- Email: [email protected]
fraction of inspired oxygen ratio in a heterogeneous sample of critically ill children
Journal of Critical Care (2013) 28, 538.e1–538.e7
Correlation of oxygen saturation as measured by pulse oximetry/fraction of inspired oxygen ratio with PaO2/fraction of inspired oxygen ratio in a heterogeneous sample of critically ill children☆ Carlos Lobete MD ⁎, Alberto Medina MD, Corsino Rey MD, Juan Mayordomo-Colunga MD, Andrés Concha MD, Sergio Menéndez MD Pediatric Intensive Care Unit, Hospital Universitario Central de Asturias, University of Oviedo, Oviedo, Spain
Keywords: Non invasive positivepressure ventilation; Respiratory insufficiency; Clinical markers; Hypoxemia; Critical care; Mechanical ventilation; Pediatrics
Abstract Purpose: Oxygen saturation as measured by pulse oximetry (SpO2)/fraction of inspired oxygen (FIO2) (SF) ratio has demonstrated to be an adequate marker for lung disease severity in children under mechanical ventilation. We sought to validate the utility of SF ratio in a population of critically ill children under mechanical ventilation, noninvasive ventilation support, and breathing spontaneously. Materials and methods: A retrospective database study was conducted in a pediatric intensive care unit of a university hospital. Children with SpO2 less than or equal to 97% and an indwelling arterial catheter were included. Simultaneous blood gas and pulse oximetry were collected in a database. Derivation and validation data sets were generated, and a linear mixed modeling was used to derive predictive equations. Model performance and fit were evaluated using the validation data set. Results: Three thousand two hundred forty-eight blood gas and SpO2 values from 298 patients were included. 1/SF ratio had a strong linear association with 1/PaO2/FIO2 (PF) ratio in both derivation and validation data sets, given by the equation 1/SF = 0.00164 + 0.521/PF (derivation). Oxygen saturation as measured by pulse oximetry/FIO2 values for PF criteria of 100, 200, and 300 were 146 (95% confidence interval [CI], 142-150), 236 (95% CI, 228-244), and 296 (95% CI, 285-308). Areas under receiver operating characteristic curves for diagnosis of PF ratio less than 100, 200, and 300 with the SF ratio were 0.978, 0.952, and 0.951, respectively, in the validation data set. Conclusions: Oxygen saturation as measured by pulse oximetry/FIO2 ratio is an adequate noninvasive surrogate marker for PF ratio. Oxygen saturation as measured by pulse oximetry/FIO2 ratio may be an ideal noninvasive marker for patients with acute hypoxemic respiratory failure. © 2013 Elsevier Inc. All rights reserved.
☆
Financial support: No financial support. ⁎ Corresponding author. Service of Pediatry, Hospital Universitario Central de Asturias, C/ Celestino Villamil s/n, E-33006 Oviedo, Asturias, Spain. Tel.: + 34 985 108066; fax: + 34 985 107887. E-mail address: [email protected] (C. Lobete). 0883-9441/$ – see front matter © 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jcrc.2012.12.006
538.e2
1. Introduction Acute lung injury and acute respiratory distress syndrome (ARDS) have a low incidence but with a large mortality in children admitted to pediatric intensive care units (PICUs) [1,2]. The diagnosis of acute lung injury/ ARDS according to American-European Consensus Conference definition required arterial blood gas (ABG) sampling [3]. Similarly, the recent Berlin definition of ARDS also includes arterial PaO2, setting 3 categories of ARDS based on the degree of hypoxemia: mild (200 mm Hg b PaO2/fraction of inspired oxygen [FIO2] ≤ 300 mm Hg), moderate (100 mm Hg b PaO2/FIO2 ≤ 200 mm Hg), and severe (PaO2/FIO2 [PF] ≤ 100 mm Hg) [4]. However, in current clinical practice, ABG performed in PICUs is much less common than in the past. An early diagnosis in these pathologies is essential because an earlier establishment of the lung protective ventilation strategies may improve the outcome of children [5-7]. Furthermore, there has been a great increase of noninvasive ventilation (NIV) use to effectively treat respiratory failure in children in recent years. Acute respiratory distress syndrome, acute hypoxemic respiratory failure, and high oxygen requirements have also been identified as NIV failure predictors in both pediatric and adult patients [8-16]. In fact, PF ratio has been suggested as a useful figure to decide whether to intubate a patient during NIV therapy [12]. Similarly, oxygen saturation as measured by pulse oximetry (SpO2)/FIO2 (SF) ratio has also been described as a NIV failure predictor in adults [17], whereas in children, SF ratio has also been suggested as a NIV outcome predictor in a preliminary study [18]. Because of being noninvasive and continuously available, SF ratio might be very useful during NIV. Arterial PaO2 is also included in many pediatric severity scores, such as Pediatric Index of Mortality 2 [19] and Pediatric Risk of Mortality III [20]. As previously stated, because of the scarcity of arterial simples nowadays, this item is frequently lost. A noninvasive surrogate could improve the usefulness of these scores if it was always available. In a similar way, the oxygenation index has been recently described to have a strong linear association to oxygen saturation index [21]. Oxygen saturation as measured by pulse oximetry/FIO2 ratio has been demonstrated to correlate well with the PF ratio in both adult and pediatric studies [21-24], and it may be used to diagnose ARDS in a noninvasive manner [21]. In children, SF ratios of 221 and 264 correspond to PF ratios of 200 and 300. In this pediatric study, the population was very homogeneous because 95% of the observations met oxygenation criteria for mild ARDS (PF ratio b 300) and close to 80% met oxygenation criteria for moderate ARDS (PF ratio b 200) [21]. We hypothesized that the correlation between SF ratio and PF ratio may be different in a sample of children under mechanical ventilation, NIV support, and breathing spontaneously without supplementary oxygen. The main objective of our study was to validate the SF ratio/PF ratio correlation in a heterogeneous sample of critically ill children.
C. Lobete et al. Secondary objective was to compare our results with previously published works.
2. Materials and methods We conducted a retrospective study of ABG and SpO2 values in a population of children, under mechanical ventilation, noninvasive ventilation support, and breathing spontaneously, who were admitted to a tertiary care noncardiac surgery PICU. Data were extracted from a clinical and research database, maintained and monitored by the PICU physicians delivering patient care. This database integrates components of the intensive care unit flow sheet and laboratories as well as diagnostic and demographic information. Corresponding measurements of SpO2, PaO2, and FIO2 were included in the database at the time of the extraction of ABG. According to our PICU protocol, these data were recorded simultaneously. Before SpO2 was recorded, it was checked the correct positioning of the sensor, adequate waveforms, and that no ventilators changes had been made in the previous 30 minutes. Data were excluded from analysis if the patient had a diagnosis of methemoglobinemia or carbon monoxide poisoning. Oxygen saturation as measured by pulse oximetry values over 97% were also excluded. Patients on nasal cannula or mask oxygen were excluded because of the inaccuracy to calculate the delivered FIO2. We used Philips FAST (Fourier Artifact Suppression Technology) pulse oximeters (Philips Healthcare, Eindhoven, Netherlands). The study was approved by the institutional review board with a waiver of written consent.
