Predictive value of perfusion index (PI) on the discrepancy between oxygen saturation measurement via pulse oximetry and blood gas analysis

Introduction

Pulse oximetry (SpO₂)

SpO₂ is a pivotal component of the monitoring and clinical management of patients within intensive care units (ICUs) (1). The Food and Drug Administration (FDA) mandates that SpO₂ accuracy should be within 2–3% of arterial oxygen saturation (SaO₂) (2). Manufacturers have published data based on healthy volunteers that demonstrate this standard (3). Factors influencing SpO₂ accuracy include the probe manufacturer, peripheral perfusion, temperature, hypoxemia, hypercapnia, ethnicity, nail decoration, movement artefact, and the site of the pulse oximeter probe (4-8). Accuracy and accessibility of SpO₂ continue to improve (9,10). It has been demonstrated that in critically ill patients SpO₂ accuracy can be worse than in other clinical settings (11). Particularly inaccuracies at physiological extremes have been highlighted in guidelines (12). Differences in skin pigmentation have been found to affect the amount of oxygen administered to patients ICU, this was attributed to variations in pulse oximeter performance in patients with darker skin pigments (13). Recent data from two ICUs in New Zealand and Australia did not find a clinically significant bias in paired SpO₂ and SaO₂ readings (14). However, the limit of agreement between SpO₂ and SaO₂ was 4% which in some cases could have a clinical significance (14). Another recent study of intensive care patients found that SpO₂ could adequately rule out hypoxia but did not meet the required FDA standard (15). The evidence on bias in SpO₂ readings is heterogeneous, with no clear pattern of under- or over-reading SaO₂ (16). In this context, it is worth investigating if perfusion index (PI) can guide the clinician to which patients are at increased risk of inaccurate SpO₂ readings.

PI

The PI is the ratio of the pulsatile blood flow to the non-pulsatile or static blood in peripheral tissue. Limited data exist on the predictive value of PI for the accuracy of oxygen saturation measurement. Pulse oximeter manufacturers assessed accuracy in ten healthy volunteers in low perfusion conditions, defining low perfusion as a PI of 0.02–0.1 (17). In this study, accuracy of SpO₂ in low perfusion states was 2.7 compared to 1.9 in normal perfusion, which was defined as all patients with PI >0.1 (17). Investigations in animal models of severe sepsis did not find PI as a useful marker for increased risk bias (18).

It is evident that inaccurate SpO₂ monitoring could have a clinically significant effect. A significant proportion of patients admitted to intensive care present with pathology that can result in low perfusion states resulting in a low PI. Therefore, further investigating the predictive value of PI index on pulse oximeter accuracy is warranted.

Aim

To assess whether a low PI increased the chance of a clinically significant discrepancy between oxygen saturations measured by blood gas analysis and SpO₂. We present this article in accordance with the STROBE reporting checklist (available at https://jeccm.amegroups.com/article/view/10.21037/jeccm-24-13/rc).

Methods

Setting

A retrospective cohort study was carried out in the ICU of the Hawkes Bay Fallen Soldiers Memorial Hospital in Hastings, New Zealand. This is a district general hospital with an ICU that has 11 beds and approximately 1,000 ICU and High Dependency Unit (HDU) admissions a year. The unit can provide mechanical ventilation and continuous renal replacement therapy, it manages adult and paediatric patients with medical and surgical presentations.

Data collection

A single investigator collected retrospective observational data. Data was collected over a period from May 2021 to December 2021. Patients were selected at random for data collection. All patients over 18 years old with an arterial line and continuous finger SpO₂ monitoring were eligible. The Hawkes Bay District Health Board (DHB) audit registration committee granted approval for the audit. Time of arterial gas sampling to the minute was recorded from the Clinical Portal. Arterial gas sampling was carried out using an ABL 800 FLEX blood gas analyser (Radiometer Medical ApS, Bronshoj, Denmark), which measures SaO₂ directly by co-oximetry. The SpO₂ and PI index were measured via a pulse oximeter using a Mindray 512E reusable sensor, finger probe. The data was recorded at time accurate to the minute and this time correlated to the time of arterial blood gas sampling. The investigator was blinded, by collecting the data points separately, SaO₂ values and the time to the minute were recorded with a patient identification number. Subsequently, at a separate time, the SpO₂ value was retrospectively taken from the monitor with the investigator unaware of the corresponding SaO₂ value in each patient.

