Your privacy, your choice

We use essential cookies to make sure the site can function. We also use optional cookies for advertising, personalisation of content, usage analysis, and social media.

By accepting optional cookies, you consent to the processing of your personal data - including transfers to third parties. Some third parties are outside of the European Economic Area, with varying standards of data protection.

See our privacy policy for more information on the use of your personal data.

for further information and to change your choices.

Skip to main content

Limitations of SpO2 / FiO2-ratio for classification and monitoring of acute respiratory distress syndrome—an observational cohort study

Abstract

Background

The ratio of pulse-oximetric peripheral oxygen saturation to fraction of inspired oxygen (SpO2/FiO2) has been proposed as additional hypoxemia criterion in a new global definition of acute respiratory distress syndrome (ARDS). This study aims to evaluate the clinical and theoretical limitations of the SpO2/FiO2-ratio when using it to classify patients with ARDS and to follow disease progression.

Methods

Observational cohort study of ARDS patients from three high-resolution Intensive Care Unit databases, including our own database ICU Cockpit, MIMIC-IV (Version 3.0) and SICdb (Version 1.0.6). Patients with ARDS were identified based on the Berlin criteria or ICD 9/10-codes. Time-matched datapoints of SpO2, FiO2 and partial pressure of oxygen in arterial blood (PaO2) were created. Severity classification followed the thresholds for SpO2/FiO2 and PaO2/FiO2 of the newly proposed global definition.

Results

Overall, 708 ARDS patients were included in the analysis. ARDS severity was misclassified by SpO2/FiO2 in 33% of datapoints, out of which 84% were classified as more severe. This can be partially explained by imprecision of SpO2 measurement and equation used to transform SpO2/FiO2 to PaO2/FiO2. A high dependence of SpO2/FiO2-ratio on FiO2 settings was found, leading to major treatment effect and limited capability for tracking change in ARDS severity, which was achieved in less than 20% of events.

Conclusions

The use of SpO2/FiO2 interchangeably with PaO2/FiO2 for severity classification and monitoring of ARDS is limited by its inadequate trending ability and high dependence on FiO2 settings, which may influence treatment decisions and patient selection in clinical trials.

Background

Current and former definitions of acute respiratory distress syndrome (ARDS) include a hypoxemia criterion [1,2,3] for diagnosis and severity classification. Traditionally, this has been assessed by the ratio of partial pressure of oxygen in arterial blood (PaO2) divided by the fraction of inspired oxygen (FiO2) [3], which was used by major interventional trials in ARDS [4,5,6]. Because of its simplicity and practicality, the PaO2/FiO2-ratio is the most widely used surrogate of oxygen transfer in the lungs to follow disease progression and response to therapy in patients with and without ARDS [7]. The main disadvantage of PaO2/FiO2-ratio is the need of invasive arterial blood gas (ABG) sampling, which is only available intermittently, may not be readily available in resource-limited regions and is associated with some complications, such as vascular injury, hematoma, infection, thrombosis and nerve injury [8]. There is increasing interest of a continuous, non-invasive surrogate for severity classification and monitoring of ARDS [9], such as the peripheral oxygen saturation (SpO2) divided by FiO2 (SpO2/FiO2) ratio. The SpO2/FiO2-ratio as a surrogate of PaO2/FiO2-ratio has been evaluated, showing good performance as long as SpO2 is ≤ 97% [10,11,12,13,14]. Recently, a new global definition of ARDS [15] added the SpO2/FiO2-ratio in an effort to provide an ARDS definition that is more suitable for resource-limited settings. Despite, that ARDS patients.

diagnosed with the SpO2/FiO2-ratio, seem to have similar outcomes compared to patients diagnosed with the PaO2/FiO2-ratio [16], these scores may not be used in parallel [17].

We hypothesized, that the SpO2/FiO2-ratio can be used as a surrogate for PaO2/FiO2 and that classification of ARDS is similar with both ratios. To assess the limitations of the SpO2/FiO2-ratio as a surrogate for PaO2/FiO2-ratio in the classification of ARDS severity, we focused on classification accuracy and monitoring of disease progression using retrospective data from three high-resolution ICU databases. Additionally, we provide theoretical and technical explanations for classification discrepancies.

Methods

Study design and datasets

This is a retrospective observational cohort study of ARDS patients using data from three high-resolution ICU databases: ICU Cockpit [18], Medical Information Mart for Intensive Care (MIMIC)-IV (version 3.0) [19, 20] and SICdb (version 1.0.6) [21, 22].

MIMIC-IV, is a large deidentified dataset of patients admitted to the Beth Israel Deaconess Medical Center in Boston (USA), between 2008 and 2022. It contains comprehensive information for over 65,000 patients admitted to an intensive care unit (ICU), including free-text radiology reports. Most of the continuous signals are provided in a resolution of one hour.

