Association Between Mechanical Power and 28-Day All-Cause Mortality in Chronic Obstructive Pulmonary Disease Patients Undergoing Invasive Ventilation: Analysis of the MIMIC-IV Database

Introduction

COPD is a progressive and heterogeneous lung disease characterized by abnormalities in the airways and/or alveoli typically accompanied by persistent airflow limitation that is not fully reversible.¹ COPD is the third leading cause of death worldwide, affecting approximately 384 million people. Alarmingly, over half of individuals with COPD remain undiagnosed, making it a major public health issue that demands significant attention.^2–4 Studies suggest that COPD patients experience between 0.5 and 3.5 acute exacerbations per year, marked by worsening cough, increased sputum production, and escalating dyspnea.^5,6 The elevated rates of intensive care unit (ICU) admissions and mortality in acute exacerbations of COPD (AECOPD) place considerable economic and social strain on healthcare systems globally, a burden compounded by the increasing aging population.^7,8

MP refers to the mechanical energy transmitted from the ventilator to the respiratory system, which can potentially lead to ventilator-induced lung injury (VILI). This results from the interaction between the load imposed by the ventilator and the load experienced by the lung parenchyma.⁹ Currently, the most practical formula for bedside measurement in clinical settings is MP = 0.098×V_t ×RR×(PIP –1/2DP), where V_t denotes tidal volume, RR stands for respiratory rate, PIP is peak inspiratory pressure, and DP refers to the driving pressure.¹⁰ The growing recognition of MP as an important factor in ventilator induced lung injury (VILI) has made it an essential parameter in critical care settings. A study has suggested that variations in DP and MP are associated with 30-day mortality in patients with acute respiratory distress syndrome (ARDS), a condition marked by impaired pulmonary oxygenation.¹¹

ARDS is a representative of diseases characterized by impaired pulmonary oxygenation. Its respiratory mechanics are characterized by decreased lung volumes and reduced compliance. Consolidated lung tissue cannot participate in gas exchange, while the remaining relatively normal lung tissue compensates by maintaining gas exchange.¹² In contrast, COPD is an obstructive lung disease marked by severe expiratory flow limitation and significant increases in lung volumes. The reduction in lung elastic recoil (due to emphysema) increases airway resistance, which causes dynamic hyperinflation of the lungs, shortened expiratory time and increased tidal volume, resulting in an elevated end-expiratory lung volume.¹³ As the severity of AECOPD increases, more patients may require invasive mechanical ventilation due to severe gas exchange abnormalities or inadequate airway protection.¹⁴ While studies have confirmed a relationship between MP and mortality in ARDS patients, the distinct respiratory mechanics of COPD compared to ARDS means the association between MP and mortality among COPD patients undergoing invasive ventilation remains unclear.

Building on this background, the purpose of this study was to investigate the association between MP and 28-day all-cause mortality among COPD patients undergoing invasive ventilation.

Materials and Methods

Data Source

Data were obtained from the MIMIC-IV database, which includes information collected from patients in the ICU at Beth Israel Deaconess Medical Center in Boston, Massachusetts.¹⁵ Ethical considerations regarding patient privacy and confidentiality were addressed through deidentification. Access to the database and data extraction were performed by one of the authors, Haipeng Li, who has successfully passed an examination for a collaborative institutional training program (certification number: 13619655). Publicly accessible databases can be accessed and used by researchers with appropriate credentials. Additionally, the ethics committee (The Institutional Review Board of the Massachusetts Institute of Technology and Beth Israel Deaconess Medical Center) approved the database at its inception, eliminating the need for further informed consent.

Participant Selection Criteria

We included all consecutive patients (aged over 18 years) in the ICU with a primary diagnosis of COPD, identified using ICD-9 codes (code = 49,120, 49,121, 49122, and 496) and ICD-10 codes (code = J44, J440, J441, and J449) from the MIMIC-IV database. Patients were excluded based on the following criteria: (1) patients with multiple ICU admissions for COPD, with data extracted only from the first admission; (2) Patients with an ICU stay shorter than 24 hours; and (3) insufficient data on key variables such as respiratory rate, tidal volume, peak pressure, positive end-expiratory pressure (PEEP), and plateau pressure during the first 24 hours in the ICU. Ultimately, 1704 patients were enrolled, and they were divided into three categories based on MP tertiles to further assess the robustness of our results. Figure 1 shows the patient screening flowchart.

Figure 1 Flowchart of the study. Abbreviations: ICU, intensive care unit; MIMIC-IV, Medical Information Mart for Intensive Care IV. Notes: Mechanical power (J/min): Tertile 1 (1.20–11.49), Tertile 2 (11.49–16.17), Tertile 3 (16.17–98.65).

