We evaluated 3,213 medical records from patients admitted to the ICU. We excluded 1,681 patients since they did not require IMV and 68 since the clinical data was missing. In the initial analysis, we included 1,464 patients who had received IMV. However, we excluded 21 patients later due to the transfer to a different ICU. Thus, we included 1,442 patients in our statistical analysis (Supplementary Figure S1).

We observed a higher frequency of male patients (n = 901; 62.4%), adults (n = 914; 63.3%), with normal BMI (n = 423; 29.3%), or overweight (n = 372; 25.8%; Table 1). Among the previous history of comorbidities, the most prevalent were systemic arterial hypertension (n = 653; 45.3%), smoking (n = 388; 26.9%), diabetes mellitus (n = 325; 22.5%), cardiopathy (n = 310; 21.5%), neurologic sequel (n = 171; 11.9%), alcoholism (n = 221; 15.3%), and pneumopathy (n = 131; 9.1%; Table 1; Supplementary Table S1).

A total of 923 (64%) patients were referred to the ICU by the surgery department and the main reason for the admissions were the need for elective surgery (n = 616; 42.7%), sepsis (n = 375; 26%), cardiopathy (n = 222; 15.4%), polytrauma (n = 210; 14.6%), and traumatic brain injury (n = 197; 13.7%; Table 1; Supplementary Table S2). Ventilation-associated pneumonia occurred in 410 (28.4%) patients, and the need for tracheostomy in 332 (23%) patients; the death of 570 (39.5%) patients was recorded.

Several patients’ characteristics were associated with enhanced lethality, such as older age [RR = 1.512 (95%CI = 1.334–1.713)], enhanced BMI, grades II and III obesity [RR = 1.426 (95%CI = 1.029–1.977)] and grade I obesity [RR = 1.354 (95%CI = 1.085–1.357)], which presented a higher risk of death (Figure 1). Individuals with a previous history of comorbidities of kidney disease [RR = 1.554 (95%CI = 1.251–1.931)], systemic arterial hypertension [RR = 1.271 (95%CI = 1.119–1.443)], and diabetes mellitus [RR = 1.262 (95%CI = 1.099–1.449)] were also at higher risk of death (Supplementary Table S3; Figure 1). The male sex was associated with decreased risk of death when compared to the female sex [RR = 0.776 (95%CI = 0.683–0.880)] (Supplementary Table S3; Figure 1).

We observed older age and higher BMI in the patients who died. Also, these patients were hospitalized for more days and diagnosed with ventilation-associated pneumonia earlier than patients who did not die (Figure 2). On the other hand, we related the lowest risk of death to the use of drugs and alcoholism, and the younger age of the patients might explain this finding in this group (data not shown). The presence of pneumonia caused by mechanical ventilation was associated with more extended hospital stays (Figure 3).

Several diagnoses were associated with enhanced lethality such as those from kidney disease [RR = 1.485 (95%CI = 1.094–2.017)], stroke [RR = 1.480 (95%CI = 1.246–1.757)], sepsis [RR = 1.391 (95%CI = 1.222–1.583)], and the clinical origin for the patients [RR = 1.387 (95%CI = 1.223–1.573)] (Supplementary Table S4; Figure 1). However, patients with traumatic brain injury [RR = 0.744 (95%CI = 0.596–0.928)], polytrauma [RR = 0.665 (95%CI = 0.290–0.836)], or those who needed elective surgery [RR = 0.677 (95%CI = 0.589–0.778)] and those who needed tracheostomy [RR = 0.644 (95%CI = 0.535–0.776)] presented a decreased risk of death (Supplementary Table S4; Figure 1); nevertheless, patients who suffered a traumatic brain injury or polytrauma were also younger (data not shown).

We associated the PEEP >8 cmH₂O at admission with a higher risk of death [RR = 1.621 (95%CI = 1.393–1.887)]. In addition, a higher risk of death also occurred in patients with hypoxemia at admission [RR = 1.365 (95%CI = 1.126–1.655)]. In contrast, a lower risk of death occurred in those with hyperoxia [RR = 0.813 (95%CI = 0.693–0.954)] at admission (Supplementary Table S4; Figure 1).