The SICdb database is a deidentified medical records dataset of patient admitted to the ICU at the University Hospital Salzburg (Austria), between 2013 and 2021 and contains vital signs, laboratory results, medication data and admission details for over 27,000 ICU admissions. SICdb provides continuous time signals with highly granular once-per-minute data.

ICU cockpit is a big data platform collecting high-resolution (up to 200 Hz for continuous signals) multimodal data since 2016, including vital signs, ventilator settings and measurements, neuromonitoring data, laboratory analyses and clinical annotations. The platform includes over 2400 patients from the Neurocritical Care Unit and the medical ICU of the University Hospital Zurich (Switzerland).

The collection and usage of data from ICU cockpit was approved by the local ethics commission (Cantonal Ethics Commission of Zurich, BASEC Nr. PB 2016–01101). Access to MIMIC-IV and SICdb was provided through PhysioNet [23] after a credential process including training and signing a data use agreement.

ARDS population

Adult patients with ARDS were selected in the ICU Cockpit database by manually applying the Berlin definition [3]. Patients in the external databases SICdb and MIMIC-IV were identified by ICD codes (ICD-9: 51,882, ICD-10: J80). The hypoxemia criterion was modified for all databases to include patients with PaO2/FiO2 ≤ 300 mmHg measured during invasive mechanical ventilation, non-invasive ventilation (NIV) or continuous positive airway pressure (CPAP) with positive end-expiratory pressure (PEEP) ≥ 5 cmH2O or during high-flow nasal oxygen (HFNO) with flow ≥ 30 L/min. If patients were admitted multiple times to the ICU during one hospital stay, only the first ICU admission was kept, as subsequent ICU admissions likely represent complicated progression of disease, rather than newly developed ARDS. Patients that were supported by extracorporeal membrane oxygenation were excluded from analysis.

Because ARDS diagnosis is frequently missed by clinicians [24], the analysis was repeated in an extended population of the MIMIC-IV database by manually selecting ICU admissions based on ICU chart data, ABG results, chest radiography notes and/or ICD codes.

Variables and data management

For each time point of ABG measurement a set of variables containing: PaO2, FiO2 and SpO2 was created (subsequently referred to as datapoints). For MIMIC-IV, whose data was mainly on hourly resolution, matching between FiO2, SpO2 and PaO2 was performed with a resolution of 30 min. In ICU Cockpit the correct time point of ABG sampling was identified with arterial pressure waveform analysis using gaps in pressure readings. When gap in pressure data was not located within 15 min prior to ABG analysis timestamp, median value between 5 and 2 min before ABG time stamp was calculated both for SpO2 and FiO2 and matched with PaO2 values to form datapoints. In SICdb, time delay allowed for FiO2 (and SpO2) and PaO2 matching was 5 min too. Datapoints were considered valid if SpO2 was ≤ 97% and measurements were taken during invasive mechanical ventilation/NIV/CPAP with PEEP ≥ 5 cmH2O or HFNO with flow ≥ 30 L/min. Incomplete sets of variables were excluded from analysis.

ARDS severity was calculated with the PaO2/FiO2-ratio and the SpO2/FiO2-ratio. Additional, most severe category per ICU admission was calculated requiring at least two consecutive classifications into an equally or more severe category. Thresholds for severity categories followed the new global definition [15] (Table 1).

Table 1 ARDS severity classifications according to the 2024 Global Definition [15]

Analysis

Clinical performance of the SpO2/FiO2-ratio was evaluated on a datapoint and an ICU admission level using overall accuracy (correct classifications divided by total classifications) as a primary outcome. We further analyzed the following secondary outcomes. Accuracy per severity category was visualized using confusion matrices. Impact of FiO2 settings on ARDS severity category, was assessed by density plots of FiO2 and through analysis of accuracy per FiO2 values (grouped by 5%). We evaluated trending ability by correlating corresponding changes in FiO2, SpO2 and PaO2 over two consecutive datapoints in the same patient. For assessing the effect of FiO2 changes over time, the effect of true changes in oxygenation capacity between consecutive datapoints was reduced by excluding datapoint-pairs with time-interval > 6 h and changes in respiratory index > 20%. The respiratory index (RI) [25] is a tension-based index of oxygenation, calculated as A-a gradient [26] divided by PaO2 and was shown to be superior to other indices of oxygenation regarding variability with different FiO2 values [27].

Limitations of SpO2 measurements were evaluated through comparison with arterial oxygen saturation (SaO2) from ABG, calculating bias and precision. A Bland–Altman plot was not used, because of a heteroscedastic bias leading to misinterpretation of the limits of agreement. Finally, conversion between PaO2/FiO2 and SpO2/FiO2 was evaluated by comparing our data and published linear [10,11,12] and log-linear [13, 14] imputations using performance metrics R2 and mean absolute error.

