Abstract
OBJECTIVES:
Acute hypoxemic respiratory failure (AHRF) is a leading cause of ICU admission and is associated with significant morbidity and mortality. Yet, epidemiological data on AHRF remain limited, and current management strategies largely derive from studies on acute respiratory distress syndrome (ARDS). We examined the prevalence of AHRF among ICU patients receiving supplemental oxygen or respiratory support for at least 4 hours, as well as initial respiratory support strategies and associated outcomes.
DESIGN:
Multicenter registry-based cohort study, 2014–2023.
SETTING:
Nine university of Toronto affiliated ICUs.
Patients:
Adult ICU patients receiving supplemental oxygen or respiratory support who met AHRF criteria, defined as Pao2/Fio2 less than or equal to 300 mm Hg or, if Pao2 was not available, peripheral oxygen saturation (SpO2)/Fio2 less than or equal to 315, within the first 24 hours of ICU admission.
INTERVENTIONS:
None.
MEASUREMENTS AND MAIN RESULTS:
The primary outcome was AHRF prevalence. Secondary outcomes included ICU length of stay, invasive mechanical ventilation (IMV) duration, ventilator-free days, and ICU mortality, stratified by AHRF severity. Of 21,714 patients recorded in the registry, 10,832 (50%) met AHRF criteria, of whom 76% required IMV. Most received lung-protective ventilation. With increasing AHRF severity, ICU length of stay and IMV duration increased, while ventilator-free days and probability of ICU discharge at 30 days decreased. ICU mortality was 23% overall and 47% among patients on IMV with severe AHRF.
CONCLUSIONS:
AHRF is common early after ICU admission among patients requiring supplemental oxygen or respiratory support and is frequently managed using ARDS-based strategies. A pragmatic, standardized, and operationally feasible definition of AHRF may improve patient recognition, enable comparisons across studies, and guide future research.
Keywords: artificial respiration, hypoxemia, intensive care units, respiratory distress syndrome, respiratory insufficiency
KEY POINTS.
Question: What is the prevalence of acute hypoxemic respiratory failure (AHRF) within the first 24 hours of ICU admission among patients receiving supplemental oxygen or respiratory support, and what are the associated respiratory support strategies and outcomes?
Findings: In this multicenter registry-based cohort study, 50% of ICU patients developed AHRF within the first 24 hours, 76% required invasive mechanical ventilation and were frequently managed using acute respiratory distress syndrome-based strategies, and mortality increased with severity, reaching 47% in severe cases.
Meaning: AHRF is highly prevalent among ICU patients requiring supplemental oxygen or respiratory support and requires a standardized, pragmatic definition to improve recognition, enable study comparisons, and guide future research.
Patients with acute hypoxemic respiratory failure (AHRF) frequently require respiratory support, and experience significant morbidity and mortality (1, 2). AHRF is often defined by new-onset oxygenation failure with a Pao2/Fio2 less than or equal to 300 mm Hg (3). It affects 15–18% of ICU admissions and up to 35% of patients receiving invasive mechanical ventilation (IMV) (1, 2, 4, 5). The estimated ICU prevalence of AHRF is around 50%, with most patients presenting with mild hypoxemia (Pao2/Fio2 201–300 mm Hg) (3, 6). Data suggest that up to 40% of AHRF patients are managed with high-flow nasal cannula (HFNC) or noninvasive ventilation (NIV), even in severe hypoxemia (Pao2/Fio2 ≤ 100 mm Hg) (3). ICU mortality averages 27%, and increases with AHRF severity, while in-hospital mortality averages 35% (2, 3, 7).
Management strategies for AHRF are largely derived from acute respiratory distress syndrome (ARDS) research, although findings may not be generalizable to the broader AHRF population (8, 9). ARDS remains under-recognized in up to 40% of eligible patients (10), mainly due to: 1) interobserver variability in identifying bilateral opacities on chest imaging (11); 2) reliance on Pao2 from arterial blood gas (ABG); and 3) the PEEP threshold in the Berlin definition excludes otherwise similar patients receiving HFNC or standard oxygen therapy, although recent global and European Society of Intensive Care Medicine (ESICM) guidelines recognize nonintubated ARDS on HFNC and allow for peripheral oxygen saturation to Fio2 (Spo2/Fio2) ratio when ABG is unavailable (8, 9).
In contrast, AHRF is more inclusive and can be identified independently of oxygen delivery mode or ABG availability. Although several multicenter cohorts have described acute respiratory failure using Pao2/Fio2-based definitions (2, 3, 6, 12), few have operationalized AHRF using the Spo2/Fio2 ratio across contemporary respiratory support strategies. The Spo2/Fio2 ratio has been proposed as a reliable, noninvasive alternative; however, clinical data supporting its use remain limited (13, 14). As a result, AHRF epidemiology remains heterogeneous due to the absence of a consensus definition, and few cohorts have characterized this broader population across contemporary modes of respiratory support (2, 5, 15).
In this large registry-based cohort across nine ICUs, we quantify early AHRF prevalence and describe respiratory support strategies and outcomes. We applied a pragmatic oxygenation-based definition within 24 hours of ICU admission, using Pao2/Fio2 or Spo2/Fio2 independent of the mode of respiratory support. This approach reflects contemporary practice across both nonintubated and intubated patients. Leveraging a detailed multicenter database spanning academic and community ICUs, we deliver a large-scale characterization of AHRF in routine ICU practice to inform clinical care and guide future research.
