Implementation of the Oxygen Saturation Index as a Predictor of Outcome in Prenatally Diagnosed CDH Neonates in the First 24 Hours of Life

ABSTRACT

Aims

This study aimed to evaluate the Oxygen Saturation Index (OSI) as a noninvasive measure for early postnatal management and outcome prediction in neonates with congenital diaphragmatic hernia (CDH). Additionally, the study analyzed the correlation and predictive ability of OSI, Oxygenation Index (OI), Horovitz Index (HI), and partial pressure of arterial oxygen (PaO₂) regarding mortality and the need for extracorporeal membrane oxygenation (ECMO).

Methods

A retrospective, single‐center study using data from 2013 to 2020. Parameters for calculating indices were extracted from patient charts every hour during the first 24 h of life. Statistical analyses included ROC analysis for predictive cut‐off values and Spearman's rank for correlation assessments.

Results

The study included 138 neonates. Postductal OSI demonstrated high sensitivity (80%–85%) and negative predictive value (NPV) for predicting mortality and ECMO need, with cut‐off values between 11.5 and 13. Optimal cut‐off values for predicting ECMO need were 10 at 12 and 24 h (sensitivity 96.7%). OSI, OI, HI, and PaO₂ showed comparable predictive capabilities with strong correlations. The lowest OI of 18 predicted mortality with a sensitivity of 75% and specificity of 90.9%. Strong correlations were found between the lowest PaO₂ and lowest HI (0.963–0.974), and between highest OI and lowest PaO₂ (−0.922 to −0.945).

Conclusion

OSI is a promising index for predicting outcomes in CDH neonates, showing strong correlation with indices like OI and HI. Despite limitations, OSI provides continuous, bedside monitoring without invasive blood sampling. Further prospective studies are needed to validate these findings and establish new cut‐off values.

Abbreviations

AUC

area under the curve

CDH

congenital diaphragmatic hernia

CLD

chronic lung disease

ECMO

extracorporeal membrane oxygenation

FiO₂

fraction of inspired oxygen

HI

Horovitz Index

MAP

mean airway pressure

NPV

negative predictive value

OI

Oxygenation Index

OSI

Oxygen Saturation Index

PaO₂

partial pressure of arterial oxygenation

PCO₂

partial pressure of CO₂

PH

pulmonary hypertension

PPV

positive predictive value

ROC

receiver operating characteristics

SaO₂

oxygen saturation of the blood

SpO₂

oxygen saturation detected by pulsoxymeter

1. Introduction

Congenital diaphragmatic hernia (CDH) is a potentially life‐threatening condition affecting newborns with an incidence of one in 4000 births [1]. Depending on the size of the defect, various organs migrate from the abdominal to the thoracic cavity leading to several pathophysiological abnormalities, including the trifecta of lung hypoplasia, pulmonary hypertension (PH), and cardiac dysfunction [2, 3]. Although it remains unclear which phenotype has the greatest impact on the need for extracorporeal membrane oxygenation (ECMO) and mortality, all three of them can lead to a significant clinical deterioration with the inability to adequately oxygenate the blood.

Traditionally, most published guidelines emphasize the Oxygenation Index (OI) as the main criterion for initiating ECMO support in neonates with CDH [4]. The OI is calculated by multiplying the mean airway pressure (MAP) with the amount of oxygen administered to the patient, divided by the partial pressure of arterial oxygen (PaO₂ [mmHg]), which has two important practical limitations. With the implementation of permissive hypercapnia and gentle lung ventilation strategies, it is usually aimed to keep the MAP within a reasonably narrow threshold (≤ 17 cmH₂O) [5, 6, 7, 8]. Because the MAP is an important parameter in this equation, the change in ventilation practice may have a direct impact on the calculation of the OI itself, questioning whether it is still the best parameter to assess the need for ECMO in this population. Second, the OI calculation requires the PaO₂, leading to repetitive arterial blood sampling, which might be associated with subsequent anemia and in the inability to calculate the OI in circumstances where arterial access is difficult to obtain [9].