In the paper, we report results for three datasets combined, while analogous results for the extended MIMIC-IV population are shown in the Supplementary Information Figs. S7S9, to address a possible selection bias introduced by using the ICD codes.

All the analysis and graphical output was performed using Python version 3.12.2.

Results

Population

Combining three databases led to the identification of 11,916 valid datapoints representing 708 ICU admissions. While MIMIC-IV had the highest number of admissions, data resolution was higher in ICU-Cockpit and SICdb (Table 2). Flowcharts with detailed results of the selection process are provided in the Supplementary Information (see Supplementary Information Figs. S1 and S2).

Table 2 Number of admissions and datapoints per database

Based on the PaO2/FiO2-ratio, datapoints were classified as no, mild, moderate, and severe ARDS in 3.2%, 15.6%, 61.7%, 19.6%. On an admission level 7.9%, 7.2%, 47.0% and 37.9% patients were classified as no, mild, moderate, and severe ARDS (see Supplementary Information Fig. S3).

Clinical performance of the SpO2/FiO2-ratio

Alignment of ARDS severity categories was 69.1% comparing ARDS severity per admission (Fig. 1a), and 67.1% considering individual datapoints (see Supplementary Information Fig. S4a). Results were similar in all databases (Fig. S5). Performance was best in more severe PaO2/FiO2 categories (Fig. 1b, Supplementary Information Fig. S4b). The SpO2/FiO2-ratio overestimated the PaO2/FiO2-ratio category in 28.0% and underestimated it in 2.9% of admissions. Accuracy differed between databases with high time-resolution (SpO2 measured every second in ICU Cockpit, and every minute in SICdb) and MIMIC-IV, which has lower granularity of one hour (Fig. 1b and Supplementary Information Fig. S4b).

Fig. 1
figure 1

Performance of the SpO2/FiO2-ratio for ARDS severity classification evaluated on admission level and presented as a confusion matrix and b recall (sensitivity) per ARDS category. Numbers are presented as percentages of PaO2/FiO2 category

Influence of FiO2 settings

A moderate/good correlation was found between SpO2/FiO2 and PaO2/FiO2 (r = 0.69). SpO2/FiO2-ratio was highly influenced by FiO2 setting, visualized by separation of ARDS severity categories by FiO2 (Fig. 2a). Similar, FiO2 settings were more widely distributed using the PaO2/FiO2-ratio and more distinct using the SpO2/FiO2-ratio categories (see Supplementary Information Fig. S6). Classification accuracy also differs by FiO2 settings, ranging from an accuracy of 22% at FiO2 70 (± 2.5)% to an accuracy of 91% at FiO2 55 (± 2.5)% as presented in Fig. 2c.

Fig. 2
figure 2

Impact of FiO2 on accuracy of SpO2/FiO2-ratio. a PaO2/FiO2 vs. SpO2/FiO2 colored by FiO2. Lines represent thresholds of SpO2/FiO2 (solid) and PaO2/FiO2 (dashed). b ARDS severity classification accuracy in respect to FiO2. Dashed line marks an accuracy of 50%

Trending ability of the SpO2/FiO2-ratio

The SpO2/FiO2-ratio correctly detected changes in PaO2/FiO2-ratios severity categories between two consecutive datapoints in only 19.6% of changes (Fig. 3a).

Fig. 3
figure 3

Trending ability of SpO2/FiO2-ratio. a Confusion matrix for changes in ARDS severity category. Proportional change of b SpO2 and c PaO2 when FiO2 changed during a stable respiratory condition (Respiratory Index ± 20%). Numbers are presented as percentages

During relatively stable respiratory conditions (RI within ± 20%), proportional changes in FiO2 were strongly correlated with proportional changes in PaO2 (r = 0.88), indicating similar PaO2/FiO2-ratios. These changes in FiO2 were not correlated with proportional changes in SpO2 (r = 0.44), indicating that FiO2 changes alter the SpO2/FiO2-ratio disproportionately (Fig. 3b and c).

Limitations of SpO2 measurements

Mean difference (bias) and precision of SpO2 measurements compared to SaO2 were 0.5 ± 2.1%. Bias showed a heteroscedastic trend towards negative bias below and positive bias above an oxygen saturation of 94% (Fig. 4a). The relationship between PaO2 and SaO2 can be described by the oxyhemoglobin dissociation curve proposed by Severinghaus [28]. This relationship was diluted for SpO2 (Fig. 4b).