MATERIALS AND METHODS
Data Source and Study Population
This multicenter, registry-based cohort study used data from the Toronto Intensive Care Observational Registry (iCORE), which included all adult patients (≥ 18 yr) admitted between April 2014 and January 2023 to one of nine University of Toronto-affiliated ICUs who received supplemental oxygen via low-flow nasal cannula or any form of respiratory support, classified as either noninvasive (HFNC, continuous positive airway pressure [CPAP], or NIV) or IMV, for at least 4 hours. These ICUs are part of academic and community hospitals across the Greater Toronto Area. Data were prospectively collected once daily at 8 am and included clinical variables, ventilation parameters, blood gas, and laboratory measurements. IMV was defined as positive pressure ventilation delivered via endotracheal tube (oral, nasal) or tracheostomy, and NIV as positive pressure ventilation delivered via face mask, nasal mask, or helmet.
We included all patients with AHRF, defined as Pao2/Fio2 less than or equal to 300 mm Hg or, if Pao2 was not available, an Spo2/Fio2 less than or equal to 315, within the first 24 hours of ICU admission (3, 4). Patients with a Pao2/Fio2 greater than 300 were excluded regardless of Spo2/Fio2 values. For patients with multiple ICU admissions during the same hospital stay, only data from the initial admission were included. This project was approved by the Research Ethics Board from the University Health Network (Protocol N°. 12-5747.24, Study title “Toronto Intensive Care Registry [iCORE] Project,” approval date: April 3, 2014), and was conducted in accordance with the ethical standards of the institutional committee and the Declaration of Helsinki. Informed consent was waived given the retrospective design and data were de-identified.
Severity of AHRF and Clinical Measurements
AHRF severity was categorized using the same Pao2/Fio2 thresholds as the Berlin definition for ARDS: mild (201–300 mm Hg), moderate (101–200 mm Hg), and severe (≤ 100 mm Hg) (16). When Pao2/Fio2 ratios were unavailable, severity was assigned using Spo2/Fio2 thresholds: mild (236–315), moderate (149–235), and severe (≤ 148) (17). This hierarchical classification reflects clinical practice, where Pao2-based categorization is prioritized when available. Fio2 was recorded directly from the device for HFNC, CPAP, NIV, and IMV. For low-flow supplemental oxygen via nasal cannula, Fio2 was estimated using conventional bedside conversion: 1 L/min equals 0.24, +0.04 per additional L/min (18). For patients receiving IMV, we described ventilation parameters, static respiratory system elastance (ERS), and use of adjunctive therapies within the first 30 days of their ICU stay, stratified by AHRF severity and ERS normalized to predicted body weight (PBW), with low normalized ERS defined as less than 2.5 and high as greater than or equal to 2.5 cm H2O/(mL/kg PBW) (19). When recorded, driving pressure (∆P) was calculated as airway plateau pressure (Pplat) minus total PEEP. Mechanical power was calculated as respiratory rate × tidal volume (Vt) × (peak inspiratory airway pressure–[∆P]/2) × 0.098) (20). Ventilatory ratio, a surrogate of physiologic dead space fraction, was calculated as (minute ventilation × Paco2)/(PBW × 100 × 37.5) (21).
Outcomes
The primary outcome was AHRF prevalence within 24 hours of ICU admission. This window was selected because iCORE data are recorded once daily at a standardized time point, allowing consistent ascertainment of early hypoxemia at ICU presentation. Secondary outcomes were focused on patients receiving IMV and included ventilation parameters, interventions and adjunctive therapies up to day 30, ICU length of stay (LOS), duration of IMV, ICU mortality, and ventilator-free days (VFDs). Patients were followed from the time they met AHRF criteria within 24 hours of admission until ICU discharge or death in the ICU, whichever occurred first. Similar to previous work, VFDs through day 28 were defined for patients intubated for more than 48 hours and calculated as 28 minus the number of calendar days intubated (22, 23). For patients intubated longer than 28 days, VFD was set to 0; for patients who died within 90 days after IMV initiation, VFD was set to –1 (22).
Statistical Analysis
Descriptive statistics were used to characterize the patient population. Continuous variables were summarized as median and interquartile range (IQR), and categorical variables as counts or percentages (%). The 30-day cumulative probability of death was estimated using a competing risk regression, where ICU discharge was considered a competing event precluding patients from dying in the ICU.
Given the descriptive nature of the study, we conducted a complete case analysis without making any assumptions about missing data. A two-sided p value of less than 0.05 was considered statistically significant. Analyses were performed using R, Version 4.5.0 (R Foundation for Statistical Computing, Vienna, Austria) and reported according to the Strengthening the Reporting of Observational Studies in Epidemiology guidelines (https://links.lww.com/CCM/I35).
Sensitivity and Subgroup Analyses
To examine whether our results were consistent under more restrictive case definitions and explore potential sources of misclassification, we performed two sensitivity analyses: 1) restricted to patients with Spo2 less than 97%, as values on the flat portion of the oxyhemoglobin dissociation curve correspond to a wide range of Pao2 levels and cannot reliably identify true hypoxemia and 2) restricted to patients on IMV with conditions more likely to be associated with ARDS, including pneumonia, sepsis, trauma, aspiration, postoperative respiratory failure, and acute respiratory failure not attributable to COPD. As a subgroup analysis, we examined patients on IMV with Pao2/Fio2 less than or equal to 150 mm Hg to assess the incidence of prone positioning within the first 30 days, aiming to explore adherence to ARDS guideline-recommended practices in patients with moderate-to-severe hypoxemia.
RESULTS
Prevalence of AHRF and Baseline Characteristics
Among the 21,714 critically ill patients receiving supplemental oxygen or respiratory support, 10,832 (50%) met criteria for AHRF and were included in this study (Fig. 1). Of these, 3831 (35%) had Spo2/Fio2 measurements only, and 34 (< 1%) had recordings of Pao2/Fio2 only. Among patients with paired measurements, severity classification showed good agreement (eTable 1, https://links.lww.com/CCM/I35). Median (IQR) age was 63 years (51–73 yr), and 37% were women (Table 1). The most common comorbidities were cardiovascular disease, respiratory disease, and diabetes. The primary indication for initiating respiratory support were acute respiratory failure from all causes and altered level of consciousness. Patients with a more severe AHRF had higher Acute Physiology and Chronic Health Evaluation III and Sequential Organ Failure Assessment scores.