Data Extraction

In this study, the software PostgreSQL (version 10.11) and Navicat Premium (version 12) were used to extract information. Age, gender, race, mortality and the length of ICU stay were extracted. The first dataset of vital signs, laboratory tests, comorbidity diseases and scoring systems were extracted from the first 24 hours of their admission to the ICU. Vital signs included heart rate (HR), systolic blood pressure (SBP), diastolic blood pressure (DBP), mean blood pressure (MBP), and temperature. Laboratory parameters included: white blood cell (WBC), red blood cell distribution width (RDW), hemoglobin (HGB), platelets, red blood cell (RBC), blood urea nitrogen (BUN), calcium, international normalized ratio (INR), sodium, potassium, prothrombin Time (PT), activated partial thromboplastin time (APTT), creatinine, and glucose. Acidity, partial pressure of oxygen (PO₂), Arterial Carbon Dioxide Pressure (PCO₂), total dioxide (TCO₂), and bicarbonate chemistry were extracted. The comorbidities diseases contained congestive heart failure (CHF), myocardial infarct, renal disease, severe liver disease, diabetes, malignant cancer and sepsis. Ventilator parameters include tidal volume (V_t), positive end-expiratory pressure (PEEP), respiratory rate (RR), peak inspiratory pressure (PIP) and driving pressure (DP). The MP was calculated using the following: MP = 0.098×V_t ×RR×(PIP – 1/2DP).¹¹ And driving pressure equals peak pressure minus PEEP.¹⁶ Since mechanical power depends on multiple ventilator parameters, it can be influenced by various factors. To maximize the accuracy of the MP measurement, we utilized the average MP over the preceding 24 hours.

Primary Outcomes

The primary outcome of this study was 28-day all-cause mortality. Follow-up began on the date of ICU admission and ended on the date of death.

Statistical Analysis

Baseline characteristics were described and compared according to the tertiles of the MP. Continuous variables are presented as mean ± standard deviation (for normally distributed data) or medians with interquartile range (for skewed data), while categorical variables are presented as number (%). The chi-square test was used to assess intergroup differences for categorical variables, while one-way ANOVA was employed to evaluate intergroup differences for continuous variables with normal distribution. The mice package in R was employed to perform multiple imputation for missing data. The Classification and Regression Trees method was applied to generate five imputed datasets, with a random seed set to enhance the accuracy, stability, and reproducibility of the results. Cox proportional hazards models were used to calculate hazard ratios (HR) and 95% confidence intervals (CI) for the association between MP and the primary outcome, adjusting for various models. Confounding variables were selected based on a p-value <0.05 from univariate analysis, and clinically relevant variables associated with prognosis were included in the multivariate model. The Model 1 was unadjusted, while Model 2 was adjusted for age, gender, and race. Model 3 included a comprehensive set of covariates: age, sex, race, SBP, DBP, HR, SpO₂, glucose, PO₂, bicarbonate, BUN, calcium, creatinine, potassium, APTT, WBC, RDW, platelet count, HGB, myocardial infarction, CHF, renal disease, malignant cancer, severe liver disease, SOFA score, sepsis criteria, SAPS II score, and vasopressor use.

Additionally, this study used a restricted cubic spline regression model to examine the nonlinear association between MP and 28-day all-cause mortality. Receiver operating characteristic (ROC) curves were used to predict the 28-day all-cause mortality rate and identify the optimal cut-off value. Kaplan-Meier survival analysis was used to estimate endpoint disparities through groups differentiated by the tertiles of MP, and participants were classified into two groups according to the inflection point of MP to assess the incidence of endpoints defined by different levels of MP. A subgroup analysis was carried out to determine whether MP had a different impact on the 28-day all-cause mortality rate based on the varied subgroup. All statistical analyses were performed using R software (version 4.4.1), with a P-value <0.05 was considered statistically significant.