Consequently, identifying alternative indices such as the Oxygen Saturation Index (OSI) and the Horovitz Index (HI) has become a new field in CDH research [10]. The OSI is calculated similarly to the OI but uses oxygen saturation from pulse oximetry (SpO₂) instead of PaO₂, making it a noninvasive option (OSI = MAP × FiO₂ × 100/SpO₂). The HI, also known as the PaO₂/FiO₂ ratio, commonly applied in the assessment and management of pediatric and adult acute respiratory distress syndrome (ARDS), is calculated by dividing PaO₂ by FiO₂, focusing solely on the oxygenation capacity of the lungs without accounting for ventilator settings. These indices may overcome limitations of the OI and might therefore be promising alternatives for predicting need for ECMO support and mortality. None of the values mandatory to calculate the OSI require blood sampling, allowing the OSI to be used for timely risk stratification as a real‐time bedside approach during early perinatal transition [10, 11].

The aim of this study was to investigate the OSI as a potential noninvasive index for early postnatal management and outcome prediction in CDH neonates. Furthermore, the correlation between OI, OSI, HI, and PaO₂, as well as their ability to predict outcome was analyzed.

2. Methods

2.1. Study Design and Patients

This was a retrospective, single‐center study of data collected from the database of the Department of Neonatology and Pediatric Intensive Care at the Children's Hospital Bonn, Germany, between 2013 and 2020. Due to its retrospective study design ethical approval is granted by the regulations of the ethics committee of the University of Bonn Medical Center (No. 113/22). For this retrospective study informed consent was waived by the ethics committee. Inclusion criteria were a prenatal diagnosis of CDH and the delivery in our hospital. Patients with concomitant severe malformations were excluded from this study. Parameters relevant for the calculation of the different indices were PaO₂, MAP, the fraction of inspired oxygen (FiO₂), and the pre‐ and postductal peripheral oxygen saturation (SpO₂) measured by pulsoxymeter. A difference of the preductal compared to the postductal SpO₂ ≥ 8% was considered to be clinically relevant and to indicate significant right‐to‐left shunting via the ductus arteriosus. For study purposes, all parameters were extracted from patients charts once every hour during the first 24 h of life. MAP values were taken at the same time as the above readings. Primary and secondary endpoints were mortality, and ECMO support, respectively.

2.2. Treatment Protocol and Prenatal Assessment

All neonates were managed according to the CDH Euro Consortium statement on standardized postnatal management of infants with CDH, allowing for permissive hypercapnia and a gentle ventilatory course [7]. Criteria for ECMO initiation were as follows: preductal oxygen saturation < 85% or postductal saturation > 70%, OI 40 consistently present, elevated PaCO₂ > 70 mmHg with pH < 7.15, peak inspiratory pressure ≥ 28 cmH₂O or MAP ≥ 17 cmH₂O, or persistent systemic hypotension (mean arterial pressure < 40 mmHg) refractory to fluid and inotropic therapy. The standard mode of ECMO in our institution is venovenous ECMO. Prenatal assessment of CDH severity was based on liver herniation and lung size measured as observed‐to‐expected lung‐to‐head ratio (o/e LHR) on fetal ultrasound at 24–28 weeks of gestation. For left CDH, severe CDH was defined as o/e LHR ≤ 25% regardless of liver position, moderate CDH as o/e LHR 26%–35% regardless of liver position and o/e LHR 36%–45% with intrathoracic liver position, and mild CDH as o/e LHR 36%–45% with intraabdominal liver position and all patients with o/e LHR ≥ 46%. For right CDH, severe CDH was defined as o/e LHR < 45% and moderate CDH as o/e LHR ≥ 45%.

2.3. Calculation of Oxygenation Indices

Blood samples were taken via an arterial line. The position of the arterial line varied between patients and was labeled as preductal and postductal, according to their position. In our department, a postductal arterial line position was preferred. The calculations for the different indices are [12]:

The OI is calculated as follows:

$$\text{OI}=\text{MAP}\left[{\text{cm H}}_2\mathrm{O}\right]\times {\text{FiO}}_2\times 100/{\text{PaO}}_2 \left[\text{mmHg}\right].$$

The pre‐ and postductal OSI was calculated as follows:

$$\text{OSI}=\text{MAP}\left[{\text{cmH}}_2\mathrm{O}\right]\times {\text{FiO}}_2\times 100/{\text{SpO}}_2[\%].$$

The HI is calculated as follows:

$$\text{HI}={\text{PaO}}_2\left[\text{mmHg}\right]/{\text{FiO}}_2.$$

2.4. Statistical Analysis

SPSS version 29 (IBM Corp. Armonk, NY, USA) was used for statistical analysis. Data were described as median (interquartile range) or absolute number (percentage). For statistical analysis, we assigned two outcomes: The primary clinical endpoint was mortality, defined as death during the hospital stay. The secondary clinical endpoint was need for ECMO, defined as neonates receiving ECMO and neonates with contraindications to ECMO who died within 24 h despite fulfilling ECMO criteria. For further analysis, the patients were divided into three groups according to their individual arterial line position: pre‐, postductal, and a mixed total group. The reason for differentiating by arterial line was to compare the correlation between invasive and noninvasive values. Receiver operating characteristic (ROC) analyses and area under the curve (AUC) were used to calculate the optimal cut‐off values for predicting mortality and ECMO initiation based on measurements at hours 1, 2, 3, 4, 6, 12, and 24 of life. To further investigate the strongest predictor of mortality, ROC analysis was performed for lowest OI, lowest OSI, highest HI, and highest PaO₂. Sensitivity, specificity, negative predictive value (NPV) and positive predictive value (PPV) were determined for each cut‐off value. Spearman's rank was used to determine correlations between preductal/postductal highest OSI, highest/lowest OI, highest/lowest PaO₂, and highest and lowest HI. The proportion of a clinically relevant difference in pre‐ to postductal SpO₂ and its correlation with mortality was calculated and compared with χ ² test. A p‐value of < 0.05 was considered statistically significant.

3. Results

3.1. Cohort Description

Overall, 138 CDH neonates met the inclusion criteria and were enrolled. The baseline characteristics of all enrolled patients are presented in Table 1. A postductal arterial line was present in 82 neonates (59.4%) and a preductal arterial line was present in 44 patients (31.9%). In 12 neonates (8.7%) the position of the arterial line was not appropriately documented (e.g., radial artery without side specification).

Table 1.

Baseline characteristics of overall included CDH infants, represented as either total number (n) or median with interquartile range.

Baseline patient characteristics
Variable	n = 138
Male (n)	86 (62.3%)
Gestational age (week)	37.9 [35.6–38.6]
Birth weight (kg)	2.8 [2.4–3.3]
Left‐sided CDH (n)	120 (87.0%)
o/e LHR (%)	38 [30–47]
Liver‐up CDH (n)	75 (54.4%)
FETO (n)	29 (21.0%)
Day of repair (day of life)	6 [4‐8]
Patch‐repair (n)	81 (63.3%)
Highest mean airway pressure (cm H2O)	14 [12–16]
ECMO (n)	56 (40.6%)
Early death (n)	4 (2.9%)
Death (n)	32 (23.2%)
Age at ECMO start (h)	9.8 [6.4–20.4]
Duration ECMO (day)	8.6 [5.2–18.7]

Abbreviations: CDH, congenital diaphragmatic hernia; ECMO, extracorporeal membrane oxygenation; FETO, fetal endoluminal tracheal occlusion; LHR, lung‐to‐head ratio.

