During the study period, 10,874 patients were admitted to the ICU, from which 2623 were ventilated for more than 24 h. Seventy per cent (1826/2623) of the study population were male; median age was 64 (53–72) years, BMI 26 (24–29), SOFA 5 (4–7), and APACHE II 21 (15–25). Median SpO2/FiO2 1 h after intubation was 217 (158–279). Seventy-one per cent (1868/2623) of the study patients had a medical reason for admission. Main comorbidities were HT (30%) and DM (15%) (Table 1).

Median ICU LOS was 12 (6–24) days and median IMV 6 (3–15) days (Table 1). Twenty-one per cent (551/2623) of study patients required tracheostomy and 8% (213/2623) required reintubation during their ICU stay. Median MP—considering all IMV time in controlled modes—was 16 J/min (13–21). Crude death rate in the ICU was 28% (733/2623).

The MP threshold beyond which patients are more likely to experience increased probability of death in the ICU was similar in values between 16.5 and 18 J/min, but maximum values by extreme decimals are actually 17.9 and 18.0 J/min. Therefore, we stablished a MP = 18 J/min as the MP cut-off point (Fig. 1A and B). Median number of hours in patients ventilated at MP > 18 J/min was 34 (8–125) (Table 1).

The univariate analysis showed that the variables age, SOFA, APACHE, SpO2/FiO2 1 h after intubation, admission for medical reasons, emergency admission, HT, DM, COPD, CKD, and heart disease were associated with ICU mortality (Table 1). The number of hours with MP > 18 J/min was also associated with ICU mortality. However, no relationship was found between the number of hours with Vt > 8 ml/kgPBW and the increase in ICU mortality (Table 1). Multivariate logistic regression analysis revealed that the number of hours with MP > 18 J/min was an independent variable associated with ICU mortality (OR = 1.001; 95%CI 1.0001–1.001; AUC 0.7, e-Fig. S1 and Table 2). This means that for each hour with MP > 18 J/min, the probability of death in the ICU increases by approximately 0.1%.

Good correlation was observed between MP hours > 18 J/min and IMV days (r = 0.79, p < 0.001, e-Fig. S2). Simple linear regression showed that 62% of the variability of IVM days can be explained by the number of hours with MP > 18 J/min (R2 = 0.62).

Good correlation was seen between hours of MP > 18 J/min and ICU LOS (r = 0.73, e-Fig. S3). Simple linear regression showed that 53% of the variability in ICU LOS can be explained by the number of hours with MP > 18 J/min (R2 = 0.53).

From the 2,623 total study patients, 277 (11%) were admitted for SARS-CoV-2 pneumonia. Table 3 shows the characteristics of this subgroup of patients. Seventy-two per cent (200/277) were male; median age was 65 years (56–71), similar to the non-COVID-19 subgroup. Median SOFA and APACHE II were 4 and 15, respectively, lower than the non-COVID subgroup (5 and 21, respectively, for SOFA and APACHE II, p < 0.001). COVID-19 patients had higher incidence of DM and HT, lower SpO2/FiO2 one hour after intubation, and longer ICU LOS (Table 3).

The comparison between non-COVID and COVID-19 patients showed that the latter presented greater alterations of variables related to protective ventilation (Table 3). No significant differences in ICU mortality was seen between the two study groups (COVID-19 30% vs. non-COVID-19 28%, p = 0.55).

Age, SOFA, APACHE, and the number of hours with MP > 18 J/min were associated with ICU mortality in the univariate analysis. Similar to what was observed among the general population, no significant differences were found for Vt > 8 ml/KgPBW for ICU mortality (Table 4).

The multiple regression model for ICU mortality showed that the number of hours with MP > 18 J/min was independently associated with higher mortality (OR = 1.003 (95% CI 1.001–1.004, Table 5) with an AUC 0.81 (e-Fig. S4). This means that for each hour with MP > 18 J/min the probability of death in the ICU increases 0.3%.

Although a good correlation was observed between MP hours > 18 J/min and IMV days (r = 0.72, eFigure S5), simple linear regression showed a lower contribution of MP hours > 18 J/min to IMV days (R2 = 0.52) compared to that observed in non-COVID-19 patients. Similarly, the correlation between MP hours > 18 J/min and ICU LOS was good (r = 0.7, eFigure S6), but a lower contribution of MP hours > 18 J/min (R2 = 0.49) on ICU LOS was also observed compared to non-COVID-19 individuals.

There were 67 (2.5%) patients in the non-hypoxemic group (SpO2/FiO2 > 355), 1259 (48%) had mild hypoxemia (SpO2/FiO2 = 355–215), 1286 (49%) moderate hypoxemia (SpO2/FiO2 = 214–90), and 11 (0.4%) severe hypoxemia (SpO2/FiO2 < 90). Patients with higher hypoxemia were more frequently male, older, admitted for medical reason, urgent admission, higher SOFA, more comorbidities (HT,DM and COPD), and higher ICU mortality, IMV days, and ICU LOS. Patients with higher degree of hypoxemia had more hours with MP > 18 J/min and Vt > 8 ml/KgPBW, as well as higher median MP and Vt (Table 6).

The univariate analysis showed a significant association between more hours with MP > 18 J/min and higher ICU mortality in the non-hypoxemic (e-Table S1), mild hypoxemic (e-Table S2), and moderate hypoxemic groups (e-Table S3), but no association was seen in patients with SpO2/FiO2 < 90 (e-Table S4). In the multivariate analysis, these differences only remain in mild and moderate hypoxemic groups (Tables 7, 8 and 9). The OR in both subgroups was 1.002, which means that the probability of death in the ICU increases 0.2% for each hour with MP > 18 J/min.

The Pearson correlation with IMV days between the hours with MP > 18 J/min was ranged between 0.75 and 0.79 in all subgroups (e-Figs. S7–S10). The contribution of MP hours > 18 J/min to IMV days was higher in severe hypoxemia (R2 = 0.83), followed by moderate hypoxemic (R2 = 0.62) and non-hypoxemic subgroups (R2 = 0.62) and, finally, followed by mild hypoxemic patients (R2 = 0.58).

As for the ICU LOS, Pearson correlation showed an R between 0.47 (severe hypoxemic patients, e-Fig. S14) and 0.77 in non-hypoxemic subgroups (e-Fig. S11). For mild hypoxemic patients, the correlation between hours with MP was 0.69 (e-Fig. S12) and 0.74 for moderate hypoxemic patients (e-Fig. S13). The contribution of MP hours > 18 J/min on ICU LOS was approximately 50% in all subgroups (non-hypoxemic: R2 = 0.58, mild hypoxemic: R2 = 0.47, moderate hypoxemic: R2 = 0.55, severe hypoxemic: R2 = 0.51).