Statistical analysis

A visual inspection of the data was performed using a scatter plot of the difference in SpO₂ and SaO₂ and PI. Based on this, the continuous PI variable was transformed into a categorical variable with three groups with divisions placed at points of greatest change in variation between the method discrepancies. These groups were a PI <0.6, PI 0.6–3, and PI >3. A Bland-Altman analysis was performed to calculate bias, limits of agreement and inspect for changes across the measurement range. Statistical analysis was performed using R version 4.3.2. Subsequent statistical analysis was carried out to assess the chance of a clinically significant discrepancy between SpO₂ and SaO₂. A clinically significant discrepancy between SpO₂ and SaO₂ was defined as ±5%. Excel was used to calculate the absolute difference between SpO₂ and SaO₂ and to identify data points with significant discrepancies. The chance of a significant outlier was calculated in each PI range as outliers/total data points in range × 100. The chance of a significant discrepancy was analysed in each of the three PI ranges. The odds ratio (OR) of a significant discrepancy and confidence intervals (CIs) of this analysis were carried out using Medcalc software.

Ethical considerations

The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013). The study was approved by the Hawkes Bay DHB audit registration committee and individual consent for this retrospective analysis was waived.

Results

Three hundred and thirty data points from 80 patients were recorded for inclusion. The average number of data points per patient was 4, with a maximum of 17. Patients with incomplete monitor data were excluded. The scatterplot, Bland-Altman analysis are displayed in Figures 1,2.

Figure 1 A scatterplot of the difference between SpO₂ and SaO₂. SpO₂, pulse oximetry; SaO₂, arterial oxygen saturation.

Figure 2 A Bland-Altman analysis of each of the PI measurement ranges. PI, perfusion index.

When the complete data set is analysed looking at the SpO₂ and SaO₂ the precision is 2.3, the accuracy is 2.36 and the bias is 0.33 (Table 1).

When the same analysis is carried out on readings in the lowest PI range of <0.6 the precision is 2.9, accuracy is 2.9 and bias is 0.6 (Table 1).

A clinically significant discrepancy between SpO₂ and SaO₂ was defined as ±5%. The data was analysed in grouped ranges depending on their PI index. These results are presented in Table 2.

The chance of a clinically significant discrepancy was 13.2% in the range <0.6. In the other ranges, the chance was approximately 3.8–5.1%.

OR of a clinically significant discrepancy between SpO₂ and SaO₂ in the PI range <0.6 compared to all other data points was 3.36, 95% CI: 1.25 to 8.98, P=0.02.

OR of a clinically significant hypoxia in the PI range <0.6 compared to all other PI ranges was 5.57, 95% CI: 1.34 to 23.02, P=0.02.

Discussion

The performance of the pulse oximeter in all PI ranges in terms of accuracy, precision and bias met manufacturer standard. When the data was analysed on a scatterplot a PI of <0.6 appeared to be a point of increased variation between SpO₂ and SaO₂ values. The Bland-Altman analysis shows a clear trend of decreased SpO₂ at low PI values <0.6 with increasing the limits of agreement as PI decreases. There is no suggestion that hypoxia decreases the accuracy of the SpO₂.

Subsequent analysis of the chance of discrepancy was performed to make the data more clinically applicable. The chance of a clinically significant difference between SpO₂ and SaO₂ reached significance (P=0.02) in the PI range <0.6. This data suggests clinicians should be wary of interpreting SpO₂ data when the PI <0.6. Given the patients admitted to ICU are potentially likely to have pathology causing low PI this is significant. In patients with a persistently low PI increased frequency of arterial blood gas analysis may be indicated.

Pulse oximeters function by measuring the absorption of infrared light by oxygenated and deoxygenated haemoglobin. Pulse oximeters measure pulsatile blood flow therefore in low perfusion states this is a potential source of error. Our results suggest this could be the case when there is a low PI value. There is a signal in the data that the SpO₂ tends to overestimate oxygen saturations at low PI (P=0.02). However, the numbers in this analysis are low and should be interpreted with caution.

Manufacturers have previously defined a low perfusion stat as a PI of 0.02–0.1 but this data suggests this should be applied to a PI of 0.6 (3).

The main limitation of this study is the narrow scope of the data that was collected for analysis. Demographic data, potential sources of SpO₂ analysis (e.g., skin pigmentation) and cofounders such as carboxyhaemoglobin are not recorded. In addition, there is no data regarding the potential cause of low PI values, such as hypotension. This prevents assessment of the factors that might be associated SpO₂ inaccuracy or low PI values. The focus of this study is entirely on the relationship between PI and the accuracy of SpO₂ compared to SaO₂. Furthermore, the nature of the PI groupings is suboptimal, however the data was analysed in this fashion to allow focus on the low PI patients where the risk of discrepancy was theorised to be greatest. Additionally, this is a single-centre study, and more data, especially in the low PI group, would have been desirable.

These results suggest further detailed study is necessary in this study population. Specifically, it would be valuable to assess the risk of overestimation of oxygen saturations by SpO₂ in patients with low PI.

Conclusions

Patients with a low PI have a significantly higher chance of having a clinically significant discrepancy between SpO₂ and SaO₂. Care should be taken to ensure these patients are not hypoxic due to monitoring inaccuracies.

Acknowledgments

We would like to thank Dr. Grant Cave for his support in project design and conception and Dr. Jennifer Stewart for her support in editing the manuscript.

