Triaging acute chest syndrome clinical decision-making using bedside SaO₂/FiO₂ ratio Open Access

Acute chest syndrome (ACS) is a leading cause of death among adult patients with sickle cell disease (SCD).^1-5 Episodes of ACS are marked by pain, shortness of breath, oxygen supplementation, prolonged hospitalizations, readmissions, and end-organ injury from vaso-occlusive disease.^3,6-8 Although episodes of ACS may be initiated by infection, fat embolism, inflammation, infarction, or hypo-ventilation, the exact pathophysiology of ACS is incompletely understood.^2,9,10 Standard treatment is supportive, including supplemental oxygen, pain medication, empiric antibiotics, and either simple or exchange transfusions.^2,8,9 

Accurately predicting the clinical trajectory of ACS remains problematic.^2,5 Whereas many patients present with mild symptoms and improve, some experience clinical deterioration very rapidly. Transfusion is a mainstay of ACS treatment. However, clinical practices vary regarding the decision to transfuse and the criteria for simple vs exchange transfusions.^2,11-13 Expert guidelines recommend that the severity of ACS should determine when to treat with simple vs exchange transfusions.^10,14 ACS severity is inconsistently defined, whether by the number of opacified lobes on chest radiography, increased work of breathing, hypoxemia despite supplemental oxygen, or the presence of pleural effusions.^7,8 We hypothesized that peripheral oxygen saturation would be a reliable and objective measure of respiratory impairment and thus could better predict ICU transfer as a surrogate marker of respiratory failure in patients with ACS. Peripheral oxygen saturation is routinely measured throughout hospitalizations for ACS, and hypoxemia itself worsens ACS pathophysiology, creating a vicious cycle.^9,15 

The ratio of arterial oxygen saturation (SaO₂), approximated by pulse oximetry, to the fraction of inspired oxygen (FiO₂), denoted as the SaO₂/FiO₂ ratio, has been used in emergency medicine and critical care settings for risk stratification and mortality prediction in patients with lung conditions such as acute respiratory distress syndrome (ARDS), pulmonary fibrosis, and COVID-19 pneumonia.^16-28 The SaO₂/FiO₂ ratio approximates the arterial partial pressure of oxygen (PaO₂) to FiO₂ ratio but is noninvasive and easily measured at the bedside.^{21,22,27,29-37} For patients in the intensive care unit (ICU), the SaO₂/FiO₂ ratio was a better predictor of mortality than the PaO₂/FiO₂ ratio,¹⁶ and the SaO₂/FiO₂ ratio has already been incorporated into clinical practice with equations to convert between SaO₂/FiO₂ and PaO₂/FiO₂ ratios.^29,38 Thus, the SaO₂/FiO₂ ratio is included in the new global definition of ARDS.³⁹ We retrospectively calculated the SaO₂/FiO₂ ratio from ACS hospitalization clinical data to assess its association with ICU transfer.

This retrospective observational study included all hospitalizations of adult patients coded with a primary or hospital-acquired diagnosis of ACS at the University of Chicago from 1 January 2017 to 31 December 2021. Hospitalizations were included if patients were aged ≥18 years at the time of admission and diagnosed with ACS by having at least 1 symptom (fever, chest pain, hypoxemia, cough, wheezing, or shortness of breath) and a new opacity on chest radiography, consistent with multiple expert guidelines.^7,8,10 

After identifying relevant hospitalizations, all data were collected via manual chart review and entered into a Research Electronic Data Capture database.^40,41 Data included baseline patient demographics (sex, age, and race/ethnicity), past medical history (body mass index [BMI], smoking history, SCD type, SCD complications, and pulmonary comorbidities), and home oxygen support; oxygen saturation as measured by MX Series pulse oximeters as well as method of delivery and amount of supplemental oxygen (rate or FiO₂) from emergency department (ED) presentation to 96 hours after ACS diagnosis; hospitalization laboratory values, antibiotic administrations, and transfusion requirements; and clinical outcomes of ICU transfer, hospital length of stay (LOS), readmission within 28 days of discharge, and inpatient mortality. SCD type was determined using standard laboratory testing (supplemental Table 1). A composite end point of ICU transfer or death was calculated but was equivalent to ICU transfer because all patients who died in the hospital were transferred to the ICU before passing away. Thus, this composite end point is hereafter referred to as ICU transfer.