3.2. OSI for Outcome Prediction

Results for the preductal and postductal OSI cut‐off values, AUC, sensitivity, specificity, PPV, and NPV at 1, 2, 3, 4, 6, 12, and 24 h of life to predict the primary and secondary endpoints are demonstrated in Tables 2A and 2B. The proportion of patients with valid OSI measurements increased from 28.3% in the first hour of life to 100% from 4 h onward. For the preductal OSI, cut‐off values between 11.5 and 12 in the first 6 h showed a sensitivity of approximately 80%–85% (Hour 1 p < .025; Hours 2–6 p < 0.001). At 12 and 24 h a cut‐off value of 10 predicted the need for ECMO with a very high sensitivity and NPV (sensitivity 96.7% and p < 0.001). The optimal cut‐off values for predicting mortality were slightly higher [12, 13] at the respective timepoints (Table 2A), with a comparable high sensitivity and NPV (p < 0.001). Sensitivity and specificity varied between 70% and 80%. In contrast, postductal OSI cut‐off values were slightly higher at each timepoint to predict need for ECMO and mortality (Table 2B). A clinically relevant pre‐ to postductal SpO₂ difference was observed in 80 patients (58.4%) at least at 1 h during the first 24 h. The mortality was 32.5% and 10.5% (p = 0.004) in patients with and without pre‐ to postductal SpO₂ difference at 1 h, respectively. In 58 patients (42.0%) pre‐ to postductal SpO₂ difference persisted in a least two consecutive hours, with a mortality of 39.7% and 11.3% (p < 0.001), respectively.

Table 2A.

Preductal OSI: prognostic value for clinical endpoints ECMO and mortality.

Outcome	Hours		Area under the curve	Included valid values (n)	p‐value	Cut‐off	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
ECMO	Osi 1		0.722 [0.528–0.916]	39	0.025	11.5	81.8	46.4	37.5	86.7
	Osi 1–2		0.814 [0.734–0.894]	104	< 0.001	12.0	79.6	63.6	66.1	77.8
	Osi 1–3		0.837 [0.769–0.905]	121	< 0.001	12.0	80.4	69.2	69.2	80.4
	Osi 1–4		0.841 [0.775–0.906]	131	< 0.001	11.5	88.1	66.7	68.4	87.3
	Osi 1–6		0.882 [0.828–0.936]	138	< 0.001	11.5	85.0	73.7	71.8	86.2
	Osi 1–12		0.934 [0.895–0.973]	138	< 0.001	10.0	96.7	66.2	69.4	96.2
	Osi 1–24		0.952 [0.921–0.984]	138	< 0.001	10.0	96.7	77.9	77.6	96.8
Mortality	Osi 1		0.954 [0.887–1.021]	39	< 0.001	16.5	100.0	91.7	50.0	100.0
	Osi 1–2		0.844 [0.761–0.928]	104	< 0.001	13.0	79.2	75.0	48.7	92.3
	Osi 1–3		0.839 [0.764–0.915]	121	< 0.001	13.0	75.3	75.0	47.7	90.9
	Osi 1–4		0.849 [0.778–0.920]	131	< 0.001	12.5	74.3	73.3	45.8	90.4
	Osi 1–6		0.845 [0.774–0.916]	138	< 0.001	12.5	76.0	75.0	49.0	90.8
	Osi 1–12		0.862 [0.797–0.926]	138	< 0.001	12.0	73.6	84.4	49.0	94.0
	Osi 1–24		0.862 [0.798–0.926]	138	< 0.001	12.5	77.4	75.0	50.0	91.1

Table 2B.

Postductal OSI: prognostic value for clinical endpoints ECMO and mortality.

