Mechanical power trajectories and 28-day mortality in patients with sepsis receiving invasive mechanical ventilation: a retrospective multidatabase cohort study

In patients with sepsis requiring invasive mechanical ventilation (IMV), mechanical power (MP)—a composite metric quantifying the total energy delivered by the ventilator to the respiratory system—varies dynamically over time. Longitudinal assessment of MP may more accurately capture the magnitude, duration, and temporal evolution of ventilator-associated energy exposure during IMV. This study aimed to identify distinct MP trajectory patterns in the first five days of IMV and to examine their association with 28-day all-cause mortality in sepsis. We analyzed data from MIMIC-IV v2.2 to identify distinct MP trajectories in patients with sepsis receiving IMV using group-based trajectory modeling. The primary outcome, 28-day mortality, was evaluated using Cox regression and Kaplan–Meier survival analysis. The eICU served as an external validation cohort. In the MIMIC-IV development cohort (n = 2490), three distinct MP trajectories were identified during the first five days of IMV: low-stable, moderate-flat, and high-rising-then-declining. Compared with the low-stable group, the moderate-flat and high-rising-then-declining groups had significantly higher 28-day mortality risks (adjusted HR 1.64 [95% CI 1.38–1.95] and 3.77 [95% CI 2.92–4.87], respectively; both P < 0.001). These trajectory patterns and their strong associations with adverse outcomes—including higher 28-day and 1-year mortality, increased ICU and hospital mortