Factors associated with the intubation of patients with acute respiratory failure and their impact on mortality: a retrospective cohort study

Abstract

Background

See More

Severe respiratory failure often requires invasive mechanical ventilation, identifying the factors that lead to this need is crucial. This study aims to identify risk factors for invasive mechanical ventilation and clinical outcomes in patients with acute respiratory failure from the time of onset of symptoms to respiratory failure.

Methods

This retrospective cohort included adults with confirmed COVID-19 admitted to Intermediate or Intensive Care Units between May 1, 2020, and May 1, 2021. Inclusion required chest computed tomography (CT) and inflammatory markers (CRP, D-dimer, ferritin, IL-6) within 72 h of admission. The primary outcome was the need for orotracheal intubation and its association with mortality. Multivariate Cox regression and time-stratified analyses were performed.

Results

Of 550 patients, 346 (63%) required intubation. The overall in-hospital mortality rate was 21.6%. Intubated patients had higher BMI (p = 0.02), SAPS-3 scores (p < 0.001), and elevated CRP, IL-6, and D-dimer. CT findings showed greater lung consolidation, especially after the second week. SAPS 3 and time from symptom onset to intubation were independent predictors of mortality. Patients intubated ≥ 15 days after symptom onset had significantly higher mortality (OR = 2.13; 95% CI: 1.07–4.23), despite similar oxygenation levels at the time of intubation. These patients also had longer use of non-invasive support.

Conclusion

Delayed intubation beyond 15 days from symptom onset is associated with increased mortality. Integrating inflammatory markers and CT findings may help identify patients at risk for clinical deterioration. Timely transition from non-invasive to invasive support may improve outcomes.

Introduction

Acute respiratory failure can occur due to different pathologies, which may be due to direct lung injury, such as pneumonia, or diseases in which the primary injury does not occur in the lung parenchyma, such as pulmonary thromboembolism [1]. Once respiratory failure occurs, patients are treated with supplemental oxygen using conventional devices, non-invasive ventilation, high-flow nasal cannula and, depending on the condition’s evolution, invasive mechanical ventilation [1, 2].

It is known that COVID-19 presents, at least initially, a non-specific clinical course and that up to 17% of patients may require mechanical ventilation, whether invasive or non-invasive [3]. Especially at the beginning of the condition, there is a disproportion in severity between the clinical findings, laboratory and radiological tests with patients deteriorating rapidly at a later stage of the disease [4]. The patient is intubated in a decision-making process that may involve level of consciousness, respiratory rate, oxygen saturation, fraction of inspired oxygen and breathing pattern. However, there is no objective, dichotomous and prospectively validated intubation criteria for orotracheal intubation (OTI) in patients with hypoxemic respiratory failure [5]. Therefore, there may be significant differences between doctors regarding the correct time to intubate a patient and this possible delay may impact the clinical outcome [6].

Studies have also demonstrated the association between inflammatory markers and increased mortality in patients affected by pneumonia, whether viral or bacterial [7,8,9].

Therefore, this study seeks to characterize the risk factors associated with tracheal intubation and mortality in patients with acute respiratory failure. Differing from other studies on the subject that characterize these data temporally from admission to the Intensive Care Unit (ICU), in the present study, in an innovative way, these outcomes as well as radiological changes and inflammatory markers are analyzed and cross-referenced from the onset of symptoms.

Methods

The present study is a retrospective cohort study and for this reason it does not have a clinical trial number. It is registered on Plataforma Brasil (https://plataformabrasil.saude.gov.br), which is a national and unified database of research records involving human beings linked to the brazilian federal government where it is approved by the Ethics Committee in institution research according to CAAE number 49630021.7.0000.0071. Likewise, this study also has approval from the Research Project Management System of the institution where it was carried out under number 4703-21. Due to its retrospective nature, the Ethics Committee waived the need for a Free and Informed Consent Form. Based on this, a cohort study was carried out that analyzed the association between initial care, tomographic findings and serum inflammatory markers from hospital admission to the 15th day of symptoms and the occurrence of respiratory failure as a result of SARS-CoV-2 infection.The research was conducted in a reference quaternary hospital in Brazil, including patients consecutively admitted to ICU and Intermediate Care Unit due to COVID-19 over the course of a year, from May 1, 2020 to May 1, 2021.