Fig. 4
figure 4

Relationship between SpO2, SaO2 and PaO2. a SpO2 vs. SpO2–SaO2. The midhorizontal line marks the mean difference between SpO2 and SaO2 (bias, + 0.5%). b Measured SpO2, SaO2 and PaO2. The grey curve represents the extrapolated oxyhemoglobin dissociation curve by Severinghaus [28]. Histograms for SaO2, SpO2 (right) and PaO2 (top) are shown. SpO2 was not restricted to ≤ 97%, but datapoints with a PaO2 above 400 mmHg were excluded for visualization

Limitations of conversion from SpO2/FiO2 to PaO2/FiO2

Out of the linear imputations of PaO2/FiO2 from SpO2/FiO2, the equation from Rice et al. [10] best fitted the data, whereas out of the log-linear equations, the one proposed by Pandharipande et al. [13] fitted best. The fits of these equations are presented in Fig. 5. Performance data is available in Table S1.

Fig. 5
figure 5

Relationship between PaO2/FiO2 and SpO2/FiO2 and comparison of best linear fit with imputations proposed in the literature [10,11,12,13,14]

Discussion

In this retrospective analysis using data from ARDS patients out of three ICU databases, we showed that classification of ARDS severity by SpO2/FiO2-ratio leads to misclassification in 33% of datapoints and 31% of admissions. Further, SpO2/FiO2-ratio showed limited trending ability of disease progression and high dependence on FiO2 setting.

The SpO2/FiO2-ratio classified two thirds of datapoints correctly into ARDS severity categories. Of the misclassified events, the majority was overestimated in severity (84% of the misclassified datapoints). The misclassification rate is strongly influenced by FiO2, with the lowest accuracy observed at a FiO2 level of 70%, where application of the SpO2/FiO2-ratio inevitably assumes severe ARDS, irrespective of SpO2. While proportional changes in FiO2 let to similar changes in PaO2 (r = 0.88), we found that SpO2 reacts only poorly (r = 0.42) to changes in FiO2. This introduces a major treatment effect, as SpO2/FiO2 is determined primarily by FiO2 settings rather than measured SpO2. Changing the FiO2 to a higher value will most likely lead to a lower SpO2/FiO2-ratio despite unchanged gas exchange conditions. This hinders any comparison of patients with different FiO2 or even the same patient at different time points.

The PaO2/FiO2 ratio, is also not a perfect surrogate of pulmonary oxygen gas transfer as it is not reliably reflecting intrapulmonary shunt [29], is not independent from FiO2 [30] and influenced by extrapulmonary factors like hemoglobin [7]. Furthermore, PaO2/FiO2-ratio is not stable during transition from HFNO or NIV to mechanical ventilation [31]. Nevertheless the PaO2/FiO2-ratio showed better stability over FiO2 changes compared to the SpO2/FiO2-ratio [32]. While RI shares some of these limitations, it was shown to correlate best with intrapulmonary shunt out of all tension-based indices of oxygenation [29] and has less variability over the range of FiO2 compared to the PaO2/FiO2 ratio [27]. For evaluating the influence of FiO2 changes, we therefore chose a stable Respiratory Index (± 20%) as a prerequisite to exclude large changes in pulmonary oxygen transfer capability, which would make interpretation difficult.

PaO2/FiO2 has been used extensively to track disease progression and response to interventions [5]. A continuous estimate of pulmonary gas transfer may enhance detection of clinical worsening and would fill the gaps between intermittent ABGs for more complex machine learning algorithms. Our data showed, that the SpO2/FiO2-ratio captures severity category changes correctly only in about 20% of events when PaO2/FiO2 severity category changed. As SpO2/FiO2 primarily reacts on treatment changes (FiO2) which probably occur after worsening oxygenation was detected on ABG, the SpO2/FiO2-ratio has limited use for monitoring of disease.

While the accuracy of pulse oximetry may be reasonable for avoiding hypoxemia [33], SpO2 may overestimate [34] or underestimate [35] arterial oxygen saturation, which probably depends on the type of oximeter [36, 37] and skin pigmentation [38]. We found a heteroscedastic bias for SpO2 of + 0.5% compared to SaO2, resulting in negative bias below and positive bias above 94%. A precision of ± 2.2% in our data is in line with other studies [36, 39], some of which also found a heteroscedastic bias for SpO2 [35]. SpO2 measurements are prone to artifacts, mainly low-perfusion states and movement artifacts [40], which were minimized in this analysis by using the median SpO2 over a short time interval. In clinical practice this means, that artifacts need to be carefully excluded before drawing assumptions based on SpO2 and device internal computations may differ between centers.

Oximetry is relatively insensitive in detection of oxygenation in patient with high PaO2 [41], because of the sigmoid shape of the oxyhemoglobin dissociation curve. Partial pressure of carbon dioxide, pH and temperature are known to shift this curve and therefore lead to imprecision [42]. However, SaO2 seem to follow the data from the Severinghaus Eq. (28) quite well, suggesting that the main reason for using a linear approach for converting SpO2/FiO2 to PaO2/FiO2 may be the missing precision of SpO2. Indeed, in our data, SpO2 measurements ≤ 97% followed best a linear equation.