Figure 1.
Study flowchart. Among the 21,714 patients initially recorded in our registry, 10,832 patients met our criteria for acute hypoxemic respiratory failure (AHRF) and subsequently were included in our analysis. CPAP = continuous positive airway pressure, ECLS = extracorporeal life support, HFNC = high-flow nasal cannula, iCORE = Toronto Intensive Care Observational Registry, IMV = invasive mechanical ventilation, NIV = noninvasive ventilation, Spo2 = peripheral oxygen saturation.
TABLE 1.
Baseline Characteristics of Patients With Acute Hypoxemic Respiratory Failure, Stratified by Severity
| Characteristic | Overall (n = 10,832) | Stratification by Acute Hypoxemic Respiratory Failure Severitya | ||
|---|---|---|---|---|
| Mild (n = 5,466) | Moderate (n = 4,070) | Severe (n = 1,296) | ||
| Baseline characteristics | ||||
| Age, yr | 63 (51–73) | 63 (52–73) | 63 (51–73) | 61 (50–72) |
| Female sex, n | 4,019 (37%) | 1,998 (37%) | 1,520 (37%) | 501 (39%) |
| Body mass index, kg/m2 | 27 (23–32) | 27 (23–32) | 28 (23–33) | 27 (23–33) |
| Acute Physiology and Chronic Health Evaluation III score | 75 (56–93) | 70 (53–89) | 78 (60–96) | 83 (62–108) |
| Sequential Organ Failure Assessment score | 5 (3–8) | 5 (2–7) | 6 (3–9) | 7 (4–10) |
| Baseline comorbidities | ||||
| Cardiovascular disease, n | 4,016 (37%) | 2,125 (39%) | 1,456 (36%) | 435 (34%) |
| Respiratory disease, n | 2,761 (25%) | 1,324 (24%) | 1,134 (28%) | 303 (23%) |
| Diabetes, n | 2,626 (24%) | 1,338 (24%) | 972 (24%) | 316 (24%) |
| Active neoplasm, n | 1,102 (10%) | 485 (8.9%) | 429 (11%) | 188 (15%) |
| Chronic kidney disease, n | 962 (8.9%) | 476 (8.7%) | 366 (9.0%) | 120 (9.3%) |
| Liver failure, n | 729 (6.7%) | 388 (7.1%) | 257 (6.3%) | 84 (6.5%) |
| Immunosuppression, n | 579 (5.3%) | 236 (4.3%) | 222 (5.5%) | 121 (9.3%) |
| Indication for respiratory support | ||||
| Acute respiratory failure, n | 7,808 (72%) | 3,803 (70%) | 3,010 (74%) | 995 (77%) |
| Altered level of consciousness, n | 3,074 (28%) | 1,715 (31%) | 1,065 (26%) | 294 (23%) |
| Acute exacerbation of chronic pulmonary disease, n | 374 (3.5%) | 162 (3.0%) | 178 (4.4%) | 34 (2.6%) |
| Neuromuscular causes, n | 57 (0.5%) | 35 (0.6%) | 15 (0.4%) | 7 (0.5%) |
| Respiratory conditions | ||||
| Pao2/Fio2, mm Hg | 196 (140–248) | 250 (225–275) | 156 (130–180) | 81 (67–91) |
| Peripheral oxygen saturation/Fio2 | 225 (163–250) | 247 (240–282) | 192 (160–213) | 100 (94–122) |
| Noninvasive respiratory support, n | 1,054 (9.7%) | 381 (7.0%) | 442 (11%) | 231 (18%) |
| High-flow nasal cannula, n | 310 (2.9%) | 49 (0.9%) | 138 (3.4%) | 123 (9.5%) |
| Continuous positive airway pressure, n | 343 (3.2%) | 194 (3.5%) | 118 (2.9%) | 31 (2.4%) |
| Noninvasive ventilation, n | 401 (3.7%) | 138 (2.5%) | 186 (4.6%) | 77 (5.9%) |
| Invasive mechanical ventilation, n | 8,280 (76%) | 4,065 (74%) | 3,247 (80%) | 968 (75%) |
Three thousand eight hundred thirty-one patients (35%) had missing Pao2/Fio2 at baseline and were classified using peripheral oxygen saturation (Spo2)/Fio2.
Values are reported as medians (interquartile range) or counts (%). Baseline data were collected at 8 am within the first 24 hr of ICU admission. Acute hypoxemic respiratory failure severity was defined using Pao2/Fio2 thresholds. When unavailable, Spo2/Fio2 thresholds were used.
Ventilation Parameters, Interventions, and Adjunctive Therapies in Patients on IMV
A total of 8280 (76%) patients received IMV within 24 hours, of which 659 (8%) had ARDS documented in the registry (eTable 2, https://links.lww.com/CCM/I35). Among patients with available measurements, most were ventilated within lung-protective limits recommended for ARDS management, with a median (IQR) Vt of 6.5 mL/kg PBW (5.9–7.8 mL/kg PBW; Table 2; Fig. 2; eFig. 1, https://links.lww.com/CCM/I35). Pressure-control was the most frequently used mode (37%), especially in more severe AHRF. Among patients receiving volume-controlled ventilation, Pplat and ∆P were recorded in 38%, of whom 97% had a Pplat less than or equal to 30 cm H2O and 78% had a ∆p less than or equal to 14 cm H2O. ∆P increased with severity, from a median (IQR) of 10 cm H2O (8–13 cm H2O) with ERS less than 2.5–17 (15–19) with ERS greater than or equal to 2.5 cm H2O/(mL/kg PBW).
TABLE 2.
Ventilatory Characteristics of Patients on Invasive Mechanical Ventilation With Acute Hypoxemic Respiratory Failure, Stratified by Severity
| Characteristic | Overall (n = 8280) | Stratification by Acute Hypoxemic Respiratory Failure Severitya | ||
|---|---|---|---|---|
| Mild (n = 4065) | Moderate (n = 3247) | Severe (n = 968) | ||
| Blood gases | ||||
| Arterial pH | 7.35 (7.28–7.40) | 7.37 (7.31–7.41) | 7.34 (7.27–7.40) | 7.28 (7.18–7.36) |
| Paco2, mm Hg | 41 (35–47) | 39 (35–45) | 42 (36–49) | 47 (40–56) |
| Spo2 or arterial oxygen saturation, % | 97 (94–98) | 97 (96–99) | 96 (94–98) | 94 (90–96) |
| Pao2/Fio2, mm Hg | 196 (140–247) | 250 (225–275) | 155 (130–180) | 81 (67–90) |
| Spo2/Fio2 | 200 (161–245) | 245 (237–277) | 190 (158–200) | 100 (93–122) |
| Ventilator parameters | ||||
| Fio2, % | 50 (40–60) | 40 (35–40) | 50 (50–60) | 95 (80–100) |
| Set respiratory rate, breaths/min | 18 (0–24) | 16 (0–22) | 20 (0–26) | 24 (18–30) |
| Tidal volume, mL/kg PBW | 6.5 (5.9–7.8) | 6.6 (5.9–7.9) | 6.5 (5.8–7.7) | 6.3 (5.6–7.4) |
| Set positive end-expiratory pressure, cm H2O | 8 (5–10) | 5 (5–8) | 8 (6–10) | 10 (8–12) |
| Peak inspiratory pressure, cm H2O | 22 (17–28) | 20 (15–24) | 24 (19–30) | 29 (24–34) |
| Plateau pressure, cm H2O | 19 (15–23) | 18 (15–21) | 20 (16–24) | 21 (13–26) |
| Driving pressure, cm H2O | 11 (8–14) | 11 (8–13) | 11 (8–14) | 11 (5–16) |
| Mechanical power, J/min | 20 (15–28) | 17 (13–23) | 22 (16–30) | 26 (17–36) |
| Ventilatory ratio | 1.59 (1.25–2.04) | 1.48 (1.17–1.83) | 1.67 (1.32–2.13) | 2.09 (1.61–2.75) |
| Elastance of the respiratory system, cm H2O/(mL/kg PBW) | 1.72 (1.28–2.24) | 1.67 (1.30–2.10) | 1.79 (1.31–2.30) | 1.75 (0.94–2.35) |
| Pressure-controlled ventilation, n | 3040 (37%) | 1312 (32%) | 1267 (39%) | 461 (48%) |
| Volume-controlled ventilation, n | 2506 (30%) | 1109 (27%) | 1030 (32%) | 367 (38%) |
| Pressure-support ventilation, n | 2633 (32%) | 1610 (40%) | 911 (28%) | 112 (12%) |
| Other ventilation modes, n | 71 (0.9%) | 28 (0.7%) | 30 (0.9%) | 13 (1.3%) |
PBW = predicted body weight, Spo2 = peripheral oxygen saturation.
Two thousand four hundred fifty-five patients (30%) had missing Pao2/Fio2 at baseline and were classified using Spo2/Fio2.
Values are reported as medians (interquartile range) or counts (%). Baseline data were collected at 8 am within the first 24 hr of ICU admission. Acute hypoxemic respiratory failure severity was defined using Pao2/Fio2 thresholds. When unavailable, Spo2/Fio2 thresholds were used.
Figure 2.
Baseline ventilatory parameters in patients receiving invasive mechanical ventilation with acute hypoxemic respiratory failure (AHRF), stratified by severity. Cumulative frequency distribution plots for plateau pressure (Pplat; A) and driving pressure (∆P; B). A total of 933 patients (96%) had a Pplat less than or equal to 30 cm H2O, and 750 (77%) had a ∆P less than or equal to 14 cm H2O. Additionally, the distributions of tidal volume (Vt) vs. Pplat (C) and Vt vs. ∆P (D) are shown. Baseline measurements were collected at 8 am within the first 24 hr of invasive mechanical ventilation. AHRF severity was defined using Pao2/Fio2 thresholds. When unavailable, peripheral oxygen saturation/Fio2 thresholds were used. PBW = predicted body weight.
Among patients on IMV, vasopressors and chest tube placements were the most frequently used interventions, while neuromuscular blockade and extracorporeal life support (ECLS) were the most frequently used adjunctive therapies during the first 30 days of ICU stay (eTable 3, https://links.lww.com/CCM/I35). Use of renal replacement therapy, neuromuscular blockade, prone positioning, inhaled nitric oxide (iNO), and ECLS increased with greater AHRF severity. Among patients with severe AHRF, 26% received neuromuscular blockade, 13% received prone positioning, 12% received iNO, and 12% received ECLS within the first 30 days of ICU admission.
ICU Length of Stay and Duration of IMV
The overall median (IQR) ICU LOS was 5 days (3–11 d), and 6 days (3–13 d) for patients on IMV (Table 3 and eTable 4, https://links.lww.com/CCM/I35). Median (IQR) duration of IMV was 4 days (2–11 d).
TABLE 3.
