Validity of ROX index in prediction of risk of intubation in patients with COVID-19 pneumonia

Introduction: One important concern during the management of COVID-19 pneumonia patients with acute hypoxemic respiratory failure is early anticipation of the need for intubation. ROX is an index that can help in identification of patients with low and those with high risk of intubation. So, this study was planned to validate the diagnostic accuracy of the ROX index for prediction of COVID-19 pneumonia outcome (the need for intubation) and, in addition, to underline the significant association of the ROX index with clinical, radiological, demographic data. Material and methods: Sixty-nine RT-PCR positive COVID-19 patients were enrolled. The following data were collected: medical history, clinical classification of COVID-19 infection, the ROX index measured daily and the outcome assessment. Results: All patients with severe COVID-19 infection (100%) were intubated (50% of them on the 3 rd day of admission), but only 38% of patients with moderate COVID-19 infection required intubation (all of them on the 3 rd day of admission). The ROX index on the 1 st day of admission was significantly associated with the presence of comorbidities, COVID-19 clinical classification, CT findings and intubation (p ≤ 0.001 for each of them). Regression analysis showed that sex and ROX.1 are the only significant independent predictors of intubation [AOR (95% CI): 16.9 (2.4– 117), 0.77 (0.69–0.86)], respectively. Cut-off point of the ROX index on the 1 st day of admission was ≤ 25.26 (90.2% of sensitivity and 75% of specificity). Conclusions: ROX is a simple noninvasive promising tool for predicting discontinuation of high-flow oxygen therapy and could be used in the assessment of progress and the risk of intubation in COVID-19 patients with pneumonia.


Introduction
Severe hypoxemia resulting from COVID-19 pneumonia is often associated with near normal respiratory system compliance, which is almost never seen in severe ARDS. However, COVID-19 pneumonia in most cases falls under the Berlin definition of ARDS [1,2].
Severely hypoxemic patients with COVID-19 pneumonia, despite sharing a single etiology, may have different presentations: markedly dyspneic or normally breathing ("happy" hypoxemia); and either responsive to prone position or not. So, the same disease presents itself with impressive heterogenity. Two primary "phenotypes": type L, characterized by low elastance (i.e., high compliance), low recruitment and low lung weight and type H, characterized by high elastance, high recruitment and high lung weight were reported [3]. For patients with COVID-19 infection, oxygen supplementation via low-flow nasal cannula may be sufficient, however; in patients with acute hypoxemic respiratory failure, higher flow of oxygen may be needed, and noninvasive modalities (HFNC and NIV) may be used rather than proceeding directly to intubation [4]. A systematic review from July 2020 identified one trial evaluating HFNC in patients with COVID-19, which suggested that it reduced the need for mechanical ventilation and improvements in oxygenation compared with standard oxygen therapy [5]. The ROX inwww.journals.viamedica.pl dex defined as the ratio of oxygen saturation as measured by pulse oximetry/FiO 2 to respiratory rate can help identify those patients with low and those with high risk of intubation [6,7]. One important concern during the management of COVID-19 pneumonia patients with acute hypoxemic respiratory failure is not to delay intubation. So, this study was planned to validate the diagnostic accuracy of the ROX index for COVID-19 pneumonia outcome (the need for intubation) and, in addition, to underline the significant association of the ROX index with clinical, radiological, and demographic data.

Material and methods
This diagnostic study was conducted on 69 RT-PCR positive COVID-19 patients with radiological evidence of pneumonia attending different quarantine places in Egypt from April 2020 to June 2020. Mild COVID-19 patients (presenting with respiratory symptoms without radiological evidence of pneumonia), individuals with chronic respiratory failure and those with conditions that affect pulse oximeter reading (e.g. nail polish) were excluded. This study was conducted within the required ethics guidelines of Mansoura institutional research board ethics committee (code number: R.20.05.832) and approved by the Ministry of Health and Population. Training and research sector (REC) (code: Com. NO/Dec.No:14-2020/2).
The following data were collected: -medical history, e. -the ROX index refers to combination of the ratio of oxygen saturation (as measured by pulse oximetry) to fraction of inspired oxygen and respiratory rate (SpO 2 /FiO 2 /RR). It was calculated daily for all patients who underwent oxygen therapy from the measured value of SpO 2 and respiratory rates (breaths/ /min) and supplemental oxygen (FiO 2 values) [6,7]; -other data indicate severity: PO 2 /FiO 2 , SpO 2 / /FiO 2 , other organ failure, d-dimer, serum ferritin, the presence of shock; -all patients were primarily evaluated for the need for oxygen therapy and follow-up for one week for either clinical improvement (fever drops, better respiratory symptoms, hemodynamics stability, less need for oxygen (< 0.4) or deterioration with the necessity for invasive mechanical ventilation. Mechanical ventilation was considered in the presence of worsening or persistent respiratory distress, respiratory rate more than 40 breaths/min, SpO 2 less than 90% despite maximum oxygen flow and FiO 2 , acidemia with pH less than 7.25, significant hemodynamic instability and multiorgan failur; -failure of oxygen therapy was considered when the patient needed invasive mechanical ventilation within one week from the beginning of treatment with oxygen.