Comparing other equations for conversion of SpO2/FiO2 to PaO2/FiO2 [10,11,12,13,14], the equation from Rice et al. [10], fitted the data better than other linear imputations. Non-linear approaches may fit the data even better, but are not easily calculated at bedside. A different equation or other thresholds for ARDS severity categories may marginally improve discrimination, but this does not solve the problem of the overweighed effect of FiO2. Rescaling (e.g. multiplying SpO2 in the equation) may be an option, but bears the risk of inflating imprecision of SpO2 measurements.

This study has some limitations, including the retrospective design, which may lead to selection bias. To minimize this risk, we included an extended population from MIMIC-IV, where patients were manually selected based on ARDS criteria and achieved similar results. The resolution of time data such as SpO2 in MIMIC-IV hinders perfect matching of SpO2 and PaO2 values and may introduced inaccuracy. We therefore chose to include three different databases, despite unequal distribution of patients. The analysis is still most affected by MIMIC-IV, which has the highest number of patients. Results remain similar when restricting the analysis to the ICU Cockpit or SICdb, which cover data with high time-resolution and allow matching of the exact ABG time point by analyzing the gap in invasive blood pressure readings (ICU Cockpit). Only a slightly better accuracy, however was found for high-resolution data. We therefore consider the results generalizable to other centers. We did not collect outcome data, because classification based on SpO2/FiO2 was not established in high resource areas, where regular ABG measurements are available. A retrospective outcome comparison would therefore ignore treatment effects. Future research should compare outcomes of patients in settings where the new global definition is in use. Regarding this study, we do not know if classification discrepancies are associated with harm or cause differences in treatment such as the application of lung protective ventilation, time of intubation, allowing spontaneous breathing efforts or changes in resource allocation. Finally, the analysis did not differentiate between HFNO and mechanical ventilation. It remains an open question whether the performance differs between the two modes of respiratory support.

In summary, while the SpO2/FiO2 ratio seems to discriminate ARDS patients into categories that have similar mortality compared to the PaO2/FiO2 ratio [16, 17], and has relatively good correlation with PaO2/FiO2 (r = 0.69), we showed that these two indices should not be used interchangeably. In resource-limited settings without the ability to draw ABGs, we suggest creating standardized settings, e.g. avoid movement artifacts, and optimize perfusion. Further, due to strong impact of FiO2, we suggest minimizing FiO2, so that SpO2 is at the lower limit of target range. However, as target SpO2 range may vary across centers, due to conflicting evidence of a best target range [43,44,45,46,47], ARDS severity classification may still not be comparable. Alternatively, better performance using non-invasive approaches to classify ARDS may be achieved through more complex machine learning models.

Conclusion

The use of the SpO2/FiO2-ratio for severity classification in ARDS interchangeably with the PaO2/FiO2-ratio is limited by its high dependence on FiO2 settings, the inability to follow disease progression and the imprecision of SpO2 measurements. This discrepancy in severity classification may influence treatment decisions and patient selection in clinical trials.

Availability of data and materials

The databases MIMIC-IV and SICdb are freely-accessible through PhysioNet [23] after training and signing the data use agreement. Retrieval codes for this data are available from the corresponding author on reasonable request. The database ICU Cockpit is not publicly available due to reasons of sensitivity, based on the protocol approved by the Cantonal Ethics Commission of Zurich. Some metadata are published on Zenodo (https://zenodo.org/records/12635089). The study report adheres to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement [48].

Abbreviations

ABG:

Arterial blood gas

ARDS:

Acute respiratory distress syndrome

CPAP:

Continuous positive airway pressure

FiO2 :

Fraction of inspired oxygen

HFNO:

High-flow nasal oxygen

ICD:

International classification of diseases

ICU:

Intensive care unit

MIMIC:

Medical information mart for intensive care

NIV:

Non-invasive ventilation

PaO2 :

Partial pressure of oxygen in arterial blood

PEEP:

Positive end-expiratory pressure

RI:

Respiratory index

SaO2 :

Arterial oxygen saturation

SpO2 :

Peripheral oxygen saturation

STROBE:

Strengthening the reporting of observational studies in epidemiology

References

  1. Murray JF, Matthay MA, Luce JM, Flick MR. An expanded definition of the adult respiratory distress syndrome. Am Rev Respir Dis. 1988;138(3):720–3.

    Article  CAS  PubMed  Google Scholar 

  2. Bernard GR, Artigas A, Brigham KL, Carlet J, Falke K, Hudson L, et al. Report of the American-European consensus conference on ARDS: definitions, mechanisms, relevant outcomes and clinical trial coordination. Consens Comm Intensive Care Med. 1994;20(3):225–32.

