European Journal of Cardiovascular Medicine

Hypoxemic respiratory failure is an important cause of Pediatric Intensive Care Unit (PICU) admissions and mechanical ventilation and is a common cause of morbidity and mortality^[1,2,3]. Both respiratory and non-respiratory illnesses can lead to respiratory failure in pediatric population which is the most common cause of cardiac arrest in children and is preventable cause of mortality in this group of population^[4]. Diagnosis and monitoring of Pediatric Respiratory Failure requires various clinical and laboratory tools among which Arterial Blood Gas (ABG) sampling is an essential parameter^[5,6]. However, the procedure for drawing an ABG sample apart from being painful, is also more of a one-time representation of the parameters of the patient^[7]. Although continuous ABG monitoring based on a combination of opto-chemical and fiberoptic detectors is an option, it requires placement of either a sensor in an artery or an indwelling arterial catheter from which repeated sampling can be done to avoid multiple punctures to the patient ^[8]. An important parameter used for diagnosing and monitoring critically ill patients with Hypoxic respiratory failure in ICU is continuous measurement of Oxygenation Index (OI), calculation of which requires multiple ABG sampling and analysis^[9-15]. Invasive monitoring methods, while usually being more accurate and far superior, they possess certain disadvantages such as inherent risk of transmission of infection and are more costly and time consuming and also require expert personnel and equipment. Hence the importance of using non-invasive parameters for monitoring of critically ill patients is based on the fact that they are safer, cheaper and retain a significant degree of sensitivity and specificity as the invasive methods. Oxygen saturation index (OSI) can be calculated by substituting SpO₂ for PaO₂ in the formula for OI, which can serve as a surrogate marker^[16-18]. In addition to being painless and non-invasive, it has the advantages of generating continuous data which can help in real time monitoring and decision making of patients with hypoxemic respiratory failure. Further it does not require the level of technical expertise needed in the whole process of sampling and analysing an ABG. In resource poor countries like India, especially in peripheral health care facilities, where the availability of an ABG analyser is a rare possibility, use of indicators that take into account a non-invasive parameter like SpO₂ in place of invasive parameters, can be of immense value for better monitoring and management of critically ill children.^[19]

Odisha from October 2017 to September 2019. After approval of the institutional ethics committee, patients aged 1 month to 14 years who were on mechanical ventilation and hemodynamically stable at the time of measurement of PaO₂ and SpO₂ were included in the study. The patients in need of immediate resuscitation, those with critical congenital heart disease, patients on inotrope or vasopressor support, patients with abnormal haemoglobin and those with perinatal related lung disease were excluded from this study. All data relating to demography, clinical and treatment details as well as outcomes were collected in a pre-designed Performa.

The recruitment protocol was used following the inclusion and exclusion criteria. Among 712 patients admitted to ICU, 216 critically ill patients required mechanical ventilation. Among them, 21 patients were excluded by exclusion criteria. As 47 patients did not give consent for study, 148 patients were included in the study. A maximum of 4 ABG samples were taken for each patient which included samples taken at 1-hour, 12 hours and 24 hours post-intubation and 12 hours prior to planned extubation. OI was calculated from the formula OI = MAP x FiO₂ / PaO₂ where MAP is Mean Airway Pressure in mm Hg, FiO₂ is fraction of Oxygen in inspired air administered during mechanical ventilation in percentage and PaO₂ is partial pressure of Oxygen determined by invasive ABG analysis in mm Hg. OSI was calculated from the formula OI = MAP x FiO₂ / SpO₂ where SpO₂ is percentage saturation of Haemoglobin (Hb) by oxygen measured by non-invasive pulse oximetry device.

Sample size calculations

This was an observational study in which we intended to estimate various proportions. Therefore, minimum sample size determination procedure for estimating a population proportion was adopted here. The formula used for this purpose is as follows:

n =z²_1-a/2P(1-P)/d²

 Where n = Minimum sample size

z²_1-a/2 = value of the standard normal variant for  1-a/2 level of significance= 1.96.

P = Anticipated population proportion 

100(1-a) % = Confidence level

d= absolute precision required on either side of the population.

In this study, following values of the above parameters have been considered keeping view the frequency of availability of the cases in the study hospital

1. Confidence level =1-a =95% 
2. Anticipated population proportion: Since multiple population proportions were estimated, anticipated population proportions ‘P’ was taken as 50% to produce the maximum of the minimum sample sizes.  
3. Absolute precision d = 9% point was taken.