Outcome	Hours	Area under the curve	Included valid values (n)	p‐value	Cut‐off	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
ECMO	Osi 1	0.790 [0.569–1.012]	31	0.01	12.0	100.0	42.9	29.4	100.0
	Osi 1–2	0.857 [0.781–0.932]	90	< 0.001	12.2	90.0	68.0	69.2	89.5
	Osi 1–3	0.865 [0.801–0.929]	111	< 0.001	12.8	80.4	75.0	73.2	81.8
	Osi 1–4	0.876 [0.818–0.935]	123	< 0.001	13.2	80.0	79.5	72.7	85.3
	Osi 1–6	0.907 [0.858–0.955]	138	< 0.001	12.5	81.0	86.8	82.5	85.7
	Osi 1–12	0.949 [0.915–0.984]	138	< 0.001	12.4	88.8	92.2	89.3	92.2
	Osi 1–24	0.965 [0.938–0.992]	138	< 0.001	12.1	88.1	93.5	91.2	91.1
Mortality	Osi 1	0.979 [0.922–1.036]	36	< 0.001	21.0	100.0	95.8	66.7	100.0
	Osi 1–2	0.855 [0.771–0.938]	103	< 0.001	14.2	87.5	73.0	41.2	96.4
	Osi 1–3	0.844 [0.766–0.921]	121	< 0.001	14.0	79.2	72.4	44.2	92.6
	Osi 1–4	0.844 [0.769–0.920]	131	< 0.001	13.2	88.5	72.2	46.0	95.9
	Osi 1–6	0.844 [0.770–0.917]	138	< 0.001	12.8	83.3	72.1	46.3	93.8
	Osi 1–12	0.855 [0.786–0.923]	138	< 0.001	13.0	84.0	68.8	56.4	89.9
	Osi 1–24	0.860 [0.794–0.926]	138	< 0.001	13.0	80.2	71.9	52.3	90.4

Figure 1 demonstrates the association of the lowest (best) preductal (Figure 1A) and postductal OSI (Figure 1B) in the first 24 h after birth with outcome following stratification of patients according to the CDH severity determined on prenatal ultrasound. There was no significant difference in the lowest OSI when comparing ECMO survivor to ECMO non‐survivor in all subgroups. All patients with mild left‐sided CDH survived.

Figure 1.

Distribution of the lowest preductal (A) and postductal OSI (B) in the first 24 h (only pre‐ECMO values) in survivors without ECMO (circles), ECMO survivors (squares), and ECMO non‐survivors (pyramids) after group allocation according to the CDH severity based on prenatal ultrasound assessment. The dashed horizontal line represents an OSI cut‐off of 12.5.

3.3. Comparison of OSI and Other Oxygenation Indices

In the group of 82 neonates with a postductal arterial line, the prognostic information of the postductal OSI, in predicting mortality was compared to the OI, the HI, and the PaO₂. All indices demonstrated a comparable AUC using ROC analysis (Table 3). A lowest OI of 18 in the first 24 h of life achieved a sensitivity of 75%, a specificity of almost 90.9% with an NPV of 94.0%. An HI and a PaO₂ of 70 resulted in a specificity of 89.4%. An OI of 10 predicted mortality a sensitivity of 80%. All of the determined cut‐off values have a significance of p ≤ 0.001.

Table 3.

Predictive value of different oxygenation indices for mortality. For comparative purposes only postductal values were included.

Mortality	Area under the curve	Included valid values (n)	p‐value	Cut‐off	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
Lowest OI	0.835 [0.708–0.961]	82	< 0.001	18.0	75.0	90.9	66.7	93.8
Lowest OSI	0.827 [0.728–0.927]	82	< 0.001	10.0	80.0	71.2	38.7	94.0
Highest HI	0.810 [0.659–0.961]	82	< 0.001	70.0	61.1	89.4	68.8	72.0
Highest PaO2	0.814 [0.664–0.965]	82	< 0.001	70.0	61.1	89.4	68.8	72.0

Abbreviations: AUC, area under the curve; HI, Horovitz Index; NPV, negative predictive value; OI, Oxygenation Index; OSI, Oxygenation Saturation Index; PaO₂, partial pressure of arterial oxygenation; PPV, positive predictive value.

Spearman analysis was used to compare prognostic value and correlation of each index and showed excellent correlations in the entire cohort (n = 138), the preductal group (n = 44), and the postductal group (n = 82; Table 4). The strongest correlation was found between the lowest PaO₂ and the lowest HI in the first 24 h (mixed 0.969, pre 0.963, post 0.974). Strong correlations were also observed between the highest OI and the lowest PaO₂ (−0.922 to −0.945). Additionally, the correlation between OI and HI was also excellent (−0.947 to −0.963). Additionally, there was also a strong correlation between OSI (pre‐ and postductal) and OI (−0.822 and −0.892, respectively).