2.1. Analysis Data set was divided into derivation and validation groups through randomization with 50% of observations in each group. The random split was done on an observation level.
2.2. Derivation data set We transformed data (1/PF and 1/SF) to satisfy assumptions of normality and improve model fit. The data set was first analyzed using 2-way scatter plots to characterize the relationship between 1/PF and 1/SF ratios. Simple correlation with Pearson correlation coefficient was obtained, and linear regression analysis was used to derive a predictive equation for 1/PF ratio prediction from 1/SF ratio. Given that multiple measurements were possible from the same patient, a general mixed model was used. Based on the derived regression equation, SF ratio values that corresponded to PF ratio values of 100, 200, and 300 were determined. Receiver operating characteristic (ROC) curves were plotted with the area under the curve (AUC) and calculated to assess the degree of SF ratio discrimination for diagnostic real PF ratio less than 100, 200, and 300. We calculated the
Correlation of SF with PF ratio in a heterogeneous sample Table 1 Baseline demographics of the patients in the derivation and validation data sets
Age, y Sex, % male PaO2, torr SpO2, % FIO2 PF ratio SF ratio Room air, % NIV Mechanical ventilation
Derivation data set (n = 1643)
Validation data set (n = 1605)
1.9 (0.6-5.4) 59.1 74.7 (63.6-88.8) 95 (92-96) 0.45 (0.35-0.6) 169.5 (114.3-247.7) 213.3 (158.3-274.3) 18 23.1 58.9
1.8 (0.6-5.4) 58.7 75.2 (64-89.3) 95 (92-96) 0.40 (0.35-0.6) 171.3 (115.6-254.9) 225 (158.3-274.3) 18.1 21.9 60
Data are presented as median (interquartile range).
sensitivity, specificity, positive and negative predictive values, and positive and negative likelihood ratios (LRs) of the SF cutoffs against the PF cutoffs of 100, 200, and 300.
2.3. Validation data set The validation data set was analyzed in a similar manner to the derivation data set. The predicted SF cutoffs from the
Fig. 1
538.e3 derivation data set were applied to the validation data set. Receiver operating characteristic plots were generated and AUC were calculated to determine how well SF ratio discriminates real PF ratio less than 100, 200, and 300 in the validation data set. We calculated the sensitivity, specificity, positive and negative predictive values, and positive and negative LR of the SF cutoffs obtained in the derivation data set against the PF cutoffs of 100, 200, and 300. Finally, with the prevalence or pretest probability of PF ratio less than 100, 200, and 300, we calculated the posttest probability of PF ratio less than 100, 200, and 300.
3. Results The derivation data set consisted of 1643 measurements from 235 patients. A total of 325 data pairs (19.8%) had a PF ratio less than 100, 1013 (61.7%) had a PF ratio less than 200, and 1389 (84.5%) had a PF ratio less than 300. The validation data set was composed of 1605 data pairs from 242 patients. A total of 310 measurements (19.3%) had a PF ratio less than 100, 985 (61.4%) had a PF ratio less than 200, and 1335 (83.2%) had a PF ratio less than 300. The baseline demographics of the 2 data sets are shown in Table 1.
Scatter plot of 1/SF ratio vs 1/PF ratio. The central line corresponds to the linear regression equation, and lateral lines, to the 95% CI.
538.e4
3.1. Derivation data set The relationship between SF ratio and PF ratio was best expressed using 1/SF ratio and 1/PF ratio. Oxygen saturation as measured by pulse oximetry/FIO2 ratio had a strong linear association with PF ratio, described by the linear regression equation 1/SF = 0.00164 + 0.521/PF [95% confidence interval (CI) 1/SF = (0.00155-0.00173) + (0.510-0.532)/PF; P b .0001, R2 = 0.843] (Fig. 1). Based on this equation, a PF ratio of 300 corresponds to an SF ratio of 296 (95% CI, 285-308); a PF ratio of 200, to an SF ratio of 236 (95% CI, 228-244); and a PF ratio of 100, to an SF ratio of 146 (95% CI, 142-150). In the derivation data set, the SF ratio had excellent discrimination ability for diagnosis PF ratio of 300 or less (AUC, 0.956), PF ratio of 200 or less (AUC, 0.944), and PF ratio of 100 or less (AUC, 0.978). In the derivation data set, the SF value of 296 had 91% sensitivity and 87% specificity in detecting cases in which the PF ratio was 300 or less, the SF value of 236 had 88% sensitivity and 86% specificity in detecting cases in which the PF ratio was 200 or less, and the SF value of 146 had 52% sensitivity and 99% specificity in detecting cases in which the PF ratio was 100 or less. The positive LR was greater than 6, and the negative LR was less than 0.2, except in diagnosis PF of 100 less (Table 2).