In the analysis of the first 20 days of intubation, the patients who died required more extended ventilatory support and presented higher PEEP values throughout the first 20 days than those who were discharged, except on the 15^th day of hospitalization (Figure 4). Curiously, the PaO₂ presented lower values in the patients who died between the day of intubation and the 5^th day of follow-up and between the 7^th and 10^th day of intubation (Figure 5). We presented the patients according to the PEEP and the outcome for the 20 days of intubation in Figure 6. It seems relevant to point out that patients who died had more time on PEEP >8 cmH₂O.

In the Pearson correlation between numeric markers (PEEP at admission, PaO₂ at admission, IMV duration, length of hospital stay, time until the pneumonia diagnosis, BMI, and age), no statistically significant correlation was observed, except for the correlation between the IMV duration and length of hospital stay (CC = 0.70; p < 0.001–strong correlation index), as well as the time until the ventilation-associated pneumonia diagnosis (CC = 0.41; p < 0.001–moderate correlation index) and hospital stay (CC = 0.35; p < 0.001–weak correlation index) (Supplementary Figure S2).

In the survival analysis, we demonstrated that PEEP >8 cmH₂O at admission is associated with a survival of 26 days. In contrast, we observed in patients with PEEP ≤8 cmH₂O the survival of 41 days (p < 0.001) and a Hazard ratio of 1.713 (95%CI = 1.340–2.345). Regarding the PaO₂ classification, we found survival values of 40, 27, and 22, respectively, for hyperoxia, normal, and hypoxemia (p < 0.001; Figure 7).

We excluded the BMI and the day of the ventilation-associated pneumonia diagnosis due to a high number of missing data. We also excluded the following markers: previous diagnosis of kidney disease, kidney disease at admission, and use of drugs.

The multivariate analysis by the binary logistic regression performed by the stepwise forward method (likelihood ratio) was significant in determining whether the patients’ characteristics evaluated were likely to predict death [X²₍₉₎ = 288.335; p < 0.001; Nagelkerke R² = 0.245]. Predictors that were significant in predicting the risk of death included older age [elderly ≥65 yo; OR = 2.226 (95%CI = 1.728–2.867)]; male sex (OR = 0.754; 95%CI = 0.593–0.959); sepsis (OR = 1.961; 95%CI = 1.481–2.595); need for elective surgery (OR = 0.469; 95%CI = 0.362–0.608); stroke (OR = 2.304; 95%CI = 1.502–3.534); length of hospital stay (OR = 0.946; 95%CI = 0.935–0.956); hypoxemia (OR = 1.635; 95%CI = 1.024–2.611), and PEEP >8 cmH₂O at admission (OR = 2.153; 95%CI = 1.426–3.250). In contrast, hyperoxia could not predict the risk of death (Table 2).

This study found a 39.5% death rate, which is only associated with patients receiving IMV. In the literature, a multicenter study that analyzed data from 361 ICUs located in the United States of America, Europe, and Latin America and included 5,183 individuals receiving mechanical ventilation reported a 52% death rate in patients that required mechanical ventilation due to respiratory insufficiency (7). Another two studies performed in Brazil and Chile also reported a high prevalence of death in ICU patients who required mechanical ventilation, with 34 and 33.9%, respectively (28, 29). Interestingly, the overall in-hospital mortality in the Brazilian study was higher than in ICU (42% vs. 34%). Our study presented a similar demographic profile to those found in the literature, which showed the prevalence of male patients, and older individuals to be higher in ICU patients with mechanical ventilation; in contrast, our study showed the main causes of mechanical ventilation to be surgery followed by pneumonia, cardiopathy, sepsis, and trauma. Those authors also reported that the factor that leads to the need for mechanical ventilation might influence the outcome. In Brazil, most patients in ICU are male (50.78%) (30), which is similar to the ones found in the United States of America (51.5%) and the United Kingdom (57.2%) (30, 31). As for the age range, adult individuals prevail in Brazil and the United States of America (30, 31).