    CAS  Google Scholar 

  3. Force ADT, Ranieri VM, Rubenfeld GD, Thompson BT, Ferguson ND, Caldwell E, et al. Acute respiratory distress syndrome: the berlin definition. JAMA. 2012;307(23):2526–33.

    Google Scholar 

  4. Combes A, Hajage D, Capellier G, Demoule A, Lavoue S, Guervilly C, et al. Extracorporeal Membrane Oxygenation for Severe Acute Respiratory Distress Syndrome. N Engl J Med. 2018;378(21):1965–75.

    Article  PubMed  Google Scholar 

  5. Guerin C, Reignier J, Richard JC, Beuret P, Gacouin A, Boulain T, et al. Prone positioning in severe acute respiratory distress syndrome. N Engl J Med. 2013;368(23):2159–68.

    Article  CAS  PubMed  Google Scholar 

  6. Brower RG, Matthay MA, Morris A, Schoenfeld D, Thompson BT, Wheeler A. 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(18):1301–8.

    Article  PubMed  Google Scholar 

  7. Feiner JR, Weiskopf RB. Evaluating pulmonary function: an assessment of PaO2/FIO2. Crit Care Med. 2017;45(1):e40–8.

    Article  PubMed  Google Scholar 

  8. Baskin SB, Oray NC, Yanturali S, Bayram B. The comparison of heparinized insulin syringes and safety-engineered blood gas syringes used in arterial blood gas sampling in the ED setting (randomized controlled study). Am J Emerg Med. 2014;32(5):432–7.

    Article  PubMed  Google Scholar 

  9. Riviello ED, Kiviri W, Twagirumugabe T, Mueller A, Banner-Goodspeed VM, Officer L, et al. Hospital incidence and outcomes of the acute respiratory distress syndrome using the kigali modification of the berlin definition. Am J Respir Crit Care Med. 2016;193(1):52–9.

    Article  PubMed  Google Scholar 

  10. Rice TW, Wheeler AP, Bernard GR, Hayden DL, Schoenfeld DA, Ware LB, et al. Comparison of the SpO2/FIO2 ratio and the PaO2/FIO2 ratio in patients with acute lung injury or ARDS. Chest. 2007;132(2):410–7.

    Article  PubMed  Google Scholar 

  11. Neto AS, Schultz MJ, Festic E, Adhikari NKJ, Dondorp AM, Pattnaik R, et al. Ventilatory support of patients with sepsis or septic shock in resource-limited settings. In: Dondorp AM, Dunser MW, Schultz MJ, editors., et al., Sepsis management in resource-limited settings. Cham: Springer; 2019. p. 131–49.

    Chapter  Google Scholar 

  12. Tripathi RS, Blum JM, Rosenberg AL, Tremper KK. Pulse oximetry saturation to fraction inspired oxygen ratio as a measure of hypoxia under general anesthesia and the influence of positive end-expiratory pressure. J Crit Care. 2010;25(3):542.e9-542.e13. https://doi.org/10.1016/j.jcrc.2010.04.009.

    Article  PubMed  Google Scholar 

  13. Pandharipande PP, Shintani AK, Hagerman HE, St Jacques PJ, Rice TW, Sanders NW, et al. Derivation and validation of Spo2/Fio2 ratio to impute for Pao2/Fio2 ratio in the respiratory component of the sequential organ failure assessment score. Crit Care Med. 2009;37(4):1317–21.

    Article  PubMed  PubMed Central  Google Scholar 

  14. Schmidt MF, Gernand J, Kakarala R. The use of the pulse oximetric saturation to fraction of inspired oxygen ratio in an automated acute respiratory distress syndrome screening tool. J Crit Care. 2015;30(3):486–90.

    Article  PubMed  Google Scholar 

  15. Matthay MA, Arabi Y, Arroliga AC, Bernard G, Bersten AD, Brochard LJ, et al. A new global definition of acute respiratory distress syndrome. Am J Respir Crit Care Med. 2024;209(1):37–47.

    Article  PubMed  Google Scholar 

  16. Chen W, Janz DR, Shaver CM, Bernard GR, Bastarache JA, Ware LB. Clinical characteristics and outcomes are similar in ARDS diagnosed by oxygen saturation/Fio2 ratio compared with Pao2/Fio2 ratio. Chest. 2015;148(6):1477–83.

    Article  PubMed  PubMed Central  Google Scholar 

  17. Qian F, van den Boom W, See KC. The new global definition of acute respiratory distress syndrome: insights from the MIMIC-IV database. Intensive Care Med. 2024;50(4):608–9.

    Article  PubMed  Google Scholar 

  18. Boss JM, Narula G, Straessle C, Willms J, Azzati J, Brodbeck D, et al. ICU Cockpit: a platform for collecting multimodal waveform data, AI-based computational disease modeling and real-time decision support in the intensive care unit. J Am Med Inform Assoc. 2022;29(7):1286–91.