For these values of the input, the minimum sample size required was computed at 118.

Statistical analysis

Data were analysed using statistical software SPSS (Statistics version 26.0) and results were calculated using cross tabulation, chi square test, linear regression, ROC curve analysis and independent t test wherever applicable. A p value of <0.05 was considered statistically significant

A total of 148 children were included in the study, and the ratio to boys:girls was 2:1. The mean(SD) duration of PICU stay was 9.53(6.52) days, and mean(SD) duration of mechanical ventilation was 6.1(5.6) days. with the majority (45.3%) being ventilated for 3-6 days. The most common indication of ICU admission that was identified was respiratory distress (31.1%) followed by congestive cardiac failure (22.3%), while respiratory failure was the most common indication of mechanical ventilation (58.78%). Pneumonia (19.59%), meningo-encephalitis (14.86%) and congenital heart disease (10.14%) were the 3 most common diagnoses. Synchronized Intermittent Mandatory Ventilation (SIMV) was the mode used in majority of patients with Pressure Controlled Ventilation + Pressure support ventilation (i.e. PCV + PSV) used in 52.1% and Volume Controlled Ventilation + Pressure support ventilation (VCV + PSV) in 45.8% of patients. 

The Mean (  OI in the study population was 4.92  2.29) while the mean OSI was 5.67  2.78). Linear regression analysis was done and following regression equation was derived correlating OI and OSI: OI = 1.488 + (0.605 x OSI), which can be further simplified to: OI = 1.5 + (0.6 x OSI)

Statistically significant correlation was found between OI and OSI (p ≤ 0.001) with a Pearson correlation coefficient of 0.73in the 583 recordings analyzed as shown in Figure 1. Corresponding to an OI value of 4, sensitivity and specificity for diagnosis of P-ARDS was calculated at an OSI value of 4.15 which was 89.7% and 91.3% respectively. Using an OSI cut off of 5.80 (corresponding to an OI of 5), sensitivity was 67.8% and specificity was 95.4% and has been shown in Figure 2. 

Significant positive correlation was also found between OSI and extubation failure (Pearson’s r = 0.32, p≤0.001) and between OI and extubation failure (Pearson’s r = 0.48, p≤0.001).

FiO2 = Fraction of oxygen in inspired air, HCO3 = Bicarbonate in Arterial blood gas, ICU = Intensive Care Unit, MAP = Mean Airway Pressure, OI = Oxygenation Index, OSI = Oxygen Saturation Index, PaO2 = Partial pressure of Oxygen in arterial blood, PEEP = Positive End Expiratory Pressure, PIP = Peak Inspiratory Pressure, SD = Standard Deviation, SpO2 = Percentage of oxygen saturation measured by pulse oximetry 

Various factors significantly affecting OI-OSI difference were identified among which indications of mechanical ventilation, aetiological diagnosis and blood gas parameters are as shown in Figure 3 and Table 1, respectively. Other such factors causing high OI-OSI difference were SIMV (PCV+PSV) mode used during mechanical ventilation (68.42%, p≤0.001), HME (heat & moisture exchange system) type humidification system (13.16%, p≤0.001).

Younger age (p≤0.001), lower weight (p≤0.001) and height (p≤0.001) was associated with higher mortality in the study population. Longer duration of PICU stay (p≤0.001) and longer duration of mechanical ventilation (p≤0.001) were also associated with higher mortality. Lower pH (p≤0.001), higher mean airway pressure (p≤0.001), higher FiO₂ requirement (p≤0.001), lower PaO₂ values (p≤0.001), lower SpO₂ recordings (p≤0.001), lower bicarbonate values (p≤0.001), higher PEEP requirement (p≤0.001) and higher PIP requirement (p≤0.001) were associated with higher mortality in the study population.

The mean (± SD) OI and OSI values recorded in this study conducted among critically ill patients undergoing mechanical ventilation were 4.9  2.3 and 5.7  2.8 respectively with a mean difference of 0.75  1.90. The correlation coefficient calculated between OI and OSI values was r = 0.734 (p  0.001) and the equation of correlation obtained by linear regression was OI = 1.5 + (0.6 x OSI).

