High-intensity mechanical ventilation in critically ill patients under general anesthesia or with respiratory failure increases the risk of ventilator-induced lung injury (VILI). VILI occurs when small airways and alveoli are repeatedly opened and closed, or become over-distended.1,2 Lung-protective ventilation is recommended in acute respiratory distress syndrome (ARDS) management to reduce VILI and improve mortality in acute respiratory failure patients.3–6 Previous studies have linked parameters such as end expiration pressure (PEEP) and driving pressure to poor outcomes in invasive ventilation,7–10 and factors like flow and breathing frequency also contribute to tissue injury. However, tidal volume (Vt) and end-expiratory lung impedance remain primary contributors to VILI-associated stress and strain.11 Mechanical power (MP)—which integrates Vt, driving pressure, airflow, and frequency—quantifies the work rate on the lungs and has been proposed as a predictor of VILI pathogenesis.4,12,13

Mechanical ventilation generates mechanical stress and strain within the respiratory system. The magnitude of this load, known as the intensity of ventilation, can be quantified by MP, which is independently associated with mortality in both Intensive care unit (ICU) and ARDS patients.14–16 Importantly, the risk of VILI at a given MP depends on factors such as lung size, recruitability, and homogeneity,11,17,18 often referred to as the “baby lung” concept.19,20 Therefore, adjusting MP by normalizing it to compliance of respiratory system (MPCRS) accounts for functional lung size and may improve prognostic accuracy in ARDS patients, as recently demonstrated.20,21 However, it remains unclear whether MPCRS predicts outcomes across all mechanically ventilated ICU patients, as relatively few studies have explored this. This study thus aims to evaluate the association of time-weighted average MPCRS (TWA-MPCRS) with all-cause mortality to assess its prognostic value in the intensive care setting.

In this retrospective cohort analysis, we confirmed that TWA-MPCRS is independently associated with ICU and in-hospital mortality in patients under invasive ventilation. The TWA-MPCRS during the second 24 h was associated with a linear positive relationship with mortality risk in patients with ventilation exceeding 48 h. The association was strong and unaffected by key covariables and confounding variables. Furthermore, our results demonstrate that the TWA-MPCRS is a superior predictor of outcomes in comparison to TWA-MP alone. To our knowledge, this is the first study to examine the prognostic value of TWA-MPCRS in critically ill patients with mechanical ventilation.

Previous studies have shown that MP is associated with poorer outcomes in critically ill patients ventilated for more than 48 h.14 However, MP does not account for the heterogeneous distribution of lung energy load, and the same MP may cause more severe VILI when aerated lung volume is reduced or pulmonary heterogeneity is increased.30 The effects of MP also vary based on lung pathophysiology, such as function, recruitability, and heterogeneity. Evidence further suggests that lung size and its heterogeneity can influence the impact of MP on VILI in patients with acute respiratory failure. Therefore, to more accurately predict VILI risk, specific power—the energy delivered per unit of ventilated lung tissue19,31,32—should be considered by normalizing MP to lung volume, thus better representing the actual energy burden delivered to the lung.18,30,33

ARDS patients often exhibit variability and significantly reduced lung size (“baby lungs”),34,35 with respiratory compliance reflecting the amount of functional lung tissue.19,31,36 MPCRS is more appropriate than normalization by body weight, as lung size is more influenced by disease severity.28,35,37 Zhang et al.38 found that MPCRS had the highest ability to predict mortality in ARDS. Similarly. Similarly, Coppola et al.20 reported that while MP had no causal association with mortality, MPCRS was independently linked to ICU mortality. Further, in severe ARDS patients receiving extracorporeal membrane oxygenation, MPCRS demonstrated better predictive power for 90-day mortality than either MP alone or MP normalized to predicted body weight.21 However, whether these findings apply broadly to all mechanically ventilated patients remains unclear. In our study, we initially examined the association between TWA-MPCRS quintiles and key clinical parameters, including SOFA scores, oxygenation index, lactate levels, as well as ICU and in-hospital mortality, in mechanically ventilated patients. The analysis revealed that a higher TWA-MPCRS was significantly associated with increased disease severity, as demonstrated by elevated SOFA scores and lactate levels, reduced oxygenation index, and higher ICU and in-hospital mortality rates. While TWA-MPCRS, like MP, does not directly cause lung injury or death, it serves as a reflection of the intensity of mechanical ventilation.15 Elevated TWA-MPCRS values are often linked to more severe underlying disease or greater respiratory dysfunction, indicating a need for higher levels of ventilatory support. Consequently, TWA-MPCRS may hold significant prognostic value in mechanically ventilated patients and can assist in identifying those at greatest risk of adverse outcomes. Importantly, the study identified a continuous linear relationship between TWA-MPCRS and mortality in mechanically ventilated patients. After adjusting for confounding variables, multivariate regression analysis confirmed TWA-MPCRS as an independent risk factor for both ICU and in-hospital mortality. These findings suggest that TWA-MPCRS may serve as a valuable stratification marker for recognizing high-risk patients, supporting clinical decision-making by prioritizing more intensive monitoring and therapeutic interventions for those with the most severe conditions.