In this study, we associated the presence of older age, obesity, systemic arterial hypertension, diabetes mellitus, and kidney insufficiency with a higher likelihood of death. This data follows the literature (7, 28, 29). Curiously, these markers seem to be part of the profile of the patients assisted in Brazil since, according to the Brazilian Intensive Medicine Association, the most frequent comorbidities found in patients admitted to ICU in the country include systemic arterial hypertension (66.40%), diabetes mellitus (32.82%), and kidney disorders (11.63%). The prevalence of male patients was also reported (51.30%) by that institution (30). Such comorbidities might lead to the risk of ICU admission, in which diabetes mellitus, for example, is associated with an increased risk of infection in several sites (skin, nervous system, bones, and articulations) (32). Systemic arterial hypertension, in turn, is the most critical morbidity and mortality risk factor in the world and is associated with an increased risk of cardiovascular diseases (33). Finally, kidney insufficiency presents a 57% increase in the mortality risk of critical patients due to its consequences, namely, metabolic acidosis, electrolytic unbalance, and uremic toxicity (34).

Obesity is also a predictor of extended hospital stay since it might affect several organs, mainly the lungs and heart. In addition, it requires differentiated mechanical ventilation management and higher ventilatory weaning expertise (34). The literature reports a relevant study carried out in the United Kingdom, including over 3.6 million individuals, which pointed out higher death incidence in patients with BMI over the band considered healthy [BMI >30 Kg/m² (obesity)]. However, that study identified the influence of age and BMI and reported that low BMI increases death risk in young individuals. At the same time, a higher BMI might have a protective effect in older people (which might be associated with higher nutritional reserve) (35). However, several meta-analyses and other studies have reported that obesity mainly influences the length of hospital stay rather than death risk (36–40). Maybe the obesity variable is part of a Simpson paradox, that is, in which a certain tendency disappears or even reverses when groups are combined, which is perhaps related to the difficulty in asses the severity of patients with obesity in ICU since the most used scores (APACHE II or III, Sequential Organ Failure Assessment and Simplified Acute Physiology Score II) do not take into account patients weight (41–43). Additionally, APACHE II score in individuals with obesity might be over or underestimate since the patients tend to have a low alveolar gradient, mainly due to a higher prevalence of hypoventilation and apnea syndrome, and also have low urine output, leading to an illusory increased kidney dysfunction (41, 44). The obesity role in the outcome of patients admitted to the ICU, mainly those requiring IMV, still needs further studies since a new “pandemic” of individuals with obesity has been observed worldwide (45). Studies also need to assess the Simpson paradox, which could bias the analysis.

It seems relevant to emphasize that comorbidities do not always develop individually; therefore, when considered together, they might increase the likelihood of adverse outcomes even more. It is essential to highlight that the risk factors can be modifiable and reduced by public health policies, awareness-raising, and better access to health services. Implementing campaigns incentivizing healthy eating habits, regular physical exercises, adherence to disease control measures, and stopping smoking and consuming alcohol help manage those diseases. For example, these actions aim to reduce the incidence of obesity, systemic arterial hypertension, and diabetes mellitus and, consequently, might reduce the occurrence of cardiovascular events (46).

Regarding diagnosis at admission, our study shows that patients in treatment with sepsis, cerebrovascular accident, and kidney disorders also present a higher death risk than individuals diagnosed with traumatic brain injury, polytrauma, elective surgeries, and those that evolved to tracheostomy. Some findings in our study disagree with those in the literature since patients with traumatic brain injury and polytrauma were younger than other patients. For example, the cerebrovascular accident, along with the need for mechanical ventilation, presents a high mortality rate (56.6%) and tends to predominate among male patients (52.7%) with a mean age of 60 yo (47, 48). We confirmed this data in our study, which showed that male sex, diagnosis of cerebrovascular accident, and age are more frequent among our patients; however, male sex was not a death predictor in our data.

When considering death risk markers, sepsis is responsible for ~30–60% of deaths in the ICU (49). The highest death risk due to sepsis results from organ failure caused by the host’s deregulated response to the infection. Despite all efforts made to prevent infections and treat patients affected by them, sepsis is still one of the most common causes of death worldwide, with varied rates according to the region (South Africa and Asia are the most affected regions), age (older age is more associated with death risk), and sex (male) (24, 50, 51). As for treatment, empirical antimicrobial therapy is still the base treatment, and its start is indicated in the first 6 h of the diagnosis. Each hour of delay in the treatment represents a 6% increase in the death risk. The literature described that prescribing unsuitable antimicrobial drugs increases death rates and bacteria resistant to antibiotic medication. In addition, antibiotic medicine might eliminate the bacteria from the blood plasma. However, it might not be efficient in preventing pathogen proliferation in the erythrocyte, which might cause the inefficiency of some treatments against sepsis (52). The sepsis profile described is similar to the profile observed in patients assisted at the University Hospital where we carried out the study.

Elective surgeries that require ICU admission represent 9.7% of this treatment. Of those, ~50.4% also present postoperative complications [e.g., pulmonary embolism and cardiac arrest], with a mortality rate from 2.4 to 9.7% (53). We can associate the lower death risk after elective surgery with the preparation that precedes the procedure.

This study described the highest death risk of patients receiving ventilation with PEEP >8 cmH₂O, maintaining hypoxemia. On the other hand, patients with hyperoxemia showed a lower death risk. Some studies have pointed out that PEEP does not reduce the incidence of pulmonary complications and should not be considered a protective factor for a favorable outcome. In addition, PEEP might increase oxygenation; however, in other cases, it might lead to static stretching, resulting in lesions (18, 54). A study carried out the analysis of surgical patients. It showed that PEEP use resulted in a 5% death risk reduction due to decreased postoperative pulmonary complications such as atelectasis and hypoxemia. However, those findings were inconclusive due to research limitations (small sample) (55). Concomitantly, we observed a higher survival rate in patients that used PEEP ≤8 cmH₂O. However, the outcome does not seem to be associated with the PEEP cut-off point in the literature (22, 23). Gatinoni and co-workers (2015) concluded that there is not “a PEEP correct value” and that it must be titrated by taking into consideration several factors (e.g., oxygenation and hemodynamics) (19). In addition, PEEP is not the only risk factor associated with worse outcomes among the ventilator parameters. In this context, it is important to evaluate parameters such as tidal volume, driving pressure, and plateau pressure that can improve the risk of ventilator-induced lung injury (known as barotrauma and volutrauma), which in turn increase the risk of death (56). Importantly, the two terms–barotrauma and volutrauma–reflect the two sides of the same phenomenon: lung injury due to a large distending volume and/or to a high airway pressure.

In extreme cases, hypoxemia might lead to organ failure (57). In contrast, hyperoxemia might lead to acute hyperoxic and acute lung injury, damaging the epithelium and endothelium due to the release of pro-inflammatory cytokines [e.g., Tumor Necrosis Factor Alpha (TNF-α) and Interferon Gamma (IFN-g)], which might start a pulmonary injury process (25, 58). Although hyperoxemia in the first 24 h of hospital admission does not seem to increase death risk in severe trauma patients (59), it is associated with a higher death risk in patients with cardiorespiratory arrest (60). The literature associated the use of supplementary oxygen in patients with hyperoxemia (PaO₂ over 150 mmHg) with the worst clinical outcome, possibly due to vasoconstriction, reduction in the coronary blood flow, and cardiac output, the release of free radicals, and microvascular perfusion modulation (58, 61).

Despite the general reduction in death risk in patients with PaO₂ over 150 mmHg in the first 24 h of ICU admission, high PaO₂ values should not be recommended when we know the etiology of the tissue oxygenation decrease (e.g., due to hampered transportation). Thus, it might not be wise to state that high levels of arterial oxygenation are always beneficial or might cause harmful side effects (62). In addition, the goals of applying PEEP are to improve gas exchange and increase functional residual capacity, but the effects of PEEP on heart function include reduced venous return, increase pulmonary vascular resistance, and afterload to the right heart, which can lead to worsening oxygenation (63).