Furthermore, the manuscript adhered to the guidelines proposed in the STROBE guidelines.

Inclusion criteria

Patients over 18 years of age admitted to the ICU or Intermediate Care Unit due to COVID-19 were included if they had a positive SARS-CoV-2 polymerase chain reaction (PCR) test, presented within 15 days of symptom onset and had both chest computed tomography (CT) and inflammatory markers (C-reactive protein, D-dimer, ferritin and IL-6) performed within 72 h of hospital admission.

Exclusion criteria

Patients with chronic lung disease using home oxygen prior to COVID-19 infection; need for tracheal intubation unrelated to COVID-19; heart failure with ejection fraction less than 40%; tracheostomy prior to hospital admission and hospital stay of less than 24 h.

Outcomes

Primary

The primary outcome was to investigate the risk factors associated with orotracheal intubation and its impact on mortality in patients with respiratory failure due to COVID-19, evaluating the association between clinical, laboratory, and radiological findings from the onset of symptoms to the need for invasive mechanical ventilation.

Secondary

Compare the clinical characteristics between patients who did or did not require invasive mechanical ventilation, evaluate the temporal evolution of inflammatory markers (C-reactive protein, D-dimer, Ferritin and Interleukin-6) and correlate them with chest CT findings, evaluate the images of chest CT within 3 weeks of the onset of symptoms and correlate the previous use of non-invasive respiratory support with the time elapsed until tracheal intubation.

Sample size calculation

The sample size was estimated using PS: Power and Sample Size Calculation software version 3.1.2 (Vanderbilt University, TN), based on an expected prevalence of respiratory failure due to COVID-19. It was assumed that 70% of patients admitted to the ICU or Intermediate Care Unit would require some form of ventilatory support, and among these, 30% would progress to orotracheal intubation (OTI), yielding a 2:1 ratio between non-intubated and intubated patients [3]. Using a type I error (α) of 0.05 and a statistical power of 95%, the initially calculated sample size included at least 345 consecutively admitted patients, with approximately 230 requiring invasive mechanical ventilation. To account for potential exclusions due to incomplete data in the electronic health records—such as missing laboratory or imaging results within the required time window—an additional 30% was added to the estimated sample size. Thus, the final required sample was adjusted to a minimum of 450 patients. All patients had complete outcome data (discharge or death), and no loss to follow-up occurred.

Collected variables

The data analyzed were collected from a de-identified database and the variables collected were:

Date of admission to the ICU or Intermediate Care Unit, age (years), sex (male or female), Simplified Acute Physiology Score 3 (SAPS-3) on admission to the ICU, presence of comorbidities (Diabetes Mellitus, Systemic Arterial Hypertension, Chronic Obstructive Pulmonary Disease, Bronchial Asthma, Smoking, Neoplasia); need for ventilatory support (yes or no); mechanical ventilation time (days); date of hospital discharge or death; length of hospital stay (days); laboratory tests (C-reactive protein, D-dimer, Ferritin and Interleukin-6) measured daily until the 15th day of symptoms and chest CT images in weeks 1, 2 and 3 of symptoms.

In order to allow a more detailed analysis of the time between the onset of symptoms and OTI, we divided this variable into quartiles with similar population size. This allowed us to identify patterns or associations between different time intervals and outcomes, providing a more comprehensive understanding of the relationship between time to intubation and in-hospital mortality.