3.2. Validation data set In the validation data set, SF ratio had a similar association with PF ratio, given by the regression equation 1/SF = 0.00149 + 0.544/PF [95% CI; 1/SF = (0.001410.00158) + (0.533-0.555)/PF; P b .0001, R2 = 0.854]. Corresponding values for PF ratio of 300 (SF, 302; 95% CI, 292-314), PF ratio of 200 (SF, 238; 95% CI, 230-245), and PF ratio of 100 (SF, 144; 95% CI, 140-148) were similar. The PF predicted by the regression equation from the derivation data set had excellent discrimination ability for diagnosis PF ratio of 300 or less (AUC, 0.951), PF ratio of 200 or less (AUC, 0.952), and PF ratio of 100 or less (AUC, 0.978) in the validation data set (Table 2). When the predicted SF values from the derivation data set were applied to the validation data set, the SF value of 296 had 90% sensitivity and 85% specificity in diagnosis PF ratio of 300 or less, the SF value of 236 had 86% sensitivity and 90% specificity in diagnosis PF ratio of 200 or less, and the SF value of 146 had 53.5% sensitivity and 99% specificity in diagnosis PF ratio of 100 or less. The positive LR was greater than 6, and the negative LR was less than 0.2 (Table 2). In this population, the prevalence or pretest probability of PF ratio of 300 or less was 83.2%, and consequently, the pretest odds was 5. With the positive LR of 6, the posttest odds for a PF ratio of 300 or less if their SF ratio of 296 or less was 30 and the posttest probability 96.8%. The negative LR of 0.12 yielded a posttest odds of 0.6 and posttest
C. Lobete et al. Table 2 Discrimination ability of the SF ratio for diagnosis of PF ratio of 100, 200, and 300 from the derivation and validation data sets PF ratio ≤ 300 AUC Sensitivity Specificity Positive predictive value Negative predictive value Positive LR Negative LR % of cases classified correctly PF ratio ≤ 200 AUC Sensitivity Specificity Positive predictive value Negative predictive value Positive LR Negative LR % of cases classified correctly PF ratio ≤ 100 AUC Sensitivity Specificity Positive predictive value Negative predictive value Positive LR Negative LR % of cases classified correctly
Derivation
Validation
0.956 91 87 97 64 6.8 0.1 90
0.951 90 85 97 63 6 0.12 89
0.944 88 86 91 82 6.2 0.14 87
0.952 86 90 93 80 8.7 0.15 88
0.972 52 99 96 77 52 0.48 81
0.978 54 99 98 78 54 0.46 82
Data are presented as area under the ROC curve plots. These values are generated from separate predictive models in each data set. Sensitivity, specificity, positive and negative predictive values, positive and negative LRs, and percentage of correctly classified cases using the cutoffs values of SF ratio 146, 236, and 296 from the derivation data set against the cutoffs values of PF (100, 200, and 300) applied to both derivation and validation data sets.
probability of 37.5%. Therefore, an SF ratio greater than 296 decreased the probability for a PF ratio of 300 or less from 83.2% to 37.5%. The prevalence or pretest probability of PF ratio of 200 or less was approximately 61.4%, and consequently, the pretest odds was 1.6. The positive LR of 8.7 yielded a posttest odds of 13.5 and a posttest probability of 93.1%. The negative LR of 0.12 yielded a posttest odds of 0.19 and posttest probability of 16%. In this population, an SF ratio greater than 236 decreased the probability for a PF ratio of 200 or less from 61% to 16%. The prevalence or pretest probability of PF ratio of 100 or less was 19.3%, and consequently, the pretest odds was 0.2. The positive LR of 54 yielded a posttest odds of 12.9 and a posttest probability of 92.8%. The negative LR of 0.46 yielded a posttest odds of 0.11 and posttest probability of 9.9%. In this population, an SF ratio greater than 146 decreased the probability for a PF ratio of 100 or less from 19.3% to 9.9%.
Rice et al Chest 2007
Khemani et al Chest 2009
Thomas et al Pediatr Crit Care Med 2010
Khemani et al Pediatr Crit Care Med 2012
Our study
Characteristics
Adults with ARDS Prospective
Children in 2 PICUs Retrospective
2673 96.9 79.7 2031 96.1 72.6 SF = 64 + (0.84 × PF) PF = (SF − 64)/0.84 R = 0.89 315 91% 56% 0.878 235 85% 85% 0.928
1298 94 80 1845 91 72 SF = 76 + (0.62 × PF) PF = (SF − 76)/0.84 R2 = 0.61 263 86% 47% 0.792 201 68% 84% 0.85
Children under mechanical ventilation Multicenter Prospective 721 95 a 80 a 469 95 a 80 a 1/SF = 0.00232 + 0.443/PF PF = 0.443/(1/SF − 0.00232)
Children in 1 PICU Retrospective
Derivation data % of PF ratio b 300 % of PF ratio b 200 Validation data % of PF ratio b 300 % of PF ratio b 200 Regression equation
Children and adolescents with LPA Multicenter Retrospective 1159 Not reported Not reported 648 Not reported Not reported Not reported Concordance coefficient 0.77 253 93% 43% 0.87 212 76% 83% 0.88
264 92 64 0.93 221 89 80 0.92
PF 300
PF 200
SF Sensibility Specificity AUC SF Sensibility Specificity AUC
1643 84.5 61.7 1605 83.2 61.4 1/SF = 0.00164 + 0.521/PF PF = 0.521/(1/SF − 0.00164) R2 = 0.843 296 90 85 0.951 236 86 90 0.952
Correlation of SF with PF ratio in a heterogeneous sample
Table 3 Principal characteristics, regression equation, SF values for PF values of 200 and 300 with the area under the ROC curve and sensitivity and specificity values of the studies that have analyzed the relation between SF and PF ratios
The PF value of 100 was not included because it was not calculated in the previous studies. a Percentages expressed for the whole population.
538.e5
538.e6
4. Discussion Our study has demonstrated that in a heterogeneous sample of critically ill children, SF ratio had a strong correlation with PF ratio. Oxygen saturation as measured by pulse oximetry/FIO2 values of 296, 236, and 146 corresponded to PF ratio criteria for mild, moderate, and severe ARDS. Our results are consistent with those previously reported by Khemani et al [21,22], Rice et al [23], and Thomas et al [24], indicating that SF ratio has an excellent correlation with PF ratio. Table 3 compares the regression equation, sample characteristics, and SF values for PF values of 200 and 300 with the area under the ROC curve as well as sensitivity and specificity values of these studies. Our cutoff values were higher than previously published in pediatric studies [21,22,24] but quite similar to those published in adult population [23]. The SF ratio values of 236 and 296 had similar sensitivity and better specificity for the diagnosis of a PF ratio below 200 and 300 compared with previous studies [21-24]. Therefore, we have found a better positive LR, which means better diagnostic efficacy for illness diagnosis. Taking into account 200 and 300 PF ratio prevalence in our sample, we saw that an SF ratio over 296 decreases more than a half the probability of a PF ratio under 300. An SF ratio over 236 means a 4-fold decrease of the probability of a PF ratio under 200. Therefore, a patient would probably meet PF criteria for mild and moderate ARDS if he met SF criteria, as suggested by Khemani et al [21]. Regarding PF ratio of 100, an SF ratio over 146 decreases about a half the probability of fulfilling severe ARDS gasometric criterion [4]. It should be highlighted that we analyzed a heterogeneous sample that might allow generalization of our results to most children patients, regardless of being or not under mechanical ventilation. The accuracy of SF ratio as a surrogate for PF ratio is lost in cases of carbon monoxide poisoning or methemoglobinemia and if SpO2 is over 97% (as the hemoglobin dissociation curve flattens over SpO2 97%, thus losing the linear correlation between SpO2 and PaO2). Therefore, SF ratio should not be used in these cases. The oxygen saturation index has been proposed as a good noninvasive lung injury severity marker [21,24]. However, oxygen saturation index calculation needs mean airway pressure value, making it not possible for patients without mechanical ventilation, for example, NIV patients. In addition, the SF ratio has demonstrated to be useful as noninvasive marker of success or failure of the NIV both at the time of its initiation and in the follow-up [16-18]. Our study has several limitations. First, although simultaneous blood gas and pulse oximetry were prospectively collected, analysis of data was retrospectively performed. Second, we could not account for the effect of mean arterial pressure or positive end-expiratory pressure over PF/SF relationship, due to the study design. This point deserves further studies. Third, we did not have a clinical target for SF values, for example, NIV failure. Fourth,
C. Lobete et al. children included were those with an arterial line, which are mainly hypoxemic, postoperative, and hemodynamic unstable. Therefore, our findings may not be generalizable to all patients. Fifth, we used 1 model of pulse oximeter (FAST), which could make our results not reproducible to other units using different pulse oximeters. Khemani et al [21] used Massimo and Nellcor pulse oximeters in their study. On the other hand, the strength of our study is that we have validated the SF ratio for clinical practice in a more heterogeneous population of critically ill children than previously described.