Results

Baseline Characteristics and Demographic Information

Table 1 presents the baseline characteristics of the participants stratified by MP tertiles. A total of 1704 patients were enrolled, with a mean age of 70.7 years (±10.9). Among these patients, 970 (56.9%) were male. The mean MP was 15.19 J/min (±7.0). In total, 372 patients (21.8%) experienced 28-day all-cause mortality. There were no significant differences among the different groups in terms of race, MBP, TCO₂, sodium, APTT, RDW, myocardial infarction, SBP, CHF, diabetes, renal disease, malignant cancer, severe liver disease, or the need for dialysis (P>0.05). As MP increased, several variables showed changes: age, SpO₂, pH, PO₂, bicarbonate, and calcium levels decreased, while DBP, HR, temperature, glucose, PCO₂, BUN, creatinine, potassium, INR, PT, WBC, RBC, platelet count, and HGB levels increased. Additionally, there was increased utilization of vasopressors and a higher incidence of sepsis. Table 2 lists the baseline characteristics of the survival and non-survival groups. We observed that MP was significantly higher in the non-survival group (17.43 (8.95) vs 14.57 (6.26), P<0.001). Non-surviving ventilated COPD patients were older (P<0.001) and had lower SpO₂, while higher DBP, HR, SAPS II, and SOFA scores when compared to the 28-day survival group. In the non-survival group, dialysis and vasopressors were used in 98.4% and 81.2% of cases, respectively, and 87.1% of patients presented with sepsis. However, only 8.1% of non-surviving patients had concurrent severe liver disease, while 16.9% had malignant cancer. Laboratory indicators revealed that the non-survival group exhibited lower levels of pH, PO₂, TCO₂, bicarbonate, and calcium compared to the survival group. Conversely, glucose, BUN, creatinine, INR, PT, APTT, WBC, RDW, and HGB were all significantly higher in the non-survival group (P<0.05) than in the survival group. There were no discernible differences between the two groups for the remaining factors (P>0.05).

Table 1 Baseline Characteristics of Patients Grouped by Mechanical Powera

Table 2 Baseline Characteristics of the Survivors and Non-Survivors Groups

Cox Proportional Hazards Models and Restricted Cubic Spline Regression Models

Three distinct Cox regression models were constructed and shown in Table 3 to assess the impact of MP on the 28-day all-cause mortality rate among COPD patients undergoing invasive ventilation. Based on the Cox regression analysis, MP was positively associated with the 28-day all-cause mortality risk to the Model 1 [HR (95% CI) 1.04 (1.03–1.05), p<0.001]. Using the T1 group as the reference, HRs for T2 and T3 groups in Model 1 were [HR (95% CI) 1.04 (0.79–1.37), p = 0.758] and [HR (95% CI) 1.84 (1.43–2.35), p<0.001]. After adjusting for age, gender, and race in Model 2, higher MP values were associated with an increased risk of 28-day mortality [HR (95% CI) 1.05 (1.04–1.06), p<0.001]. In Model 3, which adjusted for covariates including age, sex, race, SBP, DBP, HR, SpO₂, glucose, PO₂, bicarbonate, BUN, calcium, creatinine, potassium, APTT, WBC, RDW, platelet count, HGB, myocardial infarction, CHF, renal disease, malignant cancer, severe liver disease, SOFA score, sepsis criteria, SAPS II score, and vasopressor use, a higher MP value was associated with an increased risk of 28-day all-cause mortality in the T3 group [HR (95% CI) 1.33 (1.01–1.75), p = 0.045], while the association was not statistically significant in the T2 group [HR (95% CI) 0.97 (0.74–1.29), p = 0.856]. The linear trend tests conducted on 28-day mortality revealed significant outcomes in all three models, with p for trend values of <0.001 in the Model 1 and Model 2, and 0.022 in Model 3. Restricted cubic spline regression models indicated a linear rise in the risk of 28-day mortality as MP increased (P for non-linearity = 0.967) (Figure 2). The results suggest that MP may be a potential indicator of 28-day all-cause mortality among COPD patients undergoing invasive ventilation.

Table 3 Cox Proportional Hazard Ratios (HR) for 28-Day All-Cause Mortality

Figure 2 Restricted cubic spline regression analysis of mechanical power with 28-day all-cause mortality. Abbreviations: CI, confidence interval; HR, hazard ratio; MP, mechanical power. Notes: Heavy central lines represent the estimated adjusted hazard ratios, with shaded ribbons denoting 95% confidence intervals. The horizontal dotted lines represent the hazard ratio of 1.0. MP of 13.68 J/min was selected as the reference level represented by the vertical dotted lines. The p-value for non-linearity is 0.9669, suggesting a linear relationship. When MP is below 13.68 J/min, the 95% CI includes the HR = 1.0 line, indicating that the positive association between MP and 28-day all-cause mortality is not statistically significant in this range. Conversely, when MP exceeds 13.68 J/min, the 95% CI remains above the HR = 1.0 line, demonstrating a statistically significant positive association between MP and 28-day all-cause mortality.