Outcomes of Patients With Acute Hypoxemic Respiratory Failure, Stratified by Severity
| Outcomes | Overall (n = 10,832) | Stratification by Acute Hypoxemic Respiratory Failure Severitya | p b | ||
|---|---|---|---|---|---|
| Mild (n = 5,466) | Moderate (n = 4,070) | Severe (n = 1,296) | |||
| ICU outcomes | |||||
| ICU length of stay, d | 5 (3–11) | 5 (2–9) | 6 (3–13) | 6 (2–15) | < 0.001 |
| ICU mortality, n | 2,539 (23%) | 902 (17%) | 1,072 (26%) | 565 (44%) | < 0.001 |
| 30-d cumulative probability of events | |||||
| ICU discharge, % | 74 (73–75) | 82 (81–83) | 70 (69–72) | 52 (49–55) | < 0.001 |
| Death, % | 22 (22–23) | 16 (15–17) | 25 (24–26) | 42 (39–44) | < 0.001 |
Three thousand eight hundred thirty-one patients (35%) had missing Pao2/Fio2 at baseline and were classified using peripheral oxygen saturation (Spo2)/Fio2.
Kruskal-Wallis test, Pearson χ2 test, and Gray test.
Values are reported as medians (interquartile range) or counts (%). Cumulative probabilities are shown as percentages (95% CI). Acute hypoxemic respiratory failure severity was defined using Pao2/Fio2 thresholds. When unavailable, Spo2/Fio2 thresholds were used.
Mortality and VFD
Overall ICU mortality was 23% and increased with AHRF severity. Among 1054 patients (9.7%) receiving noninvasive respiratory support within 24 hours of ICU admission, the 30-day cumulative probability of death was 23% (95% CI, 20–25%; Table 4). Mortality increased and ICU discharge declined across worsening severity. Among 8280 patients (76%) receiving IMV within 24 hours of ICU admission, the 30-day cumulative probability of death was 25% (95% CI, 24–26%) and increased with severity (eTable 4, https://links.lww.com/CCM/I35). Patients with more severe AHRF, or with high baseline ERS, had fewer VFD, a lower probability of ICU discharge, and a higher probability of death at 30 days (eTables 4 and 5, https://links.lww.com/CCM/I35).
TABLE 4.
Outcomes of Patients on Noninvasive Respiratory Support With Acute Hypoxemic Respiratory Failure, Stratified by Severity
| Outcomes | Overall (n = 1054) | Stratification by Acute Hypoxemic Respiratory Failure Severitya | p b | ||
|---|---|---|---|---|---|
| Mild (n = 381) | Moderate (n = 442) | Severe (n = 231) | |||
| ICU outcomes | |||||
| ICU length of stay, d | 5 (3–10) | 4 (2–8) | 5 (3–10) | 7 (4–16) | < 0.001 |
| ICU mortality, n | 248 (24%) | 55 (14%) | 113 (26%) | 80 (35%) | < 0.001 |
| 30-d cumulative probability of events | |||||
| ICU discharge, % | 74 (71–77) | 84 (80–87) | 72 (67–76) | 62 (56–68) | < 0.001 |
| Death, % | 23 (20–25) | 14 (11–18) | 25 (21–29) | 32 (27–39) | < 0.001 |
Four hundred eighty-eight patients (46%) had missing Pao2/Fio2 at baseline and were classified using peripheral oxygen saturation (Spo2)/Fio2.
Kruskal-Wallis test, Pearson χ2 test, and Gray test.
Values are reported as medians (interquartile range) or counts (%). Cumulative probabilities are shown as percentages (95% CI). Acute hypoxemic respiratory failure severity was defined using Pao2/Fio2 thresholds. When unavailable, Spo2/Fio2 thresholds were used.
Sensitivity and Subgroup Analyses
Restricting the cohort to patients with Spo2 less than 97% at baseline yielded a more hypoxemic population with less frequent IMV use (eTable 6, https://links.lww.com/CCM/I35). Outcomes were similar, with slightly higher mortality (eTable 7, https://links.lww.com/CCM/I35). Among IMV patients with plausible ARDS etiologies, baseline characteristics, ventilator settings, and outcomes were similar to the overall IMV cohort (eTables 8–10, https://links.lww.com/CCM/I35). In the subgroup of 1740 patients on IMV with Pao2/Fio2 less than 150 mm Hg, ARDS was documented in 282 patients (16%). Prone positioning was used in 183 patients (11%), including 104 (6%) with documented ARDS and increased over time, particularly after 2017 (eFigs. 2 and 3, https://links.lww.com/CCM/I35). Over the study period, respiratory support patterns were broadly similar, with a gradual increase in low-flow nasal cannula and HFNC use (eFig. 4, https://links.lww.com/CCM/I35). ICU LOS and IMV duration showed small variations. Oxygenation ratios were slightly lower after 2019. Vt and ∆P decreased modestly over time, with no major change in PEEP.
DISCUSSION
In this large multicenter cohort of 21,714 adult ICU patients receiving supplemental oxygen or respiratory support for at least 4 hours, half met AHRF criteria within 24 hours of ICU admission. Overall ICU mortality was 23% and increased with AHRF severity. Among patients on IMV, ICU mortality increased with AHRF severity, while VFD and probability of ICU discharge declined. IMV duration and ICU LOS varied modestly by severity. Most patients on IMV received lung-protective ventilation. Use of adjunctive therapies increased with severity, although prone positioning remained infrequent despite greater utilization in later years. Outcomes among patients with Spo2 less than 97% were consistent with the full cohort, supporting our AHRF definition. Patients with plausible ARDS etiologies had similar baseline characteristics, ventilator settings, and outcomes compared with the overall IMV AHRF cohort.