5. Conclusions Oxygen saturation as measured by pulse oximetry/FIO2 ratio is a useful noninvasive marker for lung disease severity. In children under noninvasive ventilation with hypoxemic acute respiratory failure, SF ratio may be a useful clinical marker of patients at risk for severe ARDS and need of mechanical ventilation.
References [1] Bindl L, Dresbach K, Lentze MJ. Incidence of acute respiratory distress syndrome in German children and adolescents: a populationbased study. Crit Care Med 2005;33:209-312. [2] Erickson S, Schibler A, Numa A, et al. Acute lung injury in pediatric intensive care in Australia and New Zealand: a prospective, multicenter, observational study. Pediatr Crit Care Med 2007;8:317-23. [3] Bernard GR, Artigas A, Brigham KL, et al. The American-European Consensus Conference on ARDS. Definitions, mechanisms, relevant outcomes, and clinical trial coordination. Am J Respir Crit Care Med 1994;149:818-24. [4] Ranieri VM, Rubenfeld GD, Thompson BT, et al. Acute respiratory distress syndrome: the Berlin Definition. JAMA 2012;307:2526-33. [5] The Acute Respiratory Distress Syndrome Network. Ventilation with lower tidal volumes as compared with traditional tidal volumes for acute lung injury and the acute respiratory distress syndrome. N Engl J Med 2000;342:1301-8. [6] Amato MB, Barbas CS, Medeiros DM, et al. Effect of a protectiveventilation strategy on mortality in the acute respiratory distress syndrome. N Engl J Med 1998;338:347-54. [7] Villar J, Kacmarek RM, Perez-Mendez L, et al. A high positive endexpiratory pressure, low tidal volume ventilatory strategy improves outcome in persistent acute respiratory distress syndrome: a randomized, controlled trial. Crit Care Med 2006;34:1311-8. [8] Bernet V, Hug MI, Frey B. Predictive factors for the success of noninvasive mask ventilation in infants and children with acute respiratory failure. Pediatr Crit Care Med 2005;6:660-4. [9] Essouri S, Chevret L, Durand P, et al. Noninvasive positive pressure ventilation: five years of experience in a pediatric intensive care unit. Pediatr Crit Care Med 2006;7:329-34. [10] James CS, Hallewell CP, James DP, et al. Predicting the success of noninvasive ventilation in preventing intubation and re-intubation in the paediatric intensive care unit. Intensive Care Med 2011;37:1994-2001. [11] Mayordomo-Colunga J, Medina A, Rey C, et al. Predictive factors of non invasive ventilation failure in critically ill children: a prospective epidemiological study. Intensive Care Med 2009;35:527-36. [12] Antonelli M, Conti G, Esquinas A, et al. A multiple-center survey on the use in clinical practice of noninvasive ventilation as a first-line
Correlation of SF with PF ratio in a heterogeneous sample
[13]
[14]
[15]
[16]
[17]
intervention for acute respiratory distress syndrome. Crit Care Med 2007;35:18-25. Antonelli M, Conti G, Moro ML, et al. Predictors of failure of noninvasive positive pressure ventilation in patients with acute hypoxemic respiratory failure: a multi-center study. Intensive Care Med 2001;27:1718-28. Rana S, Jenad H, Gay PC, et al. Failure of non-invasive ventilation in patients with acute lung injury: observational cohort study. Crit Care 2006;10:R79. Schettino G, Altobelli N, Kacmarek RM. Noninvasive positivepressure ventilation in acute respiratory failure outside clinical trials: experience at the Massachusetts General Hospital. Crit Care Med 2008;36:441-7. Pons M, Medina A, Martinón-Torres F, Mayordomo J. Failure analysis and predictive factors in non-invasive ventilation. In: Medina A, Pons M, Martinón-Torres F, editors. Non-invasive ventilation in pediatrics. Second Edition. Madrid: Ergón Editorial; 2009. p. 91-8. Spada C, Gandhi R, Patel SR, et al. Oxygen saturation/fraction of inspired oxygen ratio is a simple predictor of noninvasive positive pressure ventilation failure in critically ill patients. J Crit Care 2011;26: 510-6.
538.e7 [18] Mayordomo-Colunga J, Pons M, Solana M, et al. Useful variables for predicting non-invasive ventilation failure in pediatric patients. A Spanish-Portuguese multicenter study. Abstract published in Pediatr Crit Care Med 2011;12(3):A41. [19] Slater A, Shann F, Pearson G. PIM2: a revised version of the Paediatric Index of Mortality. Intensive Care Med 2003;29:278-85. [20] Pollack MM, Patel KM, Ruttimann UE. PRISM III: an updated Pediatric Risk of Mortality score. Crit Care Med 1996;24:743-52. [21] Khemani RG, Thomas NJ, Venkatachalam V, et al. Comparison of SpO2 to PaO2 based markers of lung disease severity for children with acute lung injury. Crit Care Med 2012;40:1309-16. [22] Khemani RG, Patel NR, Bart RD, et al. Comparison of the pulse oximetric saturation/fraction of inspired oxygen ratio and the PaO2/fraction of inspired oxygen ratio in children. Chest 2009;135: 662-8. [23] Rice TW, Wheeler AP, Bernard GR, et al. Comparison of the SpO2/FIO2 ratio and the PaO2/FIO2 ratio in patients with acute lung injury or ARDS. Chest 2007;132:410-7. [24] Thomas NJ, Shaffer ML, Willson DF, et al. Defining acute lung disease in children with the oxygenation saturation index. Pediatr Crit Care Med 2010;11:12-7.