Data reliability

Patients consecutively admitted to the hospital’s ICU and Intermediate Care Unit during a period of one year were retrospectively reviewed to complete the calculated sample size. Those who met the inclusion criteria were eligible for analysis. Chest CT findings were evaluated based on the extent of ground-glass opacities and the presence or absence of consolidations, while the temporal behavior of key inflammatory markers was also assessed. Patients were followed until hospital discharge or death. The retrospective design and reliance on electronic health records posed some limitations, particularly for variables not consistently documented in structured fields. To ensure data integrity, we included only variables that were systematically and reliably recorded across all patients.

Ethical aspects

The study followed the ethical principles of research involving human beings of Resolution 196/96 of the National Health Council (BRAZIL, 1996) respecting the fundamental principles of autonomy, beneficence, non-maleficence, justice and equity.

Ventilatory support protocol

The hospital institution had a respiratory failure management protocol where non-invasive ventilation (BiPAP) could be attempted for a period of up to 30 min aiming for a FiO2 ≤ 50% and a target of EPAP ≤ 10 cm H2O and a pressure delta ≤ 10 cm H2O aiming for a respiratory rate ≤ 24 bpm and SpO2 ≥ 94%. If this protocol failed, the patient underwent OTI (Fig. 1) [10].

Fig. 1

Decision flowchart regarding ventilatory support. *The 30-minute NIV test is performed with a non-ventilated mask (without exhalation valve) and a double circuit on a mechanical ventilator with a barrier filter at the expiratory outlet. AHRF: acute hypoxemic respiratory failure; OTI: orotracheal intubation; VNI: non-invasive ventilation; RR: respiratory rate; BPM: breaths per minute; SpO2: peripheral oxygen saturation; FiO2: inspiratory fraction of oxygen

Intensive care unit and intermediate care unit admission protocol

Patients requiring oxygen supplementation via a nasal cannula with flow > 3 lpm or using NIV to maintain a SpO2 > 94% and/or respiratory rate ≤ 24 bpm were admitted to the Intermediate Care Unit. Patients using invasive mechanical ventilation, hemodynamic instability requiring vasopressors or lactate ≥ 36 mg/dl were admitted to the ICU.

Chest computed tomography evaluation

All CT scans were performed according to institutional protocol and interpreted by two board-certified radiologists with subspecialty training in thoracic imaging. Tomographic findings were described following the standardized nomenclature defined by the Fleischner Society for thoracic radiology, using terms such as ground-glass opacities, mosaic paving, and consolidations [11]. The extent of ground-glass opacities was assessed semi-quantitatively using an adaptation of the Chest CT Severity Score proposed by Tsakok et al., with pulmonary involvement categorized as ≤ 25%, > 25% and ≤ 50%, or > 50% [12, 13]. In the event of discrepancies regarding the extent of parenchymal involvement or presence of consolidation, a consensus reading was performed through joint review. This process aimed to minimize inter-observer variability and enhance the reliability of radiological data.

Data analysis

Categorical variables were presented as absolute and relative frequencies. Quantitative variables were expressed as mean and standard deviation or as median and interquartile range, depending on their distribution. The Kolmogorov-Smirnov test was used to evaluate the distribution pattern of continuous variables. Proportions were compared using the Chi-square test or Fisher’s exact test when appropriate. Quantitative variables were compared using the Mann-Whitney test or ANOVA for non-normally distributed variables, with Bonferroni correction applied for multiple comparisons.

For variables with missing values, we used complete case analysis without data imputation. In the case of missing data greater than 2% for a patient, the patient was excluded from the study. To assess associations between explanatory variables and outcomes, Cox regression models were used to compare event rates over the length of hospital stay. Variables included in multivariate models were those with statistical significance in univariate analysis (p < 0.05) or established clinical relevance, while those with high collinearity were excluded. The fit of the Cox regression model was evaluated using log-likelihood and Schoenfeld residuals to verify the proportional hazards assumption. Additionally, logistic regression models were assessed using the Hosmer-Lemeshow goodness-of-fit test, which indicated no evidence of poor fit. Results were reported as risk ratios with 95% confidence intervals.

