the present comparison between NEWS2 and ROX.[18] Gutta et al. (2025) investigated modified ROX indices incorporating heart rate, oxygen flow, and oxygenation measurements in patients receiving post-extubation HFNC. They investigated the possibility that other physiological variables would be useful in predicting outcome. This is important for the present investigation because NEWS2 uses several physiological parameters not used in the ROX index, but the patient populations and clinical indications are different.[19] The present results are overall consistent with previous studies indicating the utility of the ROX index to predict outcome in patients receiving HFNC. NEWS2 demonstrated numerically higher discriminatory performance, but the absence of statistically significant differences prevents a conclusion of superiority. These results should be interpreted with these caveats: the single-center design, the relatively small sample size, and the fact that fewer patients were assessed at the later time points. Further prospective studies are needed to establish whether NEWS2 provides additional predictive value over the ROX index. Limitations There are a few limitations to this study. First, the results may not be generalizable because of the single-center design and relatively small sample size. Secondly, clinical heterogeneity may have been introduced as patients with different underlying causes of acute hypoxemic respiratory failure were included. Third, there may have been differences in clinical treatment and decision-making about intubation that affected treatment outcomes. In addition, patients who were intubated early may not have been available for subsequent evaluation, which could impact the predictive utility of both indices at later time points. Furthermore, patients who did require early intubation may not have been available for subsequent evaluation, and this may impact the predictive utility of both indices at later time points. Finally, the study did not evaluate long-term clinical outcomes, including ICU length of stay and mortality.

The performance of both the NEWS2 and ROX indices was good in the prediction of HFNC failure in all patients with acute hypoxemic respiratory failure. NEWS2 demonstrated numerical advantage in discriminatory accuracy, sensitivity, and specificity, especially at later assessment time points. These changes in the two indices, however, were not statistically significant. Serial assessment using either index may support early identification of patients at risk of HFNC failure and facilitate timely clinical decision-making.

Anesi, G.L., et al., Association of ICU admission and outcomes in sepsis and acute respiratory failure. American Journal of Respiratory and Critical Care Medicine, 2022. 205(5): p. 520-528.

2. Mokrani, D. and J.-F. Timsit, Role of respiratory viruses in severe acute respiratory failure. Journal of Clinical Medicine, 2025. 14(9): p. 3175.
3. Nolasco, S., et al., High-flow nasal cannula oxygen therapy: physiological mechanisms and clinical applications in children. Frontiers in medicine, 2022. 9: p. 920549.
4. Pagliaro, R., et al., High-flow nasal cannula system in respiratory failure associated with interstitial lung diseases: a systematic review and narrative synthesis. Journal of Clinical Medicine, 2024. 13(10): p. 2956.
5. Saillard, C., et al., High-flow nasal cannula failure in critically ill cancer patients with acute respiratory failure: Moving from avoiding intubation to avoiding delayed intubation. PLoS One, 2022. 17(6): p. e0270138.
6. Soydan, E., et al., Impact of a nurse-driven high-flow nasal cannula weaning protocol on HFNC duration in infants with bronchiolitis: a randomized controlled trial. European Journal of Pediatrics, 2026. 185(4): p. 201.
7. Scott, L.J., et al., Does the use of National Early Warning Scores (NEWS or NEWS2) in healthcare settings improve patient outcomes: a systematic review. Systematic Reviews, 2026. 15(1): p. 73.
8. Prower, E., et al., The ROX index has greater predictive validity than NEWS2 for deterioration in Covid-19. EClinicalMedicine, 2021. 35.
9. Singh, R., et al., ROX index: A non-invasive tool in monitoring and guiding oxygen therapy in critically ill patients-A narrative review. Trends in Anaesthesia and Critical Care, 2022. 47: p. 15-19.
10. Kwon, H., et al., Respiratory rate‑oxygenation (ROX) index for predicting high-flow nasal cannula failure in patients with and without COVID-19. The American Journal of Emergency Medicine, 2024. 75: p. 53-58.
11. Ferrer, S., et al., ROX index as predictor of high flow nasal cannula therapy success in acute respiratory failure due to SARS-CoV-2. Respir Med, 2021. 189: p. 106638.
12. Prakash, J., et al., ROX index as a good predictor of high flow nasal cannula failure in COVID-19 patients with acute hypoxemic respiratory failure: A systematic review and meta-analysis. J Crit Care, 2021. 66: p. 102-108.
13. Junhai, Z., et al., The value of ROX index in predicting the outcome of high flow nasal cannula: a systematic review and meta-analysis. Respiratory Research, 2022. 23(1): p. 33.
14. Kim, J.H., et al., ROX index and SpO2/FiO2 ratio for predicting high-flow nasal cannula failure in hypoxemic COVID-19 patients: a multicenter retrospective study. PLoS One, 2022. 17(5): p. e0268431.
15. Poopipatpab, S., et al., The use of respiratory rate-oxygenation index to predict failure of high-flow nasal cannula in patients with coronavirus disease 2019-associated acute respiratory distress syndrome: A retrospective study. PLoS One, 2023. 18(6): p. e0287432.
16. Yau, C.E., et al., Performance of the ROX index in predicting high flow nasal cannula failure in COVID-19 patients: a systematic review and meta-analysis. Critical Care, 2023. 27(1): p. 320.
17. Zarama, V., et al., NEWS2, S/F-ratio and ROX-index at emergency department for the prediction of adverse outcomes in COVID-19 patients: An external validation study. The American Journal of Emergency Medicine, 2024. 83: p. 101-108.
18. Yu, H., et al., Machine learning models compared with current clinical indices to predict the outcome of high flow nasal cannula therapy in acute hypoxemic respiratory failure. Crit Care, 2025. 29(1): p. 101.
19. Gutta, S., et al., Comparison of ROX index with modified indices incorporating heart rate, flow rate, and PaO2/FiO2 ratio for early prediction of outcomes among patients initiated on post-extubation high-flow nasal cannula therapy. European Journal of Medical Research, 2025. 30(1): p. 166.