Moreover, we determined that the AUC of TWA-MPCRS was similar to that of SOFA score, but superior to TWA-MP alone, which is in agreement with previous research on ARDS.38 Compared to TWA-MP, TWA-MPCRS more accurately predicts ICU mortality by capturing the true mechanical power delivered to the lungs. Unlike broader scores such as SOFA, TWA-MPCRS offers a focused physiologic measure of ventilatory intensity and VILI risk. Automated calculation by ventilators or workflow integration could enhance bedside use. While SOFA remains valuable for overall organ dysfunction, TWA-MPCRS adds insight into ventilatory mechanics, helping identify high-risk patients needing closer monitoring. However, the AUC value of TWA-MPCRS indicates only a moderate discriminatory ability, which may reflect the inherent limitations of being a unidimensional metric. The heterogeneity of mechanically ventilated populations (e.g., ARDS, neurological injury, COPD) leads to variability in the mortality risk thresholds associated with MP.39,40 MP may be more strongly associated with mortality in certain subgroups but less predictive of risk in others, diluting its discriminative power when applied to the overall ICU population. Integrating TWA-MPCRS with other clinical predictors, employing stratified modeling techniques, and leveraging advanced machine learning methods could potentially enhance its discriminatory performance. Additionally, MP has limitations in assessing and quantifying VILI and the severity of lung injury. For instance, the energy required to overcome airway resistance does not necessarily lead to VILI.41 Future research in mechanically ventilated patients might explore normalizing specific components of MP, such as elastic power,42 which could have the potential to improve its prognostic predictive value. Further investigations are needed to address these challenges and enhance the utility of TWA-MPCRS in guiding ICU management decisions.

No single component of MP was found to significantly contribute to overall MP.11 Although our subgroup analyses based on PEEP, Vt, Ppeak, and lung compliance showed consistent associations between TWA-MPCRS and increased ICU and hospital mortality, the complex interactions among these variables make it difficult to clarify the specific roles of individual components in VILI. Further research is needed to address this. In addition, our analysis confirms that patients with the highest BMI have correspondingly higher TWA-MPCRS, similar to the findings of González et al., which associate this with reduced chest wall compliance in obesity and the need for higher PEEP to counterbalance the weight load.43 However, further stratified analysis reveals that BMI does not influence the relationship between TWA-MPCRS and mortality. Moreover, no significant interactions were observed for variables such as age, gender, congestive heart failure, COPD or ARDS when patients were stratified according to potential confounding factors. This suggests that the relationship between TWA-MPCRS and mortality is consistent and independent of these variables, supporting the broader applicability of TWA-MPCRS as a metric for evaluating ventilation dynamics and their association with outcomes in diverse ICU patients. Specifically, in non-ARDS patients, factors such as chest wall compliance, hyperinflation (in COPD), or fluid overload (in congestive heart failure) can alter respiratory mechanics, yet TWA-MPCRS appears to remain a robust indicator of prognosis. Its ability to integrate various components of mechanical ventilation makes it a reliable measure of ventilation-associated stress and strain on the respiratory system, even under differing underlying conditions. Further studies are warranted to explore its utility in guiding ventilatory strategies tailored to heterogeneous patient groups.

Static parameters based on plateau pressure were not used in this study because plateau pressure is often not routinely measured in clinical practice, especially without adequate sedation or standardized inspiratory pauses. This limits data accuracy in large databases like MIMIC-IV. Additionally, plateau pressure may underestimate the maximal pressure experienced by vulnerable lung units during dynamic inflation.44 Therefore, we relied on real-time, automatically recorded Ppeak and PEEP to calculate dynamic driving pressure and dynamic MP.27,45 These dynamic parameters offer the advantage of being automatically displayed by the ventilator with each breath, facilitating continuous monitoring of physiological changes and remaining independent of ventilation mode.27,46,47 Numerous studies have also supported the reliability of these dynamic indices as substitutes for static measurements, and evidence shows that elevated dynamic MP is significantly associated with increased 30-day mortality in patients with acute respiratory failure27,46 and a higher risk of postoperative pulmonary complications,47 regardless of ventilation mode. The use of dynamic MP overcomes the limitations of static measurements and provides a more universal and clinically valuable method for assessing VILI risk and enabling early intervention.

This study has several limitations. First, the conclusions can only be applied to patients who have been ventilated longer than 48 h. Second, we did not account for patients’ sedation status or restrict analysis to a single ventilation modality; patients under various ventilatory modes were included, and sedation was not uniformly documented. As MP may vary across different ventilation modes,48 this heterogeneity may affect the accuracy and interpretation of MP values, especially since the MP formula has not been fully validated for all ventilation modes. These methodological limitations may impact the generalizability of our findings. This limitation should be taken into account when interpreting our results, and future prospective studies should aim to rigorously document both sedation depth and ventilation modality to more accurately validate these associations. Third, as the study was retrospective, we measured functional lung size using compliance. Transpulmonary pressures33 or lung computed tomography (CT) scan may provide more accurate pulmonary49 function expansion and pulmonary heterogeneity or the recruitability measurements.38,50 However, as these specific data are not available in MIMIC-IV, it is necessary to further investigate the normalization of MP to lung volume using transpulmonary pressures or CT in the future.

According to our study, TWA-MPCRS can predict poor clinical outcomes in critically ill patients. TWA-MPCRS is positively associated with ICU/in-hospital mortality.

YKD and YGL carried out the design, conducted data analysis and wrote the manuscript. JQF and RH conducted data collection and interpreted the analysis. HY, JNW and YW participated in the manuscript revision. FS designed the study and conducted data analysis and reviewed the manuscript. All authors read and approved the final manuscript.

The studies involving human participants were reviewed and approved by Institutional Review Board of the Massachusetts Institute of Technology and Beth Israel Deaconess Medical Center. De‑identification was performed to ensure patients’ confidentiality. Written informed consent for participation was not required for this study in accordance with the National Legislation and the Institutional requirements.

Not applicable.

No artificial intelligence was used in the study and writing of the manuscript.

Not applicable.

The datasets supporting the conclusions of this article are available in the physionet (https://physionet.org/content/mimiciv/2.0/).