All p-values were two-tailed, and statistical significance was defined as p < 0.05. Analyses were performed using the Statistical Package for the Social Sciences (SPSS) version 26.0 (SPSS Inc., Chicago, IL, USA).

Results

During the period, 852 patients were admitted to the ICU and Intermediate Care Unit, but 149 were excluded due to missing data; 114 were ineligible for the study, with 39 also excluded due to missing laboratory or tomographic tests or a negative COVID-19 test later (Fig. 2).

Fig. 2

Participant flowchart

Regarding demographic characteristics and conditions prior to OTI, a higher Body Mass Index (BMI) (P = 0.02), a higher SAPS-3 (P < 0.001) and a shorter time from symptoms to hospital admission (P < 0.001) were evidenced in patients who required invasive mechanical ventilation. All other variables were statistically similar (Table 1).

Table 1 Baseline characteristics upon admission to the ICU

Patients who progressed to invasive mechanical ventilation had a higher CRP, Interleukin-6 and D-dimer values ​​when compared to other patients. On the other hand, Ferritin showed no difference in 8 days comparing the two groups of patients (Fig. 3).

Fig. 3

Inflammatory laboratory tests before intubation. A = Comparison of C-Reactive Protein between intubated and non-intubated patients; B = Comparison of Ferritin between intubated and non-intubated patients; C = Comparison of Interleukin-6 between intubated and non-intubated patients, D = Comparison of D-dimer between intubated and non-intubated patients. Dotted line represents intubated patients and solid line represents non-intubated patients

Analyzing the CT scans, there was a greater area of ​​consolidation, as well as a higher percentage of lung injury above 25% in patients who were intubated and they became more evident after the second and third weeks respectively (Fig. 4).

Fig. 4

Computed tomography follow-up per week. A = percentage of consolidation in intubated and non-intubated patients; B = percentage of lung injury. Black bar represents intubated patients and light bar represents non-intubated patients

Regarding inflammatory markers, it was shown that elevated CRP and serum D-dimer were associated with a greater presence of consolidation on chest CT (Fig. 5).

Fig. 5

Association of inflammatory markers with the presence of consolidation on chest computed tomography per week

Variables that were clinically relevant to the outcome and statistically significant among intubated patients were included in a COX regression model to identify independent risk factors for mortality prior to OTI. Only the SAPS-3 score and the time elapsed from the onset of symptoms to OTI emerged as significant factors for mortality in this period (Table 2).

Table 2 Risk factors for mortality in a multivariate Cox model prior to tracheal intubation

The overall in-hospital mortality rate was 21.6%.When dividing the time interval between the onset of symptoms and tracheal intubation into quartiles and relating them to hospital mortality, we observed a higher mortality rate in patients intubated 15 days after the onset of symptoms (OR = 2.13; 95% CI 1.07–4.23), while times shorter than 11 days were associated with a lower risk of death. This result remains significant even after adjustments for the SAPS 3 score and body mass index (BMI) (Fig. 6).

Fig. 6

Mortality in relation to time to intubation from the onset of symptoms. The top figure is unadjusted and the bottom figure is adjusted for SAPS 3 and BMI. A: Hospital mortality and days to indicate intubation; B: Odds ratio according to days to intubation adjusted for BMI and SAPS 3 using non-intubated patients as a reference; C: Odds ratio of only intubated patients according to days to intubation adjusted for BMI and SAPS 3 only with intubated patients

Examining the group of patients undergoing OTI, it is seen that the average respiratory rate at the time of intubation was 27.5 breaths per minute (bpm) with an average FiO2 requirement of 85% and a ROX index of 4.37. When dividing the intubation time into quartiles and comparing these variables, we observed that there were no significant differences between the quartiles in relation to clinical signs. However, it was seen that the time of using non-invasive support was significantly longer in the quartile with more than 15 days for intubation since the onset of symptoms, with a median of 5 days [interquartile interval of 3–7 days], as well such as the length of stay in the ICU and hospital until intubation (Table 3).