Statistical analysis of data
The collected data was prepared, tabulated, and statistically analyzed using statistical package for social science (SPSS) version 16. Frequencies and percentages were used to present nominal variables, while means (SD), or median (min-max) were used to present continuous data according to the results of Shapiro-Wilk testing of normality of variables. Some variables were calculated: the ROX index using the formula (SpO 2 /FiO 2 )/respiratory rate as ROX.1:ROX value on the first day of admission, ROX.2:ROX value on the 2 nd day of admission, ROX.3:ROX value on the 3 rd day of admission. Modified ROX was calculated using the formula: (PO 2 /FiO 2 )/respiratory rate. Significance testing was done with the help of chi-square test, and the independent-samples Mann-Whitney U test for categorical, non-parametric data, respectively. Also, Spearman correlation was used to test an association between non-parametric data. Multivariate logistic regression was done to determine the independent significant predictors of intubation. Variables found to have a significant association with the intubation outcome (intubatwww.journals.viamedica.pl ed/not) in univariate analysis were entered in the regression model. Receiver operator curve (ROC) analysis was used to determine the most accurate cut-off point for prediction of intubation. The level of significance was set at p < 0.05.

Discussion
Delayed intubation has been shown to be associated with poor clinical outcome, so predicting the failure of noninvasive ventilation or oxygen therapy has remained an important area of research [7]. One important concern during the management of COVID-19 pneumonia is not to delay intubation in patients with acute hypoxemic respiratory failure. An objective method to identify subjects who are likely to fail to respond to oxygen therapy is needed. The ROX index is remarkably simple and has the potential to become a routine parameter in clinical practice.
Roca and colleagues [6,7] first published an index ROX which can predict whether the patient will fail to use high frequency nasal canula (HFNC) in pneumonia in ICU, and they reported that a ROX value of > 4.88 predicted the success  The variable entered in the analysis included: sex, CT type, COVID clinical classification, comorbidity, lymphocyte count; AOR -adjusted odds ratio; CI -confidence interval; r -reference group; ROX.1 -ROX value in the first day post admission  of HFNC. Also, in Nicholas and Robin's study [8], the ROX index was validated in 191 critically ill patients enrolled at 5 centers in France and Spain, and the ROX index score of 4.88 was used as predictive of outcomes. The area under the curve at 12 hours of HFNC use was 0.752. Both studies calculated the ROX index at 0, 2, 6, 12 hours from the onset of oxygen therapy. Rodriguez et al. [9] found that the ROX index in critically ill patients in ICU was higher in the subjects who were successfully separated from HFNC at the first trial than in those who failed (12.7 vs 10.2, p = 0.002). The ROX index ≥ 9.2 predicted successful separation from HFNC at the first trial (specificity of 50%, sensitivity of 84%, positive predictive value of 93%, negative predictive value of 30%, and accuracy of 80%). They calculated the ROX index up to 48 hours.
All previous studies were conducted on critically illpatients other than COVID-19. To the best of our knowledge, there is only one recent study done by Belz and coworkers [10] using the ROX index for monitoring of oxygen therapy by HFNC in a SARS-CoV-2 severe pneumonia admitted to ICU with proven COVID-19, and the authors found that performance characteristics of ROX www.journals.viamedica.pl In this study, we used the ROX index as a simple noninvasive tool in COVID-19 pneumonia patients for prediction of the need for intubation. In contrast to previous studies, the ROX index was measured daily as we collected data from different medical facilities that dealt with COVID-19 cases, and we found that all patients with severe COVID-19 infection (100%) were intubated (Table 2), however, only 38% of moderate COVID-19 infection patients who needed oxygen therapy were intubated. All persons with moderate COVID-19 infection were intubated on the 5 th day of admission but 50% of patients with severe COVID-19 infection were intubated on the 3 rd day of admission.
The results of multivariate logistic regression of predictors of intubation among COVID-19 patients have shown that female sex and ROX.1 (the ROX value on the first day of admission) are the only significant independent predictors of intubation in this study. Cut-off point of ROX.1 (the ROX value on the 1 st day of admission) was ≤ 25.26 (90.2% of sensitivity and 75% of specificity) (AUC, p): (0.897, ≤ 0.001), ROX.2 (the ROX value on the 2 nd day of admission) was ≤ 21.34 (90% of sensitivity and 75% of specificity) and of ROX.3 (the ROX value on the 3 rd day of admission) ≤ 11.71 (90% of sensitivity and 100% of specificity). The cut-off value is higher than in other studies because we conducted this study on a heterogeneous cases of different severities as we involved some less severe cases of moderately severe COVID-19 pneumonia. ROX.1 was significantly associated with the presence of comorbidities, COVID-19 clinical classification, CT findings, intubation (p ≤ 0.001 for each of them). Also, there was a negative significant association with albumin.
Clinicians could use the ROX index to assess progress in COVID-19 patients, making serial measurements and incorporating it when considering decisions to increase care. During the 1 st day of admission, scores below the cut-offs given in this study would prompt anticipation of the need for earlier intubation. Once the 1 st day point is reached, a score > 25.26 increases clinician confidence that the patient will succeed.

Conclusions
ROX is a simple noninvasive promising tool for predicting discontinuation of high-flow oxygen therapy and could be used by clinicians in the assessment of progress and the risk of intubation in COVID-19 patients with pneumonia. Further studies on a large number of COVID-19 patients would be necessary to support our results.