Table 4.

Spearman rank correlation coefficient between the oxygenation indices for the three different cohorts: pre‐ and postductal and mixed pre‐ and postductal. Highest and lowest values used were obtained in the first 24 h of life.

	Highest OSI‐pre	Highest OSI‐post	Highest OI	Lowest PaO2	Lowest Horovitz
Pre‐ and postductal line (n = 138)
Highest OSI‐pre		0.941 (p < 0.001)	0.829 (p < 0.001)	−0.665 (p < 0.001)	−0.708 (p < 0.001)
Highest OSI‐post	0.941 (p < 0.001)		0.832 (p < 0.001)	−0.685 (p < 0.001)	−0.727 (p < 0.001)
Highest OI	0.892 (p < 0.001)	0.832 (p < 0.001)		−0.931 (p < 0.001)	−0.963 (p < 0.001)
Lowest PaO2	−0.665 (p < 0.001)	−0.685 (p < 0.001)	−0.931 (p < 0.001)		0.969 (p < 0.001)
Lowest Horovitz	−0.708 (p < 0.001)	−0.727 (p < 0.001)	−0.963 (p < 0.001)	0.969 (p < 0.001)
Preductal line (n = 44)
Highest OSI‐pre		0.946 (p < 0.001)	0.824 (p < 0.001)	−0.671 (p < 0.001)	−0.722 (p < 0.001)
Highest OSI‐post	0.946 (p < 0.001)		0.835 (p < 0.001)	−0.707 (p < 0.001)	−0.741 (p < 0.001)
Highest OI	0.824 (p < 0.001)	0.835 (p < 0.001)		−0.922 (p < .001)	−0.964 (p < 0.001)
Lowest PaO2	−0.671 (p < 0.001)	−0.707 (p < 0.001)	−0.922 (p < 0.001)		0.963 (p < 0.001)
Lowest Horovitz	−0.722 (p < 0.001)	−0.741 (p < 0.001)	−0.964 (p < 0.001)	0.963 (p < 0.001)
Postductal line (n = 82)
Highest OSI‐pre		0.952 (p < 0.001)	0.841 (p < 0.001)	−0.708 (p < 0.001)	−0.738 (p < 0.001)
Highest OSI‐post	0.952 (p < 0.001)		0.822 (p < 0.001)	−0.683 (p < 0.001)	−0.728 (p < 0.001)
Highest OI	0.841 (p < 0.001)	0.822 (p < 0.001)		−0.945 (p < 0.001)	−0.967 (p < 0.001)
Lowest PaO2	−0.708 (p < 0.001)	−0.683 (p < 0.001)	−0.945 (p < .001)		0.974 (p < 0.001)
Lowest Horovitz	−0.738 (p < 0.001)	−0.728 (p < 0.001)	−0.967 (p < 0.001)	0.974 (p < 0.001)

Abbreviations: Horovitz, Horovitz Index; OI, Oxygenation Index; OSI, Oxygenation Saturation Index, PaO₂, partial pressure of arterial oxygenation.

4. Discussion

The standardized and continuous measurement of oxygenation data in the early hours after birth is of uttermost importance and fundamental for the management of respiratory failure in CDH neonates. Our results confirm previous studies of Horn‐Oudshoorn et al. and Hari Gopal et al., which support the utility of continuous, noninvasive measurements in this setting [9, 10, 11]. In this study, continuous measurement of the pre‐ and postductal OSI in the first 24 h of life, was associated with excellent prognostic values to predict the need for ECMO and mortality. Cut‐off values for the pre‐ductal and the postductal OSI were comparable at most timepoints, although it has to be considered, that a clinically relevant difference in the pre‐ to postductal SpO₂ ≥ 8% results in an increase of the OSI of approximately 1.0. In this study, we were able to demonstrate that a pre‐ to postductal SpO₂ difference ≥ 8% at least at one or two consecutive hourly measurements was associated with a three‐ to fourfold increase in the risk of mortality. Furthermore, the OI, the HI, and the PaO₂ also demonstrated a significant association with mortality, despite a poor predictive value of the pre‐ and postductal OSI to distinguish ECMO survivor and ECMO non‐survivor in any CDH severity group based on fetal assessment. Ultimately, the correlation analysis revealed the close prognostic relationship between all investigated indices allowing the discussion over the use of values other than OI.