    Article  PubMed  PubMed Central  Google Scholar 

  19. Johnson A, Bulgarelli L, Pollard T, Gow B, Moody B, Horng S, et al. MIMIC-IV.

  20. Johnson AEW, Bulgarelli L, Shen L, Gayles A, Shammout A, Horng S, et al. MIMIC-IV, a freely accessible electronic health record dataset. Sci Data. 2023;10(1):1.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  21. Rodemund N, Kokoefer, A., Wernly, B., & Cozowicz, C. Salzburg intensive care database (SICdb), a freely accessible intensive care database (version 1.0.6). In: PhysioNet, editor. 2023.

  22. Rodemund N, Wernly B, Jung C, Cozowicz C, Koköfer A. The salzburg intensive care database (SICdb): an openly available critical care dataset. Intensive Care Med. 2023;49(6):700–2.

    Article  PubMed  PubMed Central  Google Scholar 

  23. Goldberger AL, Amaral LA, Glass L, Hausdorff JM, Ivanov PC, Mark RG, et al. PhysioBank, physiotoolkit, and physionet: components of a new research resource for complex physiologic signals. Circulation. 2000;101(23):E215–20.

    Article  CAS  PubMed  Google Scholar 

  24. Bellani G, Laffey JG, Pham T, Fan E, Brochard L, Esteban A, et al. Epidemiology, patterns of care, and mortality for patients with acute respiratory distress syndrome in intensive care units in 50 countries. JAMA. 2016;315(8):788–800.

    Article  CAS  PubMed  Google Scholar 

  25. Laghi F, Siegel JH, Rivkind AI, Chiarla C, DeGaetano A, Blevins S, et al. Respiratory index/pulmonary shunt relationship: quantification of severity and prognosis in the post-traumatic adult respiratory distress syndrome. Crit Care Med. 1989;17(11):1121–8.

    Article  CAS  PubMed  Google Scholar 

  26. Helmholz HF Jr. The abbreviated alveolar air equation. Chest. 1979;75(6):748.

    Article  PubMed  Google Scholar 

  27. Kathirgamanathan A, McCahon RA, Hardman JG. Indices of pulmonary oxygenation in pathological lung states: an investigation using high-fidelity, computational modelling. Br J Anaesth. 2009;103(2):291–7.

    Article  CAS  PubMed  Google Scholar 

  28. Severinghaus JW. Simple, accurate equations for human blood O2 dissociation computations. J Appl Physiol Respir Environ Exerc Physiol. 1979;46(3):599–602.

    CAS  PubMed  Google Scholar 

  29. Cane RD, Shapiro BA, Templin R, Walther K. Unreliability of oxygen tension-based indices in reflecting intrapulmonary shunting in critically ill patients. Crit Care Med. 1988;16(12):1243–5.

    Article  CAS  PubMed  Google Scholar 

  30. Gattinoni L, Vassalli F, Romitti F. Benefits and risks of the P/F approach. Intensive Care Med. 2018;44(12):2245–7.

    Article  CAS  PubMed  Google Scholar 

  31. Ranieri VM, Tonetti T, Navalesi P, Nava S, Antonelli M, Pesenti A, et al. High-flow nasal oxygen for severe hypoxemia: oxygenation response and outcome in patients with COVID-19. Am J Respir Crit Care Med. 2022;205(4):431–9.

    Article  CAS  PubMed  Google Scholar 

  32. Shen Y, Zhu L, Yan J. Stability of Spo2/Fio2 and respiratory rate-oxygenation indexes in critical respiratory disorders. Crit Care Med. 2022;50(8):e694–5.

    Article  CAS  PubMed  Google Scholar 

  33. Jubran A. Pulse oximetry. Critical Care. 2015;19(1):272.

    Article  PubMed  PubMed Central  Google Scholar 

  34. Seguin P, Le Rouzo A, Tanguy M, Guillou YM, Feuillu A, Malledant Y. Evidence for the need of bedside accuracy of pulse oximetry in an intensive care unit. Crit Care Med. 2000;28(3):703–6.

    Article  CAS  PubMed  Google Scholar 

  35. Perkins GD, McAuley DF, Giles S, Routledge H, Gao F. Do changes in pulse oximeter oxygen saturation predict equivalent changes in arterial oxygen saturation? Critical Care. 2003;7(4):R67.

    Article  PubMed  PubMed Central  Google Scholar 

  36. Van de Louw A, Cracco C, Cerf C, Harf A, Duvaldestin P, Lemaire F, Brochard L. Accuracy of pulse oximetry in the intensive care unit. Intensive Care Med. 2001;27(10):1606–13.