Our findings align with prior studies reporting that around half of ICU patients present with AHRF early after admission (3, 6), and extend these observations to a large cohort spanning the full spectrum of respiratory support. Among patients initially managed with noninvasive respiratory support, oxygenation-based severity categories demonstrated a stepwise gradient in outcomes, supporting the clinical relevance of pragmatic Pao2/Fio2 or Spo2/Fio2-based stratification beyond invasively ventilated ARDS populations. Documented ARDS accounted for 8% of IMV patients with AHRF, compared with 20–25% in prior multinational cohorts (24). This likely reflects misclassification and incomplete ascertainment in routine practice, limited arterial sampling, and the increasing use of noninvasive respiratory support outside the Berlin framework (11, 25, 26). The newly proposed global ARDS definition now acknowledges the role of Spo2 when ABG analysis is not feasible, to help stratify disease severity (27), as well as to include nonintubated ARDS patients. Our findings further highlight that ABG measurements are frequently missing, which may further hinder ARDS recognition in routine practice. In our cohort, 35% of patients lacked Pao2/Fio2 data and were classified using Spo2/Fio2. Patients with AHRF resembled those reported in ARDS cohorts but were less hypoxemic (24). Mortality approached that reported in ARDS with progressive hypoxemia, reaching 47% in severe AHRF. This suggests an overlap between AHRF and ARDS while capturing additional high-risk patients not identified by the Berlin criteria. Among IMV patients with plausible ARDS etiologies, baseline characteristics, ventilator settings, and outcomes were comparable, suggesting etiology did not meaningfully influence management or prognosis. Our study supports the utility of a pragmatic definition of AHRF to identify patients who may benefit from ARDS-based interventions, even in the absence of formal ARDS diagnosis.
The Pao2/Fio2 ratio remains central to ARDS management as a prognostic marker and guide for adjunctive therapies (3, 16). However, ABG sampling is invasive and may yield variable results (13, 28, 29). In our cohort, over one-third of patients lacked Pao2/Fio2 measurements. Pulse oximetry offers a noninvasive, continuous, reliable, and cost-effective alternative for assessing oxygenation, demonstrating a strong quantitative correlation between Spo2/Fio2 and Pao2/Fio2 ratios (17, 30). Calibration bias may overestimate oxygenation in patients with darker skin tones, potentially misclassifying hypoxemia when Spo2/Fio2 is used as a surrogate (27). Recent ESICM guidelines acknowledge Spo2/Fio2 ratios as a surrogate for Pao2/Fio2 in assessing hypoxemia severity, especially when ABG is unavailable (8). Our findings support Spo2/Fio2 as a practical and accessible surrogate to identify patients with AHRF who might otherwise be overlooked, particularly in settings where ABG testing is unavailable or delayed.
Most IMV patients with available measurements were ventilated with lung-protective ventilation strategies developed for ARDS, and growing evidence suggests that these strategies also benefit the broader AHRF population (31, 32). Although limiting Vt and Pplat has been shown to reduce mortality in ARDS (33, 34), our findings suggest that these practices remain insufficiently integrated into routine care. Only 12% of patients on IMV had Pplat recorded within the first 24 hours, consistent with other large cohorts and highlighting persistent gaps in routine monitoring practices (24, 35). Even among patients ventilated in volume-controlled mode, Pplat was documented in only 38% of patients. Although the frequency of these measurements increased with AHRF severity, their overall underuse likely reflects missed opportunities to optimize lung-protective ventilation. This gap extends to ERS, which reflects the elastance of the respiratory system based on end-expiratory lung volume and determines the Vt delivered for a given ∆P, thereby reflecting the size of the aerated lung. High ERS has been associated with increased mortality in ARDS (36, 37), and recent evidence suggests that Vt may need to be tailored to ERS, provided ∆P remains within protective limits (19). However, this interaction has not been evaluated in patients with AHRF. In our cohort, higher baseline ERS was associated with worse outcomes, including prolonged IMV and increased mortality, supporting its value as a marker of severity. These findings mirror observations in ARDS and support the need to investigate personalized ventilation approaches based on lung mechanics, even outside traditional ARDS populations (38).
Our study has several strengths and limitations. Because iCORE includes ICU patients receiving supplemental oxygen or ventilatory support for at least 4 hours, estimates reflect AHRF burden within this respiratory support population rather than across all ICU admissions. iCORE is one of the largest multicenter Canadian databases on respiratory support, providing a detailed and comprehensive data source for investigating patients with AHRF. We used a pragmatic AHRF definition based on Pao2/Fio2 and Spo2/Fio2 thresholds that mitigated common diagnostic challenges associated with the ARDS definition (11, 25) and aligned more closely with how patients are stratified and managed in clinical practice. Oxygenation measurements were recorded once daily, so AHRF classification was based on a single value, which may have introduced severity misclassification. Sensitivity analyses showed consistent patterns in baseline characteristics and outcomes across strata, increasing confidence in the robustness of our definition. As with routinely collected data, missingness was present, particularly for ABG and static pressure measurements. However, this reflects real-world variability in monitoring practices and clinical decision-making. Future research should incorporate long-term follow-up to better define the impact of AHRF on prognosis.
CONCLUSIONS
This large multicenter cohort study found that half of ICU patients receiving supplemental oxygen or respiratory support for at least 4 hours developed AHRF within the first 24 hours of ICU admission, with 76% requiring IMV. ICU LOS, duration of IMV, and mortality increased with AHRF severity. These findings support a pragmatic and operationally feasible definition of AHRF that reflects real-world clinical practice. Standardizing the definition of AHRF may improve patient identification and management, facilitate comparisons across studies, and optimize care.
ACKNOWLEDGMENTS
We acknowledge all members of the Toronto Intensive Care Observational Registry project team for collecting and providing the data for this study.
Supplementary Material
Footnotes
Dr. von Düring was involved in conceptualization, investigation, methodology, formal analysis, writing—original draft, visualization, and project administration. Drs. Goligher, Amaral, Rubenfeld, and Ferguson were involved in methodology, validation, and writing—review and editing. Drs. Urner and Fan were involved in conceptualization, investigation, methodology, validation, writing—original draft, and supervision.
Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website (http://journals.lww.com/ccmjournal).