Table 3 Clinical conditions at the time of tracheal intubation comparing the groups according to the time to intubation from the onset of symptoms

Analyzing the devices used separately before OTI and after the onset of symptoms, we found that the time to intubation from the onset of symptoms was longer (P = 0.002) for patients who used a HFNC with a median of 12 days compared to the other respiratory supports used prior to OTI (Fig. 7).

Fig. 7

Median time to intubation from the onset of symptoms in relation to respiratory support. The bars represent the median time with respective 95% confidence intervals

In the analysis of patients stratified by the time from the onset of symptoms to OTI, we observed significant differences in therapy and outcomes. Although oxygenation remained stable, there were a greater number of prone positions in patients who had a longer time until intubation. Furthermore, the use of NIV and HFNC was more frequent in patients with longer intervals until obtaining a definitive airway (42% and 56.5% respectively, in intervals ≥ 15 days) (Table 4).

Table 4 Clinical results among intubated patients, separating the time of intubation from the onset of symptoms into quartiles

Discussion

Patients who develop hypoxemic respiratory failure and require OTI have greater inflammatory activity prior to intubation and develop radiological worsening of the lung injury. Furthermore, risk factors for hospital mortality prior to respiratory failure were higher SAPS-3 values ​​and a longer time for tracheal intubation from the onset of symptoms.

Through the analysis of the relationship since the onset of symptoms, temporal surveillance of inflammatory markers as well as radiological evolution was possible. Intubation performed 15 days after the onset of symptoms demonstrated worsening survival.

The appropriate time for OTI is a central concern since in addition to hypoxemia, the patient outside of invasive mechanical ventilation may present a higher ventilatory effort, leading to high transpulmonary pressures, stress and strain, which may culminate in the development of Patient Self-Inflicted Lung Injury (P-SILI) and thus potentiate lesions in an already diseased lung [14,15,16].

The prolonged time for OTI often correlates with the use of non-invasive ventilatory support, however its use is not without risks [17]. Respiratory fatigue, pulmonary atelectasis and inflammatory activity (biotrauma) can be exacerbated, causing damage to distant organs in addition to the already sick respiratory system, worsening the patient’s prognosis [5, 15,16,17,18,19,20]. It is known that patients with exacerbated COPD managed with non-invasive ventilatory support undergo intubation in up to 15–20% of cases, a number that is already high, however in patients with hypoxemic respiratory failure these numbers jump to 40–60% of failure rate with the non-invasive method [21, 22]. As an aggravating factor, these patients with hypoxemic respiratory failure who progress to OTI will have higher mortality [21, 23], which is one of the reasons for the extreme importance of their monitoring in relation to method failure, whether through tidal volume in patients using NIV or ROX Index in patients using HFNC [2, 24]. At the beginning of the COVID-19 pandemic, the indication was to intubate immediately [25, 26]. However, with greater understanding of the disease, it was seen that these patients could tolerate a degree of hypoxemia with few clinical manifestations, causing these patients to undergo invasive mechanical ventilation later [27]. Therefore, later guidelines recommended that patients, before being intubated, undergo therapy with non-invasive ventilatory support [25].

Even with a physiological rationale for P-SILI in patients on NIV, a meta-analysis of more than 8,000 patients with COVID-19 suggested that the timing of intubation may not have an impact on mortality. However, given the great heterogeneity of the studies analyzed on the topic, this cannot be considered a certainty [28].