For patients on high-flow nasal cannula, bilevel positive airway pressure, or mechanical ventilation, the FiO₂ was obtained from the electronic medical record flow sheet. Missing data were imputed by carrying forward the most recent value.

To evaluate the SaO₂/FiO₂ ratio in the context of ACS, we took several statistical approaches because there are multiple published definitions of ACS severity.^7,42 These approaches included assessing the number of lobes with opacities, the type of transfusions used, and whether the patient was transferred to the ICU. Through manual abstraction and visual inspection, we assessed the number of lobes with opacities on chest radiography at ACS diagnosis.⁷ We also categorized the type of transfusion administered during the hospitalization as none, simple only, or red cell exchange (RCE).⁷ Transfusions were performed per clinicians’ judgment. Finally, we characterized whether the hospitalization included transfer to the ICU. We examined the association between the SaO₂/FiO₂ ratio and each of these 3 proxies for ACS severity: opacified lobes on chest radiography, use of RCE, and ICU transfer. Three key time points were evaluated in the patient’s clinical course to determine the utility of triaging a hospitalization based on the SaO₂/FiO₂ ratio: (1) presentation in the ED; (2) initiation of antibiotics for ACS; and (3) chest radiography confirmation of ACS diagnosis. Initiation of antibiotics may have preceded or followed ACS diagnosis.

Baseline patient demographics and clinical characteristics were explored by ACS hospitalization using descriptive statistics. BMI was categorized as underweight, normal, or overweight/obese, and normal BMI served as a reference for binary comparisons. Smoking history was dichotomized as never vs prior/current. SCD type was dichotomized as hemoglobin SS vs non-SS. Associations between demographic or clinical variables and ICU transfer were tested using multilevel logistic regression models, which accounted for clustering of hospitalizations at the patient level.

Subsequently, we compared the median SaO₂/FiO₂ ratio at the time of ACS diagnosis by key medical history characteristics (BMI, smoking history, asthma, and chronic obstructive pulmonary disorder), the number of opacified lobes at diagnosis on chest radiography, and the clinical decision of RCE or not. We calculated median SaO₂/FiO₂ ratios rather than mean, given the nonnormal distribution. After log-transforming SaO₂/FiO₂ ratios at diagnosis, we constructed univariate multilevel linear regression models to account for clustering of hospitalizations by patient to determine significant associations between key medical and clinical characteristics and the diagnostic SaO₂/FiO₂ ratio. After aggregate analyses across all hospitalizations, we stratified hospitalizations by home oxygen support given the impact of supplemental oxygen on the SaO₂/FiO₂ ratio. We also compared key clinical outcomes of ICU transfer and inpatient mortality by the same key medical history characteristics and methods of ACS severity characterization using univariate multilevel logistic regression models to account for clustering of hospitalizations by patient.

Finally, we used the SaO₂/FiO₂ ratio at each of the 3 key time points (ED presentation, antibiotic initiation, and ACS diagnosis) to determine SaO₂/FiO₂ ratio cutoffs for predicting ICU transfer using the Youden Index.⁴³ Based on the time point with the strongest relationship to ICU transfer, we selected an SaO₂/FiO₂ ratio cutoff for patients overall as well as cutoffs after stratifying those with and without home oxygen support. We then calculated the sensitivity, specificity, and areas under the receiver-operating characteristic curves for these SaO₂/FiO₂ ratio cutoffs. We compared clinical outcomes of transfusion type, ICU transfer, LOS, inpatient mortality, and 28-day readmission by these cutoffs using univariate multilevel logistic and linear regression models to account for clustering of hospitalizations by patient. We also assessed relationships between ICU transfer and these SaO₂/FiO₂ ratio cutoffs using univariate and multivariable multilevel logistic regression models. Models were clustered by patient and adjusted for demographics, clinical, and laboratory characteristics that were either significantly associated with ICU transfer or deemed clinically relevant. Thus, the final multivariable model included BMI category, smoking history, asthma history, number of opacified lobes, hemoglobin concentration, and platelet count at diagnosis. We constructed Kaplan-Meier curves and multivariable mixed parametric survival models to compare hospitalizations above and below each of these cutoffs to determine the hazard ratio for ICU transfer. Statistical analyses were done using Stata 16.1 (College Station, TX).

The study was reviewed and approved by the University of Chicago Institutional Review Board.