Dr. von Düring is supported by a research grant from the Valeria Rossi di Montelera Foundation and a Geneva University Hospitals grant for advanced training abroad. He has also received speaking honoraria from Löwenstein Medical. Dr. Goligher is supported by the Canadian Institutes of Health Research, the National Sanitarium Association, and Vyaire. He also reports receiving personal fees from BioAge, Dräger, Fisher & Paykel, Getinge, Heecap, LungPacer, Stimit LLC, Vyaire, and Zoll outside the submitted work. Dr. Amaral’s institution received funding from the Physicians’ Services Incorporated (PSI) Foundation. Dr. Ferguson reports consulting fees from Baxter. Dr. Urner is supported by the Canadian Institutes of Health Research through the Platform of Randomized Adaptive Clinical Trials in Critical Illness (PRACTICAL) and through the International Observational Data Network (INCEPTION) Project. Dr. Fan is supported by the Canadian Institutes of Health Research, the PSI Foundation, and the National Institutes of Health. He also reports receiving personal fees from Aerogen, Baxter, Boehringer Ingelheim, Getinge, Inspira, Mallinckrodt, and Vasomune outside the submitted work. Dr. Rubenfeld has disclosed that he does not have any potential conflicts of interest.
Drs. Urner and Fan contributed equally as senior authors.
Contributor Information
Stephan von Düring, Email: stephan.vonduring@hug.ch.
Ewan Goligher, Email: Ewan.Goligher@uhn.ca.
Andre Carlos Kajdacsy-Balla Amaral, Email: AndreCarlos.Amaral@sunnybrook.ca.
Gordon D. Rubenfeld, Email: Gordon.Rubenfeld@sunnybrook.ca.
Niall D. Ferguson, Email: Niall.Ferguson@uhn.ca.
Martin Urner, Email: Martin.Urner@uhn.ca.
Eddy Fan, Email: eddy.fan@sinaihealth.ca.
REFERENCES
- 1.Choi WI, Shehu E, Lim SY, et al. ; Korean Study group on Respiratory Failure (KOSREF): Markers of poor outcome in patients with acute hypoxemic respiratory failure. J Crit Care 2014; 29:797–802 [DOI] [PubMed] [Google Scholar]
- 2.Villar J, Mora-Ordonez JM, Soler JA, et al. : The PANDORA study: Prevalence and outcome of acute hypoxemic respiratory failure in the pre-COVID-19 era. Crit Care Explor 2022; 4:e0684. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.SRLF Trial Group: Hypoxemia in the ICU: Prevalence, treatment, and outcome. Ann Intensive Care 2018; 8:82. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Roupie E, Lepage E, Wysocki M, et al. : Prevalence, etiologies and outcome of the acute respiratory distress syndrome among hypoxemic ventilated patients. SRLF Collaborative Group on Mechanical Ventilation. Societe de Reanimation de Langue Francaise. Intensive Care Med 1999; 25:920–929 [DOI] [PubMed] [Google Scholar]
- 5.Pham T, Pesenti A, Bellani G, et al. : Outcome of acute hypoxaemic respiratory failure: Insights from the LUNG SAFE study. Eur Respir J 2021; 57:2003317. [DOI] [PubMed] [Google Scholar]
- 6.Ling RR, Ponnapa Reddy M, Subramaniam A, et al. : Epidemiology of acute hypoxaemic respiratory failure in Australian and New Zealand intensive care units during 2005-2022. A binational, registry-based study. Intensive Care Med 2024; 50:1861–1872 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Ferreyro BL, Angriman F, Munshi L, et al. : Association of noninvasive oxygenation strategies with all-cause mortality in adults with acute hypoxemic respiratory failure: A systematic review and meta-analysis. JAMA 2020; 324:57–67 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Grasselli G, Calfee CS, Camporota L, et al. ; European Society of Intensive Care Medicine Taskforce on ARDS: ESICM guidelines on acute respiratory distress syndrome: Definition, phenotyping and respiratory support strategies. Intensive Care Med 2023; 49:727–759 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Qadir N, Sahetya S, Munshi L, et al. : An update on management of adult patients with acute respiratory distress syndrome: An official American Thoracic Society Clinical Practice Guideline. Am J Respir Crit Care Med 2024; 209:24–36 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Bellani G, Laffey JG, Pham T, et al. ; LUNG SAFE Investigators and the ESICM Trials Group: The LUNG SAFE study: A presentation of the prevalence of ARDS according to the Berlin definition! Crit Care 2016; 20:268. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Sjoding MW, Hofer TP, Co I, et al. : Interobserver reliability of the Berlin ARDS definition and strategies to improve the reliability of ARDS diagnosis. Chest 2018; 153:361–367 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Vincent JL, Akca S, De Mendonca A, et al. ; SOFA Working Group. Sequntial Organ Failure Assessment: The epidemiology of acute respiratory failure in critically ill patients(*). Chest 2002; 121:1602–1609 [DOI] [PubMed] [Google Scholar]
- 13.Bilan N, Dastranji A, Ghalehgolab Behbahani A: Comparison of the Spo2/Fio2 ratio and the Pao2/Fio2 ratio in patients with acute lung injury or acute respiratory distress syndrome. J Cardiovasc Thorac Res 2015; 7:28–31 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Choi KJ, Hong HL, Kim EJ: The association between mortality and the oxygen saturation and fraction of inhaled oxygen in patients requiring oxygen therapy due to COVID-19-associated pneumonia. Tuberc Respir Dis (Seoul) 2021; 84:125–133 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Esteban A, Frutos-Vivar F, Muriel A, et al. : Evolution of mortality over time in patients receiving mechanical ventilation. Am J Respir Crit Care Med 2013; 188:220–230 [DOI] [PubMed] [Google Scholar]