Correlation of oxygen saturation as measured by pulse oximetry/fraction of inspired oxygen ratio with PaO2/fraction of inspired oxygen ratio in a heterogeneous sample of critically ill children☆ Carlos Lobete MD ⁎, Alberto Medina MD, Corsino Rey MD, Juan Mayordomo-Colunga MD, Andrés Concha MD, Sergio Menéndez MD Pediatric Intensive Care Unit, Hospital Universitario Central de Asturias, University of Oviedo, Oviedo, Spain
Keywords: Non invasive positivepressure ventilation; Respiratory insufficiency; Clinical markers; Hypoxemia; Critical care; Mechanical ventilation; Pediatrics
Abstract Purpose: Oxygen saturation as measured by pulse oximetry (SpO2)/fraction of inspired oxygen (FIO2) (SF) ratio has demonstrated to be an adequate marker for lung disease severity in children under mechanical ventilation. We sought to validate the utility of SF ratio in a population of critically ill children under mechanical ventilation, noninvasive ventilation support, and breathing spontaneously. Materials and methods: A retrospective database study was conducted in a pediatric intensive care unit of a university hospital. Children with SpO2 less than or equal to 97% and an indwelling arterial catheter were included. Simultaneous blood gas and pulse oximetry were collected in a database. Derivation and validation data sets were generated, and a linear mixed modeling was used to derive predictive equations. Model performance and fit were evaluated using the validation data set. Results: Three thousand two hundred forty-eight blood gas and SpO2 values from 298 patients were included. 1/SF ratio had a strong linear association with 1/PaO2/FIO2 (PF) ratio in both derivation and validation data sets, given by the equation 1/SF = 0.00164 + 0.521/PF (derivation). Oxygen saturation as measured by pulse oximetry/FIO2 values for PF criteria of 100, 200, and 300 were 146 (95% confidence interval [CI], 142-150), 236 (95% CI, 228-244), and 296 (95% CI, 285-308). Areas under receiver operating characteristic curves for diagnosis of PF ratio less than 100, 200, and 300 with the SF ratio were 0.978, 0.952, and 0.951, respectively, in the validation data set. Conclusions: Oxygen saturation as measured by pulse oximetry/FIO2 ratio is an adequate noninvasive surrogate marker for PF ratio. Oxygen saturation as measured by pulse oximetry/FIO2 ratio may be an ideal noninvasive marker for patients with acute hypoxemic respiratory failure. © 2013 Elsevier Inc. All rights reserved.
☆
Financial support: No financial support. ⁎ Corresponding author. Service of Pediatry, Hospital Universitario Central de Asturias, C/ Celestino Villamil s/n, E-33006 Oviedo, Asturias, Spain. Tel.: + 34 985 108066; fax: + 34 985 107887. E-mail address: [email protected] (C. Lobete). 0883-9441/$ – see front matter © 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jcrc.2012.12.006
538.e2
1. Introduction Acute lung injury and acute respiratory distress syndrome (ARDS) have a low incidence but with a large mortality in children admitted to pediatric intensive care units (PICUs) [1,2]. The diagnosis of acute lung injury/ ARDS according to American-European Consensus Conference definition required arterial blood gas (ABG) sampling [3]. Similarly, the recent Berlin definition of ARDS also includes arterial PaO2, setting 3 categories of ARDS based on the degree of hypoxemia: mild (200 mm Hg b PaO2/fraction of inspired oxygen [FIO2] ≤ 300 mm Hg), moderate (100 mm Hg b PaO2/FIO2 ≤ 200 mm Hg), and severe (PaO2/FIO2 [PF] ≤ 100 mm Hg) [4]. However, in current clinical practice, ABG performed in PICUs is much less common than in the past. An early diagnosis in these pathologies is essential because an earlier establishment of the lung protective ventilation strategies may improve the outcome of children [5-7]. Furthermore, there has been a great increase of noninvasive ventilation (NIV) use to effectively treat respiratory failure in children in recent years. Acute respiratory distress syndrome, acute hypoxemic respiratory failure, and high oxygen requirements have also been identified as NIV failure predictors in both pediatric and adult patients [8-16]. In fact, PF ratio has been suggested as a useful figure to decide whether to intubate a patient during NIV therapy [12]. Similarly, oxygen saturation as measured by pulse oximetry (SpO2)/FIO2 (SF) ratio has also been described as a NIV failure predictor in adults [17], whereas in children, SF ratio has also been suggested as a NIV outcome predictor in a preliminary study [18]. Because of being noninvasive and continuously available, SF ratio might be very useful during NIV. Arterial PaO2 is also included in many pediatric severity scores, such as Pediatric Index of Mortality 2 [19] and Pediatric Risk of Mortality III [20]. As previously stated, because of the scarcity of arterial simples nowadays, this item is frequently lost. A noninvasive surrogate could improve the usefulness of these scores if it was always available. In a similar way, the oxygenation index has been recently described to have a strong linear association to oxygen saturation index [21]. Oxygen saturation as measured by pulse oximetry/FIO2 ratio has been demonstrated to correlate well with the PF ratio in both adult and pediatric studies [21-24], and it may be used to diagnose ARDS in a noninvasive manner [21]. In children, SF ratios of 221 and 264 correspond to PF ratios of 200 and 300. In this pediatric study, the population was very homogeneous because 95% of the observations met oxygenation criteria for mild ARDS (PF ratio b 300) and close to 80% met oxygenation criteria for moderate ARDS (PF ratio b 200) [21]. We hypothesized that the correlation between SF ratio and PF ratio may be different in a sample of children under mechanical ventilation, NIV support, and breathing spontaneously without supplementary oxygen. The main objective of our study was to validate the SF ratio/PF ratio correlation in a heterogeneous sample of critically ill children.
C. Lobete et al. Secondary objective was to compare our results with previously published works.
2. Materials and methods We conducted a retrospective study of ABG and SpO2 values in a population of children, under mechanical ventilation, noninvasive ventilation support, and breathing spontaneously, who were admitted to a tertiary care noncardiac surgery PICU. Data were extracted from a clinical and research database, maintained and monitored by the PICU physicians delivering patient care. This database integrates components of the intensive care unit flow sheet and laboratories as well as diagnostic and demographic information. Corresponding measurements of SpO2, PaO2, and FIO2 were included in the database at the time of the extraction of ABG. According to our PICU protocol, these data were recorded simultaneously. Before SpO2 was recorded, it was checked the correct positioning of the sensor, adequate waveforms, and that no ventilators changes had been made in the previous 30 minutes. Data were excluded from analysis if the patient had a diagnosis of methemoglobinemia or carbon monoxide poisoning. Oxygen saturation as measured by pulse oximetry values over 97% were also excluded. Patients on nasal cannula or mask oxygen were excluded because of the inaccuracy to calculate the delivered FIO2. We used Philips FAST (Fourier Artifact Suppression Technology) pulse oximeters (Philips Healthcare, Eindhoven, Netherlands). The study was approved by the institutional review board with a waiver of written consent.