There were 252 hospitalizations (136 patients) for ACS initially identified using primary and hospital-acquired diagnoses from admissions between 1 January 2017 and 31 December 2021. Twenty-five hospitalizations (10%) were excluded for not meeting the diagnostic criteria for ACS (ie, no new opacity on chest radiography). Thus, 227 hospitalizations (128 patients) were analyzed; 54% were female, the mean age at admission was 28.9 years (standard deviation, 9.7), and 70% had hemoglobin (Hb) SS (Table 1). BMI category was significantly associated with SCD type (P = .001), with higher proportions of patients with Hb SC and Hb Sβ⁺ having overweight and obese BMIs. Most patients (85%) had a prior ACS hospitalization. Of the hospitalizations in which COVID-19 infection status was assessed (n = 79), only 3 hospitalizations (4%) had current COVID-19 infection, and 4 hospitalizations (5%) had prior COVID infection; only 1 of these 7 hospitalizations (14%) included an ICU transfer. Platelet count at ACS diagnosis was lower for hospitalizations with ICU transfer than those without (284 000/μL vs 369 000/μL; P = .002), but hemoglobin concentration did not significantly differ (7.2 g/dL vs 7.7 g/dL; P = .12). Additionally, patients who were overweight or obese were more likely to be transferred to the ICU than those with normal BMI (36% vs 16%; P = .01).

At ACS diagnosis, chest radiography involved 1 lobe for 114 hospitalizations (50%), 2 lobes for 91 hospitalizations (40%), and ≥3 lobes for 22 hospitalizations (10%). Whereas 45 hospitalizations (20%) had no transfusions, 137 hospitalizations (60%) only included a simple transfusion, and 45 hospitalizations (20%) included RCE. Home oxygen support was noted in 43 hospitalizations (19%). The median diagnostic SaO₂/FiO₂ ratio was lower for hospitalizations of patients with chronic obstructive pulmonary disorder or with home oxygen, as well as for those who had more opacified lobes on diagnostic chest radiography and those who received RCE (Table 2). These associations remained true for those without home oxygen, but only RCE use was associated with lower diagnostic SaO₂/FiO₂ ratio for those with home oxygen. Hospitalizations of those who were overweight or obese or who had ≥3 opacified lobes on diagnostic radiography or received RCE were more likely to include an ICU transfer. Inpatient mortality (n = 4) was not frequent enough to test for significant associations with ACS severity.

SaO₂/FiO₂ ratio cutoffs were determined at each of the 3 key time points. Although the SaO₂/FiO₂ ratio cutoff of 429 at ED presentation was not strongly associated with subsequent ICU transfer (area under the curve [AUC], 0.59; sensitivity, 42%; specificity, 76%), the cutoff of 303 at antibiotic initiation (AUC, 0.74; sensitivity, 60%; specificity, 88%) and of 310 at ACS diagnosis (AUC, 0.73; sensitivity, 63%; specificity, 82%) were strongly associated with subsequent ICU transfer. The time points of antibiotic initiation and ACS diagnosis had roughly equivalent SaO₂/FiO₂ ratios (Figure 1), and antibiotic initiation preceded ACS diagnosis by at least 1 hour for 27% of hospitalizations. Thus, stratified cutoffs by home oxygen support were calculated based on ACS diagnosis because it had the highest sensitivity: 314 for those without home oxygen support (AUC, 0.69; sensitivity, 54%; specificity, 85%) and 276 for those with home oxygen support (AUC, 0.75; sensitivity, 65%; specificity, 85%).

We compared clinical outcomes by the diagnostic SaO₂/FiO₂ cutoff of 310 and stratified cutoffs of 314 and 276 (Table 3). If the diagnostic SaO₂/FiO₂ ratio was below the cutoff, patients more frequently received RCE. Given how the cutoffs were calculated, patients with a diagnostic SaO₂/FiO₂ ratio below the cutoff were frequently transferred to the ICU. LOS was longer for those below the diagnostic SaO₂/FiO₂ ratio cutoff of 310 but not after stratifying by home oxygen support. The 28-day readmission rate did not differ by diagnostic SaO₂/FiO₂ ratio. Of the 4 patients who died while admitted, all were transferred to the ICU, and none of them died directly from ACS-related respiratory failure that was identifiable through retrospective chart review but instead from fat emboli syndrome, hyperhemolysis, cardiac arrest, and pulmonary intravascular talcosis with pulmonary hemorrhage.