- 16.Force ADT, Ranieri VM, Rubenfeld GD, et al. : Acute respiratory distress syndrome: The Berlin definition. JAMA 2012; 307:2526–2533 [DOI] [PubMed] [Google Scholar]
- 17.Matthay MA, Arabi Y, Arroliga AC, et al. : A new global definition of acute respiratory distress syndrome. Am J Respir Crit Care Med 2024; 209:37–47 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Wettstein RB, Shelledy DC, Peters JI: Delivered oxygen concentrations using low-flow and high-flow nasal cannulas. Respir Care 2005; 50:604–609 [PubMed] [Google Scholar]
- 19.Goligher EC, Costa ELV, Yarnell CJ, et al. : Effect of lowering Vt on mortality in acute respiratory distress syndrome varies with respiratory system elastance. Am J Respir Crit Care Med 2021; 203:1378–1385 [DOI] [PubMed] [Google Scholar]
- 20.Gattinoni L, Tonetti T, Cressoni M, et al. : Ventilator-related causes of lung injury: The mechanical power. Intensive Care Med 2016; 42:1567–1575 [DOI] [PubMed] [Google Scholar]
- 21.Sinha P, Calfee CS, Beitler JR, et al. : Physiologic analysis and clinical performance of the ventilatory ratio in acute respiratory distress syndrome. Am J Respir Crit Care Med 2019; 199:333–341 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Hosseini R, Chen Z, Goligher E, et al. : Designing a Bayesian adaptive clinical trial to evaluate novel mechanical ventilation strategies in acute respiratory failure using integrated nested Laplace approximations. Contemp Clin Trials 2024; 142:107560. [DOI] [PubMed] [Google Scholar]
- 23.Goligher EC, Bradbury CA, McVerry BJ, et al. ; REMAP-CAP Investigators: Therapeutic anticoagulation with heparin in critically ill patients with Covid-19. N Engl J Med 2021; 385:777–789 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Bellani G, Laffey JG, Pham T, et al. ; LUNG SAFE Investigators: Epidemiology, patterns of care, and mortality for patients with acute respiratory distress syndrome in intensive care units in 50 countries. JAMA 2016; 315:788–800 [DOI] [PubMed] [Google Scholar]
- 25.Bellani G, Pham T, Laffey JG: Missed or delayed diagnosis of ARDS: A common and serious problem. Intensive Care Med 2020; 46:1180–1183 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Goddard SL, Rubenfeld GD, Manoharan V, et al. : The randomized educational acute respiratory distress syndrome diagnosis study: A trial to improve the radiographic diagnosis of acute respiratory distress syndrome. Crit Care Med 2018; 46:743–748 [DOI] [PubMed] [Google Scholar]
- 27.Sjoding MW, Dickson RP, Iwashyna TJ, et al. : Racial bias in pulse oximetry measurement. N Engl J Med 2020; 383:2477–2478 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Brown SM, Grissom CK, Moss M, et al. ; NIH/NHLBI PETAL Network Collaborators: Nonlinear imputation of Pao2/Fio2 from Spo2/Fio2 among patients with acute respiratory distress syndrome. Chest 2016; 150:307–313 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Sanz F, Dean N, Dickerson J, et al. : Accuracy of PaO2/FiO2 calculated from SpO2 for severity assessment in ED patients with pneumonia. Respirology 2015; 20:813–818 [DOI] [PubMed] [Google Scholar]
- 30.Rice TW, Wheeler AP, Bernard GR, et al. ; National Institutes of Health, National Heart, Lung, and Blood Institute ARDS Network: Comparison of the SpO2/FIO2 ratio and the PaO2/FIO2 ratio in patients with acute lung injury or ARDS. Chest 2007; 132:410–417 [DOI] [PubMed] [Google Scholar]
- 31.Serpa Neto A, Cardoso SO, Manetta JA, et al. : Association between use of lung-protective ventilation with lower tidal volumes and clinical outcomes among patients without acute respiratory distress syndrome: A meta-analysis. JAMA 2012; 308:1651–1659 [DOI] [PubMed] [Google Scholar]
- 32.Neto AS, Simonis FD, Barbas CS, et al. ; PROtective Ventilation Network Investigators: Lung-protective ventilation with low tidal volumes and the occurrence of pulmonary complications in patients without acute respiratory distress syndrome: A systematic review and individual patient data analysis. Crit Care Med 2015; 43:2155–2163 [DOI] [PubMed] [Google Scholar]
- 33.Brower RG, Matthay MA, Morris A, et al. ; Acute Respiratory Distress Syndrome Network: 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; 342:1301–1308 [DOI] [PubMed] [Google Scholar]
- 34.Amato MB, Meade MO, Slutsky AS, et al. : Driving pressure and survival in the acute respiratory distress syndrome. N Engl J Med 2015; 372:747–755 [DOI] [PubMed] [Google Scholar]
- 35.Bellani G, Grassi A, Sosio S, et al. : Plateau and driving pressure in the presence of spontaneous breathing. Intensive Care Med 2019; 45:97–98 [DOI] [PubMed] [Google Scholar]
- 36.Panwar R, Madotto F, Laffey JG, et al. : Compliance phenotypes in early acute respiratory distress syndrome before the COVID-19 pandemic. Am J Respir Crit Care Med 2020; 202:1244–1252 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Gattinoni L, Pesenti A: The concept of “baby lung.” Intensive Care Med 2005; 31:776–784 [DOI] [PubMed] [Google Scholar]
- 38.Constantin JM, Jabaudon M, Lefrant JY, et al. ; AZUREA Network: Personalised mechanical ventilation tailored to lung morphology versus low positive end-expiratory pressure for patients with acute respiratory distress syndrome in France (the LIVE study): A multicentre, single-blind, randomised controlled trial. Lancet Respir Med 2019; 7:870–880 [DOI] [PubMed] [Google Scholar]