The cutoff point for early and late intubation is divided between 24 and 48 h depending on the source used. Studies that used 24 h after ICU admission as a cutoff found no significant differences in mortality [18, 29]. Studies that used 48 h as a cutoff point reported contradictory results, with some observing a higher mortality rate in patients intubated after this period and others finding no significant differences [30]. In another publication, both cutoff points were analyzed in a prospective paired analysis study in Spain, showing an increased risk of mortality in patients undergoing invasive mechanical ventilation both after 24 and 48 h [31]. A retrospective observational study including only 40 patients with COVID-19 found lower mortality in patients intubated before 50 h after ICU admission, but it is clear that the small sample size does not allow definitive conclusions to be drawn about the best time for intubation [32].

In the present study, unlike other existing studies, the temporal analysis takes place not from hospital or ICU admission, but rather with the onset of COVID-19 symptoms, providing a more detailed analysis of the patient whether in terms of their clinical, radiological or laboratories trends. Analyzing the numbers presented, the median from the onset of symptoms to hospitalization was 10 days and the median from the onset of symptoms to OTI was 11 days, the fact indicates that OTI occurred 24 h after hospitalization, corroborating all studies which showed a positive influence from this early intervention. Assessing the time window for OTI from the onset of symptoms provides a more accurate perspective on the evolution of the disease from the appearance of the first signs of infection and can help with early identification of rapidly deteriorating patients. This makes it possible to implement interventions before the condition reaches more advanced stages.

ICU arrival time can vary significantly between patients and some may be admitted late due to factors such as access to medical care, initial screening or individual characteristics of the disease. By starting the count from the symptoms, these variations are sought to be mitigated. Analysis from the onset of symptoms allows for a more in-depth understanding of how the initial immunological response may influence the need for OTI and other clinical outcomes.

Regarding inflammatory markers, we found higher values ​​of CRP, IL-6 and D-dimer in patients undergoing OTI. The topic of hyperinflammatory and hypoinflammatory patterns of ARDS has been increasingly studied in recent years and it is known that patients with greater inflammatory activity throughout the disease require higher doses of vasoactive drugs, administration of fluids, PEEP values ​​as well as they have higher mortality [33,34,35]. The importance of constant assessment of the patient in relation to their fluid needs is highlighted here, since at the same time that hyperinflammatory patients need to use higher PEEP values ​​in an attempt to correct their hypoxemia, these same individuals also receive a greater amount of fluids, which can in itself worsen the lung ventilation/perfusion relationship, leading to worsening hypoxemia.

Furthermore, we found an association of patients with greater inflammatory activity and failure to non-invasive pulmonary ventilation strategies and the consequent need for OTI. In addition, we also showed that patients infected by SARS-CoV-2 and who presented a hyperinflammatory pattern had a greater presence of pulmonary consolidations on chest CT. These points are worth highlighting, since today there is much discussion about the ARDS phenotype, but little is known about what to do about it. Many studies today focus on whether a personalized approach to treating ARDS is valid, as occurs in oncology, as there are still doubts about its impact on the clinical outcome of these individuals [35, 36].

Although this study has several strengths, including a representative sample in a highly specialized hospital, one-year follow-up encompassing different pandemic phases, and robust statistical analyses adjusted for key severity and demographic variables, it also has limitations. The observational nature of the study prevents causal inference [37]. Data collection was performed at a single center, which limits its generalizability. Corticosteroids were administered to the vast majority of patients (98%), limiting their analytical utility. The use of neuromuscular blockers and vasopressor therapy were not systematically recorded in the database because they were recorded in an open field in text format. Data on lung compliance and cumulative fluid balance were also not analyzed due to variability in the way they were done and recorded in the medical records, preventing their reliable analysis.

Conclusions

In patients with COVID-19, the decision to intubate late was associated with a deleterious effect. Our data suggest that clinicians consider a window of less than 12 days from symptom onset to perform intubation. Furthermore, the increased inflammatory response may be a potential warning to determine appropriate ventilatory care for these patients. In patients with COVID-19 receiving non-invasive ventilatory support, especially patients using a high-flow nasal cannula, intubation cannot be delayed. Based on the observational nature of the data, this information should be validated in future studies.