Univariate multilevel logistic regression models indicate that all 3 diagnostic SaO₂/FiO₂ ratio cutoffs are significantly associated with ICU transfer: an odds ratio (OR) of 10.1 (95% confidence interval [CI], 4.00-25.6) if <310; OR 9.97 (95% CI, 3.00-33.1) if <314 and not using home oxygen; and OR 8.08 (95% CI, 1.85-35.4) if <276 and using home oxygen. After adjusting for BMI, smoking status, asthma, opacified lobes, hemoglobin, and platelets at ACS diagnosis, multivariable multilevel logistic regression models indicated that all 3 diagnostic SaO₂/FiO₂ ratio cutoffs were still significantly associated with ICU transfer: adjusted odds ratio (aOR) 6.52 (95% CI, 2.33-18.2) if <310; aOR 10.3 (95% CI, 1.59-66.7) if <314 and not using home oxygen; and aOR 7.24 (95% CI, 1.13-46.4) if <276 and using home oxygen. Within 24 hours of ACS diagnosis, 35% of hospitalizations with a diagnostic SaO₂/FiO₂ ratio <310 transferred to the ICU (Figure 2), which equates to 90% oxygen saturation on 2 L/min nasal cannula support (Figure 3). Multivariable mixed parametric survivor models indicated that all 3 diagnostic SaO₂/FiO₂ ratio cutoffs were significantly associated with increased risk of ICU transfer: adjusted hazard ratio (aHR) 4.06 (95% CI, 2.14-7.70) if <310; aHR 4.59 (95% CI, 1.54-13.7) if <314 and not using home oxygen; and aHR 10.3 (95% CI, 1.98-53.2) if <276 and using home oxygen.

To our knowledge, this is the first study to characterize ACS severity using the SaO₂/FiO₂ ratio. We have shown that the SaO₂/FiO₂ ratio is a powerful index for assessing ACS severity and clinical outcomes in SCD. The SaO₂/FiO₂ ratio at the time of diagnosis was associated with the number of lobes with opacities on chest radiography, the use of RCE transfusion, and ICU transfer. This simple bedside point-of-care measure shows promise as a potential marker of ACS severity and should be further explored prospectively to guide ICU transfer and RCE treatment decisions.

The National Heart, Lung, and Blood Institute guidelines from 2014 strongly recommend RCE transfusions for patients with SCD who have “rapid progression of ACS,” which they clarify as “oxygen saturation below 90% despite supplemental oxygen, increased respiratory distress, progressive pulmonary infiltrates, and/or decline in hemoglobin concentration despite simple transfusion,” although this is based on low-quality evidence.⁸ Dichotomizing ACS severity by transfusion type, in which severe ACS is treated with RCE and nonsevere ACS with simple transfusions, does not consider the center’s abilities to administer RCE or the clinical practices of local physicians.⁴⁴ Instead, we propose using this simple bedside measure to help triage ACS severity given the ubiquitous nature of pulse oximeters across all centers and the objectivity of a calculated ratio. This retrospective study demonstrates the feasibility of obtaining the SaO₂/FiO₂ ratio in adult patients with ACS and the potential clinical utility of this measure in defining oxygenation cutoffs for patient care interventions such as ICU transfer or RCE.

We propose using the SaO₂/FiO₂ ratio to aid in clinical decision-making for patients with SCD hospitalized for ACS. A general approach could be used for all patients, or they could be stratified by whether they use home oxygen support. If a general approach is used, our data suggest considering ICU transfer for patients with a SaO₂/FiO₂ ratio cutoff of <310. This would include any patient on at least 3 L/min nasal cannula support at ACS diagnosis or those with an oxygen saturation <90% on 2 L/min. If a stratified approach is used based on home oxygen support, our data suggest an SaO₂/FiO₂ ratio cutoff of <314 for those without home oxygen and <276 for those with home oxygen. This would include patients without home oxygen who are on at least 3 L/min nasal cannula support or those with an oxygen saturation <91% on 2 L/min, as well as those with home oxygen support on at least 4 L/min or with an oxygen saturation <91% on 3 L/min nasal. The SaO₂/FiO₂ ratio cutoffs corresponds to PaO₂/FiO₂ ratios of 293, 298, and 252, respectively, based on existing literature.³⁰ These are all within the range of PaO₂/FiO₂ ratios of 200 to 300 defining mild ARDS.⁴⁵ Although transfer to the ICU mostly occurred within the first 36 hours after ACS diagnosis for the first 2 diagnostic SaO₂/FiO₂ ratio cutoffs (<310 for all and <314 for those without home oxygen), half of patients with home oxygen support who had a diagnostic SaO₂/FiO₂ ratio <276 transferred to the ICU within 12 hours of ACS diagnosis. Additionally, hospitalizations that included an ICU transfer were more common for patients who were overweight or obese, which could relate generally to the roles of social determinants of health in patients with SCD or may be driven by restrictive physiology related to obesity.^46-48 Further work is needed to better describe the impact of BMI on ACS, and future prospective studies could consider stratifying randomization by BMI.