We identified the following markers as the main predictors for death: female sex, elderly, sepsis, cerebrovascular accident, hypoxemia, and PEEP >8 cmH₂O. Concomitantly, patients undergoing elective surgery and male sex presented lower chances of death.

We developed the study at a trauma referral center in the region. This fact could lead to an increase in the death risk in male patients, which would confirm other epidemiological studies on trauma centers in Brazil (located in Parana, Bahia, and Paraiba states). However, the male sex was associated with the lowest death incidence. A fact that could explain our findings is that these male patients might have had their age as the primary protective factor since they were all younger patients (data not presented).

Among the elderly, traumatic brain injury might increase mortality when associated with several comorbidities, such as falls, which can even contribute to the cause of trauma (64–66). A retrospective cohort study that analyzed data from 8,598 patients reported that most ICU admissions were male patients. However, the analysis did not show a difference between the sexes when comparing the length of hospital stays, but the hospital discharge rate was higher for female patients (67). In addition, older patients are more vulnerable and might develop multiple organ failures faster, leading to an increased death rate in that population (68).

Sepsis is accountable for 25% of ICU admissions in Brazil and shows high mortality rates, which might reach 65%, while sepsis mortality means around the world might reach 40% (69). Being an organ failure caused by the deregulated and unsuitable host response to infection, sepsis is potentially fatal, and its mortality rate is higher in environments of low or medium resources (70).

Elective surgeries usually present a low mortality rate (between 1 and 4%), and pre-operative care procedures are essential for safe surgical treatment. However, the ideal level of such care has not been defined yet, and death still occurs, mainly due to postoperative complications, for example, pulmonary embolism and cardiac arrest (53).

Both hypoxemia and the use of PEEP >8 cmH₂O were factors that increased mortality rates in our analysis. A study developed with rats that analyzed PEEP to prevent postoperative pulmonary complications reported that PEEP >8 cmH₂O prevented such complications (71). However, that study reported a postoperative analysis only. In addition, regarding PaO₂, health professionals are most concerned with hypoxemia than with the harmful effects of hyperoxemia. For this reason, PaO₂ at admission is oftenly higher than recommended. However, the mortality curve related to PaO₂ at admission presents a U shape. The mortality risk increases as much with low PaO₂ as with high. Also, it is relevant to highlight that the oxygen supplementary offer and the PEEP influence the PaO₂ (72). Although PEEP reduces the collapse of alveolar units and the incidence of atelectasis, one of the factors causing hypoxemia (73), the use of high PEEP values might lead to injury induced by static stretching of alveolar units, mainly when we consider the time in mechanical ventilation support since it is usually longer in patients of clinical or trauma origin (18, 74). The PEEP ideal value remains an unanswered question, and if underestimated, it might collapse the alveoli hampering gas exchange. On the other hand, if overestimated, it might lead to alveolar hyperdistention, inhibiting gas exchange, and venous return (19, 20). Therefore, we must compare PEEP titration to the drug administration, which must be applied rationally based on the patient’s condition.

PEEP increases linearly the mechanical power, which is the energy delivered to the alveolus because of the ventilatory parameters set (75). The mechanical power equation might help the clinical team to estimate injuries associated with mechanical ventilation support by observing the variables present in its formula (current volume, respiratory rate, and inspiratory time). Since PEEP increases the mechanical power volume linearly, it also increases the risk of injury associated with ventilation and death (56). Our study showed increased death risk with PEEP >8 cmH₂O, which might be related to lesions caused by the ventilation, which agrees with the literature.

A recent study incorporated PEEP into the PaO₂/FiO₂ ratio to evaluate the mortality predisposition of patients receiving mechanical ventilation, and it was seen to be a good marker. That study also reported that PEEP incorporated into the PaO₂/FiO₂ ratio alters the classification of gas exchange severity in critical patients (76). The pandemic caused by the new coronavirus (COVID-19) raised great interest in PEEP since this disease affects the lungs severely in some cases leading to a condition like that of acute respiratory discomfort syndrome, requiring better mechanical ventilation performance (77).