Funding: None.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://jeccm.amegroups.com/article/view/10.21037/jeccm-24-13/rc

Data Sharing Statement: Available at https://jeccm.amegroups.com/article/view/10.21037/jeccm-24-13/dss

Peer Review File: Available at https://jeccm.amegroups.com/article/view/10.21037/jeccm-24-13/prf

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jeccm.amegroups.com/article/view/10.21037/jeccm-24-13/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013). The study was approved by the Hawkes Bay DHB audit registration committee and individual consent for this retrospective analysis was waived.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Jubran A. Pulse oximetry. Crit Care 2015;19:272. [Crossref] [PubMed]
2. Food and Drug Administration. FDA Safety Warning. Accessed August 24, 2022. Available online: https://www.fda.gov/news-events/fda-brief/fda-brief-fda-warns-about-limitations-and-accuracy-pulse-oximeters
3. Assessing the Accuracy of Pulse Oximetry in True Clinical Settings. [Cited 2022 Jul 27]. Available online: https://www.semanticscholar.org/paper/Assessing-the-Accuracy-of-Pulse-Oximetry-in-True/1def9974918a1bd7ed14607d43aa73dd1dd74cd2
4. Louie A, Feiner JR, Bickler PE, et al. Four Types of Pulse Oximeters Accurately Detect Hypoxia during Low Perfusion and Motion. Anesthesiology 2018;128:520-30. [Crossref] [PubMed]
5. Katayama S, Shima J, Tonai K, et al. Accuracy of two pulse-oximetry measurements for INTELLiVENT-ASV in mechanically ventilated patients: a prospective observational study. Sci Rep 2021;11:9001. [Crossref] [PubMed]
6. Hassan AT, Ahmed SM, AbdelHaffeez AS, et al. Accuracy and precision of pulse oximeter at different sensor locations in patients with heart failure. Multidiscip Respir Med 2021;16:742. [Crossref] [PubMed]
7. Pilcher J, Ploen L, McKinstry S, et al. A multicentre prospective observational study comparing arterial blood gas values to those obtained by pulse oximeters used in adult patients attending Australian and New Zealand hospitals. BMC Pulm Med 2020;20:7. [Crossref] [PubMed]
8. Luks AM, Swenson ER. Pulse Oximetry for Monitoring Patients with COVID-19 at Home. Potential Pitfalls and Practical Guidance. Ann Am Thorac Soc 2020;17:1040-6. [Crossref] [PubMed]
9. Cannesson M, Talke P. Recent advances in pulse oximetry. F1000 Med Rep 2009;1:66. [Crossref] [PubMed]
10. Nemomssa HD, Raj H. Development of Low-Cost and Portable Pulse Oximeter Device with Improved Accuracy and Accessibility. Med Devices (Auckl) 2022;15:121-9. [Crossref] [PubMed]
11. Van de Louw A, Cracco C, Cerf C, et al. Accuracy of pulse oximetry in the intensive care unit. Intensive Care Med 2001;27:1606-13. [Crossref] [PubMed]
12. Pretto JJ, Roebuck T, Beckert L, et al. Clinical use of pulse oximetry: official guidelines from the Thoracic Society of Australia and New Zealand. Respirology 2014;19:38-46. [Crossref] [PubMed]
13. Gottlieb ER, Ziegler J, Morley K, et al. Assessment of Racial and Ethnic Differences in Oxygen Supplementation Among Patients in the Intensive Care Unit. JAMA Intern Med 2022;182:849-58. [Crossref] [PubMed]
14. Ebmeier SJ, Barker M, Bacon M, et al. A two centre observational study of simultaneous pulse oximetry and arterial oxygen saturation recordings in intensive care unit patients. Anaesth Intensive Care 2018;46:297-303. [Crossref] [PubMed]
15. Harskamp RE, Bekker L, Himmelreich JCL, et al. Performance of popular pulse oximeters compared with simultaneous arterial oxygen saturation or clinical-grade pulse oximetry: a cross-sectional validation study in intensive care patients. BMJ Open Respir Res 2021;8:e000939. [Crossref] [PubMed]
16. Singh AK, Sahi MS, Mahawar B, et al. Comparative Evaluation of Accuracy of Pulse Oximeters and Factors Affecting Their Performance in a Tertiary Intensive Care Unit. J Clin Diagn Res 2017;11:OC05-8. [PubMed]
17. Hyle M. An accuracy study of TruSignal SpO2 technology during challenging patient conditions. GE Healthcare. Accessed September 3, 2022. Available online: https://clinicalview.gehealthcare.com/white-paper/accuracy-trusignal-spo2-technology
18. Hummler HD, Engelmann A, Pohlandt F, et al. Decreased accuracy of pulse oximetry measurements during low perfusion caused by sepsis: Is the perfusion index of any value? Intensive Care Med 2006;32:1428-31. [Crossref] [PubMed]