2.1. Analysis Data set was divided into derivation and validation groups through randomization with 50% of observations in each group. The random split was done on an observation level.
2.2. Derivation data set We transformed data (1/PF and 1/SF) to satisfy assumptions of normality and improve model fit. The data set was first analyzed using 2-way scatter plots to characterize the relationship between 1/PF and 1/SF ratios. Simple correlation with Pearson correlation coefficient was obtained, and linear regression analysis was used to derive a predictive equation for 1/PF ratio prediction from 1/SF ratio. Given that multiple measurements were possible from the same patient, a general mixed model was used. Based on the derived regression equation, SF ratio values that corresponded to PF ratio values of 100, 200, and 300 were determined. Receiver operating characteristic (ROC) curves were plotted with the area under the curve (AUC) and calculated to assess the degree of SF ratio discrimination for diagnostic real PF ratio less than 100, 200, and 300. We calculated the
Correlation of SF with PF ratio in a heterogeneous sample Table 1 Baseline demographics of the patients in the derivation and validation data sets
Age, y Sex, % male PaO2, torr SpO2, % FIO2 PF ratio SF ratio Room air, % NIV Mechanical ventilation
Derivation data set (n = 1643)
Validation data set (n = 1605)
1.9 (0.6-5.4) 59.1 74.7 (63.6-88.8) 95 (92-96) 0.45 (0.35-0.6) 169.5 (114.3-247.7) 213.3 (158.3-274.3) 18 23.1 58.9
1.8 (0.6-5.4) 58.7 75.2 (64-89.3) 95 (92-96) 0.40 (0.35-0.6) 171.3 (115.6-254.9) 225 (158.3-274.3) 18.1 21.9 60
Data are presented as median (interquartile range).
sensitivity, specificity, positive and negative predictive values, and positive and negative likelihood ratios (LRs) of the SF cutoffs against the PF cutoffs of 100, 200, and 300.
2.3. Validation data set The validation data set was analyzed in a similar manner to the derivation data set. The predicted SF cutoffs from the
Fig. 1
538.e3 derivation data set were applied to the validation data set. Receiver operating characteristic plots were generated and AUC were calculated to determine how well SF ratio discriminates real PF ratio less than 100, 200, and 300 in the validation data set. We calculated the sensitivity, specificity, positive and negative predictive values, and positive and negative LR of the SF cutoffs obtained in the derivation data set against the PF cutoffs of 100, 200, and 300. Finally, with the prevalence or pretest probability of PF ratio less than 100, 200, and 300, we calculated the posttest probability of PF ratio less than 100, 200, and 300.
3. Results The derivation data set consisted of 1643 measurements from 235 patients. A total of 325 data pairs (19.8%) had a PF ratio less than 100, 1013 (61.7%) had a PF ratio less than 200, and 1389 (84.5%) had a PF ratio less than 300. The validation data set was composed of 1605 data pairs from 242 patients. A total of 310 measurements (19.3%) had a PF ratio less than 100, 985 (61.4%) had a PF ratio less than 200, and 1335 (83.2%) had a PF ratio less than 300. The baseline demographics of the 2 data sets are shown in Table 1.
Scatter plot of 1/SF ratio vs 1/PF ratio. The central line corresponds to the linear regression equation, and lateral lines, to the 95% CI.
538.e4
3.1. Derivation data set The relationship between SF ratio and PF ratio was best expressed using 1/SF ratio and 1/PF ratio. Oxygen saturation as measured by pulse oximetry/FIO2 ratio had a strong linear association with PF ratio, described by the linear regression equation 1/SF = 0.00164 + 0.521/PF [95% confidence interval (CI) 1/SF = (0.00155-0.00173) + (0.510-0.532)/PF; P b .0001, R2 = 0.843] (Fig. 1). Based on this equation, a PF ratio of 300 corresponds to an SF ratio of 296 (95% CI, 285-308); a PF ratio of 200, to an SF ratio of 236 (95% CI, 228-244); and a PF ratio of 100, to an SF ratio of 146 (95% CI, 142-150). In the derivation data set, the SF ratio had excellent discrimination ability for diagnosis PF ratio of 300 or less (AUC, 0.956), PF ratio of 200 or less (AUC, 0.944), and PF ratio of 100 or less (AUC, 0.978). In the derivation data set, the SF value of 296 had 91% sensitivity and 87% specificity in detecting cases in which the PF ratio was 300 or less, the SF value of 236 had 88% sensitivity and 86% specificity in detecting cases in which the PF ratio was 200 or less, and the SF value of 146 had 52% sensitivity and 99% specificity in detecting cases in which the PF ratio was 100 or less. The positive LR was greater than 6, and the negative LR was less than 0.2, except in diagnosis PF of 100 less (Table 2).