The main limitation of this study is its retrospective study design at a single center. Although a total of 128 patients across 227 hospitalizations were included in this analysis, the results may be limited in its generalizability to other hospitals given the variation in clinical practice patterns, including when to initiate RCE.¹³ The 227 hospitalizations were distributed across the 3 hospitals at the University of Chicago campus: Center for Care and Discovery (38%), Mitchell Hospital (38%), and Comer Children’s Hospital (24%). Additionally, there were missing data in this retrospective chart review, which required the imputation of FiO₂ for some of the SaO₂/FiO₂ ratios. Because of the retrospective nature, the SpO₂ measurements could not be benchmarked to PaO₂ in all study patients. We may also have missed ACS hospitalizations during the time period that were not coded with either primary or hospital-acquired ACS diagnoses. Furthermore, we were not able to fully assess for the presence of multisystem organ failure as described by Chaturvedi et al, given the incomplete documentation of all relevant clinical data.⁴⁹ We plan to prospectively evaluate for this and for trends in hemoglobin and platelets over the ACS hospitalization in future studies. Furthermore, the total number of packed red blood cell units given by simple transfusion during hospitalization was not recorded. Ideally, categorization would have followed Ballas et al with up to 2 units of simple transfusion being mild ACS, 3 units simple transfusion being moderate, and RCE being severe,⁷ but we did not have this level of detail. Finally, the clinical decisions to perform an RCE or to transfer to the ICU in this retrospective study had some degree of subjectivity and depended on the clinical judgment of the physicians treating that patient.

ACS is a complex syndrome that encompasses several disorders that result in a similar clinical phenotype, including infection, fat emboli, pulmonary emboli, and vaso-occlusive sickling. There are variable clinical phenotypes that are not yet clearly delineated, spanning from patients who recover completely with limited intervention to those who decompensate very quickly. In the future, the diagnosis of ACS may shift toward methods without radiation exposure such as bedside ultrasound or auscultation.^50,51 The SaO₂/FiO₂ ratio warrants further exploration in patients with SCD who present with ACS. It shows promise given how simply it is obtained and how it normalizes the SpO₂ readings to the level of oxygen support. It relies on objective data rather than the more subjective interpretation of chest radiography lobe opacification number or the clinical practice of RCE. The study team partnered to create an easily accessible online SaO₂/FiO₂ ratio calculator to help improve real-time prediction in routine clinical practice.

Despite data demonstrating racial bias in SpO₂ and higher rates of occult hypoxemia in Black patients than White patients,^52,53 as well as concern regarding the accuracy of pulse oximetry when patients with SCD have acute worsening of their anemia given the shift in the oxyhemoglobin dissociation curve,⁵⁴ in our study, the SaO₂/FiO₂ ratio was associated with relevant clinical outcomes and measures of ACS severity in an entirely Black cohort of patients with SCD, further underscoring the robustness of its prognostic value. If escalating nasal cannula support to at least 3 L/min (or at least 4 L/min for those with home oxygen support), clinicians could consider prompt ICU transfer or RCE. Future prospective studies in both academic and community hospitals should evaluate the use of the SaO₂/FiO₂ ratio threshold to guide ICU transfer and RCE recommendations and evaluate its impact on clinical outcomes and patient safety in adults and children with SCD. The SaO₂/FiO₂ ratio is an accessible tool that can be used by clinicians at academic or community hospitals, regardless of their prior experience treating patients with SCD.

Contribution: A.W. designed the research, analyzed data, and wrote the manuscript; M.W. and A.J.W. performed research and reviewed the manuscript; A.A. contributed vital analytical tools and reviewed the manuscript; and M.J.R. and G.L.-C. designed the research and reviewed the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Austin Wesevich, Department of Medicine, The University of Chicago, 5841 S Maryland Ave, MC 2115, Chicago, IL 60637; email: austin.wesevich@uchicagomedicine.org.