The findings of this study were similar to other studieswho have found a high correlation of 0.91 between OI and OSI in their study and suggested the potential for such non-invasive parameters to help reduce need for invasive monitoring, requirement of workload and cost.^[16,20]Studies conducted in neonates have also found a good accuracy of OSI in assessment of severity of hypoxic respiratory failure in place of OI by substituting SpO₂ in place of PaO₂.^[11]Significant associations have been derived between OSI on the day of ARDS diagnosis and mortality as well as number of fewer ventilator free days and suggested OSI as a reliable surrogate for OI in patients with ARDS.^[17]Similar study conducted in neonates with respiratory failure have concluded that OSI can be used as non-invasive surrogate marker for OI in patients with respiratory failure with good reliability.^[21]

A sensitivity of 89.7% at an OSI cut off of 4.15 (= OI of 4) found in our study indicates excellent reliability of using this parameter as a diagnostic parameter for P-ARDS. Taking a higher cut off of 5.80 increases the specificity significantly i.e. 95.4% while compromising the sensitivity (67.8%).

Although good correlation exists between OI and OSI, many factors that significantly affect the difference between the two parameters were identified. High OI-OSI difference observed in cases of Respiratory failure and Congestive cardiac failure may be caused by the compromised perfusion status in such patients. Higher MAP, FiO₂, PEEP and PIP settings also showed high OI-OSI difference which indicates the more severe nature of illness requiring extremely high mechanical ventilation settings, which implies OSI cannot be used as a reliable parameter at very high ventilatory requirements. Similar findings were associated with longer duration of ICU stay and longer duration of mechanical ventilation.  

The poorer outcome associated with lower age, weight and height may be attributed to malnutrition and its associated co-morbidities. Lower pH, PaO₂, SpO₂ and bicarbonate found in association with high mortality reflect the state of circulatory insufficiency associated with severe respiratory and cardiac illnesses, which are also reflected in the longer days of ICU admission, days of mechanical ventilation and higher MAP, FiO₂, PEEP and PIP values. 

The degree of correlation between predicting success of extubation and OI obtained in this study was r = 0.479 & that between predicting success of extubation and OSI was r = 0.324. This refers to the greater reliability of OI than OSI in being used as a predictor for success of extubation in patients with respiratory failure who are being planned for extubation. Although OI is a better indicator in such situations, the need for invasive ABG sampling for its calculation makes it a less favourable one. Rather the use of OSI has the advantages of almost similar predictability while being completely non-invasive and painless for calculation.

The limitations of the present study are –

1. A higher sample size could have provided an even better correlation between OI and OSI
2. A similar study including neonates as the study population could be performed which could provide valuable guidance in neonatal critical care
3. Use of CO-oximetry could have been taken into account which could have added value to the study by including some more cases with abnormal haemoglobin (e.g. Methemoglobinemia), but was not available.

Although good correlation exists between OI and OSI, many factors significantly affect the difference between the two. Therefore, OSI can be used as a reliable monitoring method in controlled settings after ensuring good patient selection, proper method of sampling and sample handling, good quality electronic devices and invasive monitoring facilities.

FUNDING:  None

CONFLICTS OF INTEREST:  None

1. Eriksen V, Nielsen LH, Klokker M,GreisenG.Follow-upof5 to11 year-old children treated for persistent pulmonary hypertension of the newborn ActaPaediatr.2009;98(2):304-309.
2. WalshSukysMC,BauerRE,CornellDJ,FriedmanHG,StorkEK,Hack M. Severe respiratory failure in neonates: mortality and morbidity rates and neurodevelopmental outcomes JPediatr. 1994;125(1):104-110.
3. Wheeler DS, Wong HR, Shanley TP. Science and Practice of Pediatric Critical Care Medicine. London: Springer; 2009. p. 1
4. Shebl E, Burns B. Respiratory Failure. [Updated 2019 May 6]. In: StatPearls. Treasure Island (FL): StatPearls Publishing; 2019 Jan-. Available from: https://www.ncbi.nlm.nih.gov/books/NBK526127
5. Doradla LPS, Vadivelan M, et al. Invasive monitoring in the intensive care unit. Ind J Clin Pract 2013; 24: 430-5
6. Sood P, Paul G, Puri S. Interpretation of arterial blood gas. Indian J Crit Care Med. 2010;14(2):57-64.
7. Al-Mukhaini KS, Al-Rahbi NM Noninvasive Ventilation and High-Flow Nasal Cannulae Therapy for Children with Acute Respiratory Failure: An overview. Sultan Qaboos Univ Med J. 2018;18(3):e278-e285.
8. Weiss IK, Fink S, Harrison R, Feldman JD, Brill JE Clinical Use of Continuous Arterial Blood Gas Monitoring in the Pediatric Intensive Care Unit. Pediatrics.1999;103(2):440-445.