3.2. Validation data set In the validation data set, SF ratio had a similar association with PF ratio, given by the regression equation 1/SF = 0.00149 + 0.544/PF [95% CI; 1/SF = (0.001410.00158) + (0.533-0.555)/PF; P b .0001, R2 = 0.854]. Corresponding values for PF ratio of 300 (SF, 302; 95% CI, 292-314), PF ratio of 200 (SF, 238; 95% CI, 230-245), and PF ratio of 100 (SF, 144; 95% CI, 140-148) were similar. The PF predicted by the regression equation from the derivation data set had excellent discrimination ability for diagnosis PF ratio of 300 or less (AUC, 0.951), PF ratio of 200 or less (AUC, 0.952), and PF ratio of 100 or less (AUC, 0.978) in the validation data set (Table 2). When the predicted SF values from the derivation data set were applied to the validation data set, the SF value of 296 had 90% sensitivity and 85% specificity in diagnosis PF ratio of 300 or less, the SF value of 236 had 86% sensitivity and 90% specificity in diagnosis PF ratio of 200 or less, and the SF value of 146 had 53.5% sensitivity and 99% specificity in diagnosis PF ratio of 100 or less. The positive LR was greater than 6, and the negative LR was less than 0.2 (Table 2). In this population, the prevalence or pretest probability of PF ratio of 300 or less was 83.2%, and consequently, the pretest odds was 5. With the positive LR of 6, the posttest odds for a PF ratio of 300 or less if their SF ratio of 296 or less was 30 and the posttest probability 96.8%. The negative LR of 0.12 yielded a posttest odds of 0.6 and posttest
C. Lobete et al. Table 2 Discrimination ability of the SF ratio for diagnosis of PF ratio of 100, 200, and 300 from the derivation and validation data sets PF ratio ≤ 300 AUC Sensitivity Specificity Positive predictive value Negative predictive value Positive LR Negative LR % of cases classified correctly PF ratio ≤ 200 AUC Sensitivity Specificity Positive predictive value Negative predictive value Positive LR Negative LR % of cases classified correctly PF ratio ≤ 100 AUC Sensitivity Specificity Positive predictive value Negative predictive value Positive LR Negative LR % of cases classified correctly
Derivation
Validation
0.956 91 87 97 64 6.8 0.1 90
0.951 90 85 97 63 6 0.12 89
0.944 88 86 91 82 6.2 0.14 87
0.952 86 90 93 80 8.7 0.15 88
0.972 52 99 96 77 52 0.48 81
0.978 54 99 98 78 54 0.46 82
Data are presented as area under the ROC curve plots. These values are generated from separate predictive models in each data set. Sensitivity, specificity, positive and negative predictive values, positive and negative LRs, and percentage of correctly classified cases using the cutoffs values of SF ratio 146, 236, and 296 from the derivation data set against the cutoffs values of PF (100, 200, and 300) applied to both derivation and validation data sets.
probability of 37.5%. Therefore, an SF ratio greater than 296 decreased the probability for a PF ratio of 300 or less from 83.2% to 37.5%. The prevalence or pretest probability of PF ratio of 200 or less was approximately 61.4%, and consequently, the pretest odds was 1.6. The positive LR of 8.7 yielded a posttest odds of 13.5 and a posttest probability of 93.1%. The negative LR of 0.12 yielded a posttest odds of 0.19 and posttest probability of 16%. In this population, an SF ratio greater than 236 decreased the probability for a PF ratio of 200 or less from 61% to 16%. The prevalence or pretest probability of PF ratio of 100 or less was 19.3%, and consequently, the pretest odds was 0.2. The positive LR of 54 yielded a posttest odds of 12.9 and a posttest probability of 92.8%. The negative LR of 0.46 yielded a posttest odds of 0.11 and posttest probability of 9.9%. In this population, an SF ratio greater than 146 decreased the probability for a PF ratio of 100 or less from 19.3% to 9.9%.
Rice et al Chest 2007
Khemani et al Chest 2009
Thomas et al Pediatr Crit Care Med 2010
Khemani et al Pediatr Crit Care Med 2012
Our study
Characteristics
Adults with ARDS Prospective
Children in 2 PICUs Retrospective
2673 96.9 79.7 2031 96.1 72.6 SF = 64 + (0.84 × PF) PF = (SF − 64)/0.84 R = 0.89 315 91% 56% 0.878 235 85% 85% 0.928
1298 94 80 1845 91 72 SF = 76 + (0.62 × PF) PF = (SF − 76)/0.84 R2 = 0.61 263 86% 47% 0.792 201 68% 84% 0.85
Children under mechanical ventilation Multicenter Prospective 721 95 a 80 a 469 95 a 80 a 1/SF = 0.00232 + 0.443/PF PF = 0.443/(1/SF − 0.00232)
Children in 1 PICU Retrospective
Derivation data % of PF ratio b 300 % of PF ratio b 200 Validation data % of PF ratio b 300 % of PF ratio b 200 Regression equation
Children and adolescents with LPA Multicenter Retrospective 1159 Not reported Not reported 648 Not reported Not reported Not reported Concordance coefficient 0.77 253 93% 43% 0.87 212 76% 83% 0.88
264 92 64 0.93 221 89 80 0.92
PF 300
PF 200
SF Sensibility Specificity AUC SF Sensibility Specificity AUC
1643 84.5 61.7 1605 83.2 61.4 1/SF = 0.00164 + 0.521/PF PF = 0.521/(1/SF − 0.00164) R2 = 0.843 296 90 85 0.951 236 86 90 0.952
Correlation of SF with PF ratio in a heterogeneous sample
Table 3 Principal characteristics, regression equation, SF values for PF values of 200 and 300 with the area under the ROC curve and sensitivity and specificity values of the studies that have analyzed the relation between SF and PF ratios
The PF value of 100 was not included because it was not calculated in the previous studies. a Percentages expressed for the whole population.
538.e5
538.e6
4. Discussion Our study has demonstrated that in a heterogeneous sample of critically ill children, SF ratio had a strong correlation with PF ratio. Oxygen saturation as measured by pulse oximetry/FIO2 values of 296, 236, and 146 corresponded to PF ratio criteria for mild, moderate, and severe ARDS. Our results are consistent with those previously reported by Khemani et al [21,22], Rice et al [23], and Thomas et al [24], indicating that SF ratio has an excellent correlation with PF ratio. Table 3 compares the regression equation, sample characteristics, and SF values for PF values of 200 and 300 with the area under the ROC curve as well as sensitivity and specificity values of these studies. Our cutoff values were higher than previously published in pediatric studies [21,22,24] but quite similar to those published in adult population [23]. The SF ratio values of 236 and 296 had similar sensitivity and better specificity for the diagnosis of a PF ratio below 200 and 300 compared with previous studies [21-24]. Therefore, we have found a better positive LR, which means better diagnostic efficacy for illness diagnosis. Taking into account 200 and 300 PF ratio prevalence in our sample, we saw that an SF ratio over 296 decreases more than a half the probability of a PF ratio under 300. An SF ratio over 236 means a 4-fold decrease of the probability of a PF ratio under 200. Therefore, a patient would probably meet PF criteria for mild and moderate ARDS if he met SF criteria, as suggested by Khemani et al [21]. Regarding PF ratio of 100, an SF ratio over 146 decreases about a half the probability of fulfilling severe ARDS gasometric criterion [4]. It should be highlighted that we analyzed a heterogeneous sample that might allow generalization of our results to most children patients, regardless of being or not under mechanical ventilation. The accuracy of SF ratio as a surrogate for PF ratio is lost in cases of carbon monoxide poisoning or methemoglobinemia and if SpO2 is over 97% (as the hemoglobin dissociation curve flattens over SpO2 97%, thus losing the linear correlation between SpO2 and PaO2). Therefore, SF ratio should not be used in these cases. The oxygen saturation index has been proposed as a good noninvasive lung injury severity marker [21,24]. However, oxygen saturation index calculation needs mean airway pressure value, making it not possible for patients without mechanical ventilation, for example, NIV patients. In addition, the SF ratio has demonstrated to be useful as noninvasive marker of success or failure of the NIV both at the time of its initiation and in the follow-up [16-18]. Our study has several limitations. First, although simultaneous blood gas and pulse oximetry were prospectively collected, analysis of data was retrospectively performed. Second, we could not account for the effect of mean arterial pressure or positive end-expiratory pressure over PF/SF relationship, due to the study design. This point deserves further studies. Third, we did not have a clinical target for SF values, for example, NIV failure. Fourth,
C. Lobete et al. children included were those with an arterial line, which are mainly hypoxemic, postoperative, and hemodynamic unstable. Therefore, our findings may not be generalizable to all patients. Fifth, we used 1 model of pulse oximeter (FAST), which could make our results not reproducible to other units using different pulse oximeters. Khemani et al [21] used Massimo and Nellcor pulse oximeters in their study. On the other hand, the strength of our study is that we have validated the SF ratio for clinical practice in a more heterogeneous population of critically ill children than previously described.