The OI is the most commonly used score in the postnatal assessment of CDH neonates and a significant association of the OI with outcome such as the initiation of ECMO, mortality, and the development of chronic lung disease has been demonstrated [9]. A higher OI value indicates more severe respiratory failure and is associated with a more complicated hospital course [13]. The timing of ECMO initiation is important, as the treatment itself may have an impact on neurodevelopmental outcome and premature commencement may cause permanent damage to the already fragile lung tissue [14]. CDH leads to the typical picture of PH, lung hypoplasia, and left ventricular dysfunction, resulting in distinguishable shunting pattern via the patent ductus arteriosus and the foramen ovale [15]. Therefore, PaO₂ values might differ when a blood gas analysis is obtained from a pre‐ductal versus a postductal arterial line position. To our knowledge, no study has compared the predictive capacity for pre‐ and postductal OI in CDH. Considering that most hospitals use preductal measurements to evaluate their therapy, our study allows comparison by dividing the three cohorts into pre‐ductal, postductal, and a mixed total group, showing that the measurements correlate well and provide comparable results.

These points draw attention to the OSI, which does not require an arterial line, but is calculated from pulse oximetry saturation values and may provide the clinician with continuous noninvasive information on the oxygenation status. Similar to the OI, higher values indicate more severe respiratory failure. However, the OSI is also associated with some limitations and subject to potential misinterpretation. Reduced peripheral perfusion or the potential falsification of values due to pigmented skin are potential sources of error [16, 17, 18]. In our study, the OSI cut‐off value to predict mortality was highest in the first hour of life with a subsequent decline at later timepoints. Due to the retrospective design of our study, it is difficult to draw a substantial conclusion whether this was associated with a potential selection bias or reflects a drop in the pulmonary vascular resistance or an improvement in cardiac function with better peripheral perfusion. A similar pattern with the highest OSI cut‐off values in the first 1–6 h of life to predict mortality has been observed in a recent study by Horn‐Oudshoorn et al. [11] Although the cut‐off values were chosen somewhat higher in their study, the AUC for the OSI to predict mortality and need for ECMO support was comparable to our study. Additionally, a moderate correlation of the mean OSI values and the presence of moderate to severe PH was observed. We refrained to include PH severity in our study for two reasons. First, there is still no universally accepted definition for the diagnosis of PH in the first days of life and second, PH might be aggravated by impaired left ventricular function leading to post‐capillary PH which is distinguishable from pre‐capillary PH in terms of pathophysiology and treatment strategies [19]. Despite the aforementioned limitations, the OSI appears to be a reasonable alternative to OI in the first 24 h of life. Being a noninvasive parameter with an ultrashort turnaround time, the OSI provides clinicians with real‐time oxygenation data directly at the bedside without the need for blood sampling. Measuring PaO₂ and calculation of the OI using blood gas analysis should not be completely omitted in CDH neonates, due to the risk of hyperoxia and its detrimental effects. We recommend the OSI as a fast, noninvasive score that clinicians may use to monitor clinical deterioration and early risk, with the acknowledgment that further well‐designed prospective studies are needed to validate cut‐off values.