ROC Curve Analysis and Kaplan–Meier Curve

Figure 3 shows ROC curves for the MP for forecasting 28-day all-cause mortality for COPD patients undergoing invasive ventilation. The AUC for MP was 0.602, suggesting a modest ability to predict 28-day all-cause mortality. This study simultaneously gained the best cut-off value of 17.38 J/min for MP, which had a sensitivity of 41.6% and a specificity of 76.12%. COPD patients undergoing invasive mechanical ventilation were divided into high MP group (MP ≥17.38 J/min, n = 473) and low MP group (MP <17.38 J/min, n = 1231) according to the cut-off value. The 28-day survival rates for patients in the T1 and T2 groups were considerably greater than those in the T3 group (p<0.0001), according to the Kaplan-Meier survival analysis (Figure 4a). Figure 4b revealed that patients in the high MP (MP ≥17.38 J/min) group had a significantly higher mortality rate than those in the low MP (MP <17.38 J/min) group (P<0.0001). Figure 5 examines the stability of the relationship between MP and 28-day all-cause mortality across various subgroups among COPD patients undergoing invasive mechanical ventilation.

Figure 3 Receiver operating characteristic curve for mechanical power. Abbreviations: AUC, area under the curve; ROC, receiver operating characteristic.

Figure 4 Kaplan–Meier survival analysis curves for all-cause mortality according to groups at 28 days. (a) Kaplan–Meier survival analysis curves for all-cause mortality according to the tertiles of MP; (b) Kaplan–Meier survival analysis curves for all-cause mortality according to the optimal cut-off value of MP. Notes: MP tertiles: T1 (1.20–11.49), T2 (11.49–16.17), T3 (16.17–98.65).

Figure 5 Forest plots of hazard ratios (HRs) for the 28 day all-cause mortality in different subgroups. Abbreviations: PO2, partial pressure of oxygen; CHF, congestive heart failure; CI, confidence interval.

Subgroup Analysis

Upon stratified analysis considering factors such as age, gender, race, sepsis, PO₂, malignant cancer, and congestive heart failure, the forest plot (Figure 5) revealed no significant interaction between MP and subgroup (P for interaction: 0.114–0.967), suggesting that MP may be a potential indicator of 28-day all-cause mortality.

Discussion

In this study, we explore the association between MP and 28-day all-cause mortality among COPD patients undergoing invasive ventilation, using data from the MIMIC-IV database. Our research demonstrates that higher MP values are significantly associated with an increased risk of 28-day mortality in this cohort. Importantly, the association persists even after controlling for various confounding factors, indicating that MP may serve as a potential indicator of disease severity in COPD patients undergoing invasive mechanical ventilation.

Moreover, our analysis identified a cut-off value for high MP at 17.38 J/min, which provides a potentially valuable metric for assessing the severity of COPD patients undergoing invasive ventilation. However, we recognize that the sensitivity and specificity of this threshold are not perfect, as indicated by our results. Future studies are needed to refine and validate this threshold across diverse cohorts of mechanically ventilated COPD patients. This value may be a reference point for evaluating the severity of COPD patients undergoing invasive ventilation. By integrating MP thresholds into clinical practice, we may improve the identification of high-risk individuals who require intensified monitoring. However, the clinical utility of MP should be considered in conjunction with other parameters of mechanical ventilation to ensure that ventilatory support is neither excessive nor insufficient. Although the exact strategies to lower mortality based on MP are unclear, it may serve as an important parameter for risk stratification and guide clinicians in providing more personalized care.

Currently, MP has shown promise in predicting mortality across various diseases contexts, including ARDS. The findings¹⁷ indicated that MP was independently linked to the ICU mortality among patients with comparable ARDS severity. Sarah Wahlster’s research¹⁸ revealed that in the initial stages of mechanical ventilation, the association between MP and hospital mortality was at its strongest. Furthermore, the findings from Huangpin Wu’s team¹⁹ revealed that MP was a significant factor linked to 28-day mortality in critically ill pneumonia patients requiring ventilation. A retrospective cohort study²⁰ demonstrated that the 24-hour variation in MP was an independent risk factor for ICU mortality in patients with acute respiratory failure after adjusting for confounding factors. Xiaofeng Jiang’s study²¹ involved sustained acute brain injury participant who needed invasive ventilation. The findings indicated that increased MP was linked to higher mortality rates in the ICU. However, given that COPD has unique pathological and respiratory mechanical characteristics, it remains uncertain whether there is a similar relationship between MP and mortality rates among COPD patients undergoing invasive ventilation. Consistent with our research findings, elevated MP may serve as a marker of the severity of underlying lung disease and respiratory dysfunction, potentially guiding clinical decision-making to prioritize intensive monitoring or more aggressive therapeutic approaches for high-risk patients. However, COPD is distinct from ARDS in its pathophysiology, and patients with COPD may tolerate higher levels of mechanical ventilation. Studies have shown that chronic hypercapnic COPD patients often experience improvements in hypercapnia and quality of life with appropriate mechanical ventilation, and long-term survival may be enhanced with optimal ventilatory support.^22,23