5. Conclusions Oxygen saturation as measured by pulse oximetry/FIO2 ratio is a useful noninvasive marker for lung disease severity. In children under noninvasive ventilation with hypoxemic acute respiratory failure, SF ratio may be a useful clinical marker of patients at risk for severe ARDS and need of mechanical ventilation.
References [1] Bindl L, Dresbach K, Lentze MJ. Incidence of acute respiratory distress syndrome in German children and adolescents: a populationbased study. Crit Care Med 2005;33:209-312. [2] Erickson S, Schibler A, Numa A, et al. Acute lung injury in pediatric intensive care in Australia and New Zealand: a prospective, multicenter, observational study. Pediatr Crit Care Med 2007;8:317-23. [3] Bernard GR, Artigas A, Brigham KL, et al. The American-European Consensus Conference on ARDS. Definitions, mechanisms, relevant outcomes, and clinical trial coordination. Am J Respir Crit Care Med 1994;149:818-24. [4] Ranieri VM, Rubenfeld GD, Thompson BT, et al. Acute respiratory distress syndrome: the Berlin Definition. JAMA 2012;307:2526-33. [5] The Acute Respiratory Distress Syndrome Network. Ventilation with lower tidal volumes as compared with traditional tidal volumes for acute lung injury and the acute respiratory distress syndrome. N Engl J Med 2000;342:1301-8. [6] Amato MB, Barbas CS, Medeiros DM, et al. Effect of a protectiveventilation strategy on mortality in the acute respiratory distress syndrome. N Engl J Med 1998;338:347-54. [7] Villar J, Kacmarek RM, Perez-Mendez L, et al. A high positive endexpiratory pressure, low tidal volume ventilatory strategy improves outcome in persistent acute respiratory distress syndrome: a randomized, controlled trial. Crit Care Med 2006;34:1311-8. [8] Bernet V, Hug MI, Frey B. Predictive factors for the success of noninvasive mask ventilation in infants and children with acute respiratory failure. Pediatr Crit Care Med 2005;6:660-4. [9] Essouri S, Chevret L, Durand P, et al. Noninvasive positive pressure ventilation: five years of experience in a pediatric intensive care unit. Pediatr Crit Care Med 2006;7:329-34. [10] James CS, Hallewell CP, James DP, et al. Predicting the success of noninvasive ventilation in preventing intubation and re-intubation in the paediatric intensive care unit. Intensive Care Med 2011;37:1994-2001. [11] Mayordomo-Colunga J, Medina A, Rey C, et al. Predictive factors of non invasive ventilation failure in critically ill children: a prospective epidemiological study. Intensive Care Med 2009;35:527-36. [12] Antonelli M, Conti G, Esquinas A, et al. A multiple-center survey on the use in clinical practice of noninvasive ventilation as a first-line
Correlation of SF with PF ratio in a heterogeneous sample
[13]
[14]
[15]
[16]
[17]
intervention for acute respiratory distress syndrome. Crit Care Med 2007;35:18-25. Antonelli M, Conti G, Moro ML, et al. Predictors of failure of noninvasive positive pressure ventilation in patients with acute hypoxemic respiratory failure: a multi-center study. Intensive Care Med 2001;27:1718-28. Rana S, Jenad H, Gay PC, et al. Failure of non-invasive ventilation in patients with acute lung injury: observational cohort study. Crit Care 2006;10:R79. Schettino G, Altobelli N, Kacmarek RM. Noninvasive positivepressure ventilation in acute respiratory failure outside clinical trials: experience at the Massachusetts General Hospital. Crit Care Med 2008;36:441-7. Pons M, Medina A, Martinón-Torres F, Mayordomo J. Failure analysis and predictive factors in non-invasive ventilation. In: Medina A, Pons M, Martinón-Torres F, editors. Non-invasive ventilation in pediatrics. Second Edition. Madrid: Ergón Editorial; 2009. p. 91-8. Spada C, Gandhi R, Patel SR, et al. Oxygen saturation/fraction of inspired oxygen ratio is a simple predictor of noninvasive positive pressure ventilation failure in critically ill patients. J Crit Care 2011;26: 510-6.
538.e7 [18] Mayordomo-Colunga J, Pons M, Solana M, et al. Useful variables for predicting non-invasive ventilation failure in pediatric patients. A Spanish-Portuguese multicenter study. Abstract published in Pediatr Crit Care Med 2011;12(3):A41. [19] Slater A, Shann F, Pearson G. PIM2: a revised version of the Paediatric Index of Mortality. Intensive Care Med 2003;29:278-85. [20] Pollack MM, Patel KM, Ruttimann UE. PRISM III: an updated Pediatric Risk of Mortality score. Crit Care Med 1996;24:743-52. [21] Khemani RG, Thomas NJ, Venkatachalam V, et al. Comparison of SpO2 to PaO2 based markers of lung disease severity for children with acute lung injury. Crit Care Med 2012;40:1309-16. [22] Khemani RG, Patel NR, Bart RD, et al. Comparison of the pulse oximetric saturation/fraction of inspired oxygen ratio and the PaO2/fraction of inspired oxygen ratio in children. Chest 2009;135: 662-8. [23] Rice TW, Wheeler AP, Bernard GR, et al. Comparison of the SpO2/FIO2 ratio and the PaO2/FIO2 ratio in patients with acute lung injury or ARDS. Chest 2007;132:410-7. [24] Thomas NJ, Shaffer ML, Willson DF, et al. Defining acute lung disease in children with the oxygenation saturation index. Pediatr Crit Care Med 2010;11:12-7.