Another index we studied is the HI, which plays an important role in risk stratification in pediatric and adult ARDS. A higher HI reflects the pulmonary ability of adequate oxygenation and this index is not influenced by ventilator settings. In our study, the HI was similar to the OI and the OSI associated with excellent results in predicting neonatal survival and correlated strongly with the OI. The significance and impact of PaO₂ on prognosis has already been shown by Terui et al. in a study with preductal PaO₂ [20]. The relationship between the OI and the HI is close and different MAP values might have a moderate effect on the OI when kept within a recommended range (i.e., ≤ 17 cmH₂O). Although optimal ventilatory settings for CDH neonates during early postnatal transition have yet to be defined, it is important to note that increasing the MAP may result in higher PaO₂ levels by improving alveolar opening. There is a risk of overdistension of the lung at high MAPs, which may lead to compression of the alveolar capillaries, worsening the perfusion–ventilation mismatch and the pulmonary vascular resistance with a subsequent decrease of the PaO₂ [17]. In this context, the interaction between MAP and PEEP and ultimately the influence on OI must be considered. According to data from a study based on electrical impedance tomography monitoring, a low PEEP, which ultimately lowers MAP, might be beneficial for lung aeration in CDH neonates. This would have implications for the use of OI and its current reference values [21].

Due to the impaired postnatal transition in CDH neonates, cut‐off values of the applied indices in the first hours of life may differ from those at 24 h. This is related to many factors, such as the ventilator setting as well as the individual cardiovascular adaptation of the neonate to the extrauterine environment [22]. Finding the right treatment settings and balancing the individual parameters for children with CDH remains challenging. Several pathophysiological factors need to be considered in the assessment of the clinical status of CDH neonates during early postnatal transition. Beside oxygenation failure due to lung hypoplasia, both PH and cardiac dysfunction are often observed in CDH neonates. All three conditions may lead to poor oxygenation, either due to a small lung volume for gas exchange, or a high pulmonary vascular resistance, or impaired cardiac function [23]. Therefore, oxygenation indices should be interpreted in the context of other comorbidities such as PH or cardiac dysfunction. Additionally, specific cardiopulmonary interactions may be associated with poor oxygenation in this population. The interpretation of changes in the oxygenation status should always include an assessment of the cardiovascular system as well. In this context, echocardiography is an important technique to guide treatment strategies in CDH neonates with deteriorating oxygenation indices.

The main limitations of our study are its retrospective nature and the varying location of the arterial line. Due to individual circumstances, we were not able to use the same arterial line position to collect our information, however, by allocating patients into groups according to their arterial line position and analyzing them separately we have attempted to overcome this issue. The potential limitations of pulse oximetry measurements mentioned earlier might be a limiting factor that needs to be investigated in further studies. Data from the very first hour of life was rather small in comparison, although the overall sample size with a large data pool over 24 h allowed for a robust statistical analysis.

5. Conclusion

Our study demonstrates the close correlation of the OI, the OSI, the HI, and the PaO₂ and the usefulness of the OSI as a parameter for outcome prediction in CDH neonates. Despite some inherited limitations, the OSI seems to be a reasonable alternative to the OI in the first 24 h of life. Being a noninvasive parameter with an ultra‐short turn‐around‐time, the OSI is available directly at the patient's bedside, without the need for blood sampling and processing. Well‐designed prospective studies are needed to elucidate the role of all different indices for outcome prediction of CDH neonates.

Author Contributions

Lennart Hale: methodology, software, data curation, investigation, validation, formal analysis, visualization, writing–original draft. Judith Leyens: methodology, investigation, supervision, writing–original draft. Bartolomeo Bo: investigation, supervision, writing–review and editing. Clara Engel: data curation, investigation, writing–review and editing, project administration. Christoph Berg: methodology, investigation, validation, supervision, writing–review and editing. Lukas Schroeder: investigation, supervision, project administration, writing–review and editing. Andreas Mueller: conceptualization, supervision, writing–review and editing, project administration. Florian Kipfmueller: conceptualization, methodology, software, data curation, investigation, validation, formal analysis, supervision, visualization, project administration, funding acquisition, resources, writing–review and editing.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

Data supporting the findings of this study are available from the corresponding author upon reasonable request within the legal regulations of data protection rights.