Clinical background
High-flow nasal cannula (HFNC) therapy is now established for hypoxaemic respiratory failure, but a central concern is delayed intubation: patients who will not manage on HFNC deteriorate while waiting, and late intubation is associated with increased mortality. At the same time, it is not clinically obvious at an early stage which patients will succeed and which will need invasive ventilation. Oxygenation alone is an unreliable marker, since HFNC can keep the SpO₂ acceptable even as the work of breathing and exhaustion increase. The ROX index combines oxygenation and respiratory rate into a simple measure that can be obtained at the bedside without an arterial blood gas, and aims to distinguish early those patients likely to succeed on HFNC from those who require escalation.
Calculating the ROX index
ROX is calculated as the ratio of the oxygenation index to the respiratory rate:
where SpO₂ is given as a percentage, FiO₂ as a fraction (for example 0.40 for 40%) and the respiratory rate in breaths per minute. A higher value indicates better oxygenation relative to the work of breathing and argues for success with HFNC.
The index was derived by Roca et al. in a prospective multicentre cohort of 191 patients with acute hypoxaemic respiratory failure due to pneumonia treated with HFNC [1]. The cohort was assembled over two years at several centres in Spain and France. The outcome was the need for intubation, which occurred in 68 patients (35.6%). The authors used Cox proportional hazards models to identify the association of ROX with the outcome and found that its predictive ability increased over time after starting HFNC: the AUC was 0.679 at 2 hours, 0.703 at 6 hours and 0.759 at 12 hours [1]. A ROX ≥ 4.88 at all three time points was consistently associated with a lower risk of intubation, with hazard ratios between 0.29 and 0.43. To identify patients at high risk of failure, lower thresholds were proposed: < 2.85 at 2 hours, < 3.47 at 6 hours and < 3.85 at 12 hours [1].
Interpretation in practice
ROX should be measured repeatedly after starting HFNC, not as a one-off assessment. The calculator uses the validated success threshold of ≥ 4.88, but in clinical use three risk zones should be considered, with time-specific lower limits for failure from the derivation study [1]:
| ROX at 12 hours | Interpretation | Clinical action |
|---|---|---|
| ≥ 4.88 | Low risk of intubation | Continue HFNC, reduce FiO₂ gradually, continue monitoring |
| 3.85 to 4.87 | Intermediate risk | More frequent assessment, consider an arterial blood gas, prepare for escalation |
| < 3.85 | High risk of intubation | Prepare for intubation, involve intensive care, avoid further delay |
A rising ROX over time is reassuring. A falling or persistently low value, particularly below the time-specific lower limit, indicates that HFNC is not sufficient and that escalation should be actively considered. At 2 hours the index is still weakly predictive (AUC 0.679 in the derivation cohort) and should be interpreted with caution; at 6 and especially 12 hours it becomes more reliable [1].
Validation and performance
A systematic review and meta-analysis by Zhou et al. included 13 observational studies with a total of 1,751 patients with pneumonia and acute hypoxaemic respiratory failure [2]. The hierarchical summary AUC (AUHSROC) was 0.81 (95% CI 0.77 to 0.84), with a pooled sensitivity of 0.71 (95% CI 0.64 to 0.78) and specificity of 0.78 (95% CI 0.70 to 0.84). The cut-off values varied considerably between studies, however, with a mean of 4.8 and a median of 5.3, and most values centring between 4.5 and 6.0 [2]. Subgroup analyses showed comparable predictive ability whether the measurement was made within 6 hours or between 6 and 12 hours, and the index performed reliably in the COVID-19 subgroup as well.
External validation in a Japanese multicentre cohort of 300 COVID-19 patients (Okano et al.) showed poorer early discrimination, however [3]. The AUROC for ROX at 2 hours was only 0.57, and at 6 hours 0.62. Only at 12 hours did it reach 0.68, and at 48 hours 0.75. With the original lower limit of < 2.85 at 2 hours, sensitivity was only 3% at a specificity of 100%, meaning that almost all the patients who did go on to fail were missed early. Calibration assessment showed that ROX was not well calibrated in this population [3].
Vega et al. studied 120 COVID-19 patients treated outside intensive care at three centres [4]. Here the AUC at 12 hours was 0.79, but the optimal threshold was 5.99 (specificity 96%, sensitivity 62%), considerably higher than 4.88. The Roca threshold of 4.88 could not significantly discriminate between success and failure in this cohort (p = 0.4 on log-rank testing) [4]. The authors discuss how the different mechanism of hypoxaemia in COVID-19 pneumonia, with microthrombosis and a high dead space, may explain why a higher ROX is needed to predict success.
In summary, ROX performs best in non-COVID pneumonia, where it was derived, and when measured after 12 hours. In COVID-19 cohorts, discrimination is poorer early on and the thresholds may need to be adjusted upwards.
Limitations
ROX is validated for hypoxaemic respiratory failure due to pneumonia. Extrapolation to other aetiologies, for example heart failure, ARDS of other origin or postoperative respiratory failure, lacks strong empirical support and should be done with caution.
The index captures only three variables and ignores important clinical parameters such as level of consciousness, haemodynamics, pH and the degree of respiratory effort. A patient with a rising pCO₂, fatigue or haemodynamic instability may have an apparently acceptable ROX and still require immediate intubation. ROX should never be used as the sole basis for postponing intubation when there are clinical signs of exhaustion.
The decision to intubate is itself subjective and is influenced by the availability of intensive care beds, local practice and the expectations of the treating clinicians. This is a source of bias in all validation studies and partly explains the variation in reported thresholds and performance between cohorts.
SpO₂ is an unreliable variable at very high FiO₂, since pulse oximetry has limited resolution at saturations close to 100%, which can mask deterioration. In marked anaemia or peripheral circulatory compromise, the SpO₂ may also be misleading.
References
- Roca O, Caralt B, Messika J, et al. An Index Combining Respiratory Rate and Oxygenation to Predict Outcome of Nasal High-Flow Therapy. Am J Respir Crit Care Med. 2019;199(11):1368-1376. PMID: 30576221
- Zhou X, Liu J, Pan J, et al. The ROX index as a predictor of high-flow nasal cannula outcome in pneumonia patients with acute hypoxemic respiratory failure: a systematic review and meta-analysis. BMC Pulm Med. 2022;22(1):121. PMID: 35365110
- Okano H, Yamamoto R, Iwasaki Y, et al. External validation of the HACOR score and ROX index for predicting treatment failure in patients with coronavirus disease 2019 pneumonia managed on high-flow nasal cannula therapy: a multicenter retrospective observational study in Japan. J Intensive Care. 2024;12(1):7. PMID: 38360681
- Vega ML, Dongilli R, Olaizola G, et al. COVID-19 Pneumonia and ROX index: Time to set a new threshold for patients admitted outside the ICU. Pulmonology. 2022;28(1):13-17. PMID: 34049831