    Article  PubMed  Google Scholar 

  37. Blanchet MA, Mercier G, Delobel A, Nayet E, Bouchard PA, Simard S, et al. Accuracy of multiple pulse oximeters in stable critically Ill patients. Respir Care. 2023;68(5):565–74.

    Article  PubMed  PubMed Central  Google Scholar 

  38. Jubran A, Tobin MJ. Reliability of pulse oximetry in titrating supplemental oxygen therapy in ventilator-dependent patients. Chest. 1990;97(6):1420–5.

    Article  CAS  PubMed  Google Scholar 

  39. Webb RK, Ralston AC, Runciman WB. Potential errors in pulse oximetry: II. Effects of changes in saturation and signal quality. Anaesthesia. 1991;46(3):207–12.

    Article  CAS  PubMed  Google Scholar 

  40. Clayton DG, Webb RK, Ralston AC, Duthie D, Runciman WB. A comparison of the performance of 20 pulse oximeters under conditions of poor perfusion. Anaesthesia. 2007;46(1):3–10.

    Article  Google Scholar 

  41. Jubran A. Pulse oximetry. Critical Care. 1999;3(2):R11–7.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  42. Collins JA, Rudenski A, Gibson J, Howard L, O’Driscoll R. Relating oxygen partial pressure, saturation and content: the haemoglobin-oxygen dissociation curve. Breathe. 2015;11(3):194–201.

    Article  PubMed  PubMed Central  Google Scholar 

  43. Barrot L, Asfar P, Mauny F, Winiszewski H, Montini F, Badie J, et al. Liberal or conservative oxygen therapy for acute respiratory distress syndrome. N Engl J Med. 2020;382(11):999–1008.

    Article  CAS  PubMed  Google Scholar 

  44. Mackle D, Bellomo R, Bailey M, Beasley R, Deane A, Eastwood G, Finfer S, Freebairn R, King V, Linke N, Litton E. Conservative oxygen therapy during mechanical ventilation in the ICU. New England J Med. 2019;382(11):989–98.

    Google Scholar 

  45. Schjorring OL, Klitgaard TL, Perner A, Wetterslev J, Lange T, Siegemund M, et al. Lower or higher oxygenation targets for acute hypoxemic respiratory failure. N Engl J Med. 2021;384(14):1301–11.

    Article  PubMed  Google Scholar 

  46. Semler MW, Casey JD, Lloyd BD, Hastings PG, Hays MA, Stollings JL, et al. Oxygen-saturation targets for critically Ill adults receiving mechanical ventilation. N Engl J Med. 2022;387(19):1759–69.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  47. Girardis M, Busani S, Damiani E, Donati A, Rinaldi L, Marudi A, et al. Effect of conservative vs conventional oxygen therapy on mortality among patients in an intensive care unit: the oxygen-ICU randomized clinical trial. JAMA. 2016;316(15):1583–9.

    Article  CAS  PubMed  Google Scholar 

  48. von Elm E, Altman DG, Egger M, Pocock SJ, Gotzsche PC, Vandenbroucke JP, Initiative S. The strengthening the reporting of observational studies in epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet. 2007;370(9596):1453–7.

    Article  Google Scholar 

Download references

Acknowledgements

Not applicable

Funding

The authors declare that they received funding solely from Innosuisse and did not receive any financial support from Hamilton Medical AG.

Author information

Authors and Affiliations

Authors

Contributions

RE, UP and EK contributed to the study conception and design. Material preparation, data collection, analysis and interpretation were performed by all authors. The first draft of the manuscript was written by RE and UP and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Rolf Erlebach.

Ethics declarations

Ethics approval and consent to participate

This retrospective observational study involving human data was in accordance with the ethical standards of the institutional and national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The collection and usage of data from ICU cockpit was approved by the local ethics commission (Cantonal Ethics Commission of Zurich, BASEC Nr. PB_2016-01101). Access to MIMIC-IV and SICdb was provided through PhysioNet [23] after a credential process including training and signing a data use agreement.

Consent for publication

Not applicable.

Competing interests

The study is part of a research project funded by the Swiss Innovation Agency (Innosuisse) in collaboration with Hamilton Medical AG (Switzerland). The project uses high resolution data to improve early recognition of ARDS and studies the impact of ventilator settings on gas exchange. The authors declare that they received funding solely from Innosuisse and did not receive any financial support from Hamilton Medical AG. Hamilton Medical AG had no role in the conduction of the study or writing of the report.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Erlebach, R., Pale, U., Beck, T. et al. Limitations of SpO2 / FiO2-ratio for classification and monitoring of acute respiratory distress syndrome—an observational cohort study. Crit Care 29, 82 (2025). https://doi.org/10.1186/s13054-025-05317-7

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s13054-025-05317-7

Keywords