The original MP calculation involves three important components.²⁴ The first key component is the elastance of the respiratory system, which relates to the energy associated with VT and ΔP. The second crucial component is airway resistance, which concerns the energy related to gas flow. The third component represents the energy required to overcome the surface tension caused by PEEP. Additionally, a simplified power equation has been proposed, which demonstrates a strong association with the original equation, while being more straightforward and not requiring pressure-volume curves. When calculated in patients without ARDS, the average difference was only 0.196 J/min.²⁵ This simplified power equation is used in this study.

The energy delivered to the respiratory system generates mechanical stress and strain during mechanical ventilation. This load is considered the intensity of mechanical ventilation and can be quantified as a value of MP.²⁵ In COPD patients, higher MP values may be associated with more severe disease and a greater degree of ventilatory support required. However, MP itself does not directly cause lung injury or death. Rather, it reflects the intensity of mechanical ventilation and is associated with more severe respiratory dysfunction.^26,27 High MP values may signal the need for greater ventilatory support and potentially indicate worse outcomes in severe COPD cases, but it must be interpreted in the broader context of the patient’s overall ventilatory status and the mechanical load being applied. In our study, we identified a critical threshold of 17.38 J/min for high MP in COPD patients undergoing invasive mechanical ventilation, based on ROC curve analysis. While statistically significant, this threshold’s predictive value requires further validation.

Numerous studies have indicated that various blood biomarkers are associated with mortality in COPD patients. A relevant research found that serum anion gap is positively associated with all-cause mortality among COPD patients.²⁸ The red cell distribution width-to-platelet ratio is associated with the in-hospital mortality rate of AECOPD patients.²⁹ Furthermore, lower serum albumin levels are inversely associated with higher in-hospital mortality rates in COPD patients in the ICU.³⁰ Elevated blood urea nitrogen levels are associated with an increased risk of mortality in AECOPD patients.³¹ However, despite the association between blood marker and mortality among COPD patients, there is currently no dedicated scoring system to evaluate the severity and risk stratification of COPD in patients undergoing mechanical ventilation. Developing such a system, which integrates MP and relevant biomarkers, would enhance risk stratification and improve patient management.

While our study found an association between high MP and increased mortality in COPD patients undergoing invasive ventilation, several limitations should be considered. First, our investigation is a retrospective cohort study reliant on data from the MIMIC-IV database. While we found a relationship between MP and mortality in this population, causality cannot be established. To confirm these findings, future multicenter, prospective, randomized double-blind controlled trials are necessary for verification. Second, despite utilizing multivariable Cox regression analysis and subgroup analyses, residual confounding factors may still affect clinical outcomes, potentially limiting the applicability of our results. When selecting patients from the MIMIC-IV database, we excluded COPD patients undergoing invasive mechanical ventilation due to missing data. This missing information could potentially impact the conclusions of the study. Although we included COPD patients undergoing invasive mechanical ventilation, we were unable to differentiate the methods of artificial airway establishment, such as nasal, oral, or tracheostomy intubation techniques. Lastly, the study did not statistically evaluate the use of sedatives or neuromuscular blockers in patients, preventing us from distinguishing whether they were in a sedated or awake state during mechanical ventilation. MP may be influenced by patients’ spontaneous breathing; therefore, we attempted to minimize potential confounding factors by using the average MP over the preceding 24 hours. Finally, the MIMIC-IV database lacks specific data on the severity of COPD, such as detailed information on lung function or the MMRC grade. As COPD severity is known to influence patient outcomes, the absence of this information may limit the accuracy of our conclusions regarding the direct impact of MP on mortality. We were unable to account for the varying degrees of disease severity, which could lead to potential confounding effects. More severe COPD patients may be more prone to mortality irrespective of their MP levels, and the absence of such stratifications could impact the strength and generalizability of our findings. In future prospective studies, we plan to address this limitation by incorporating a more detailed stratification of COPD stages to explore their potential impact on the findings.

Conclusions

In summary, this study emphasizes that MP is associated with 28-day all-cause mortality in COPD patients undergoing invasive ventilation. The cut-off value of 17.38 J/min may serve as a reference point for clinicians in assessing disease severity. Further studies are needed to explore whether MP could be a useful monitoring parameter in COPD patients undergoing invasive ventilation management and to determine its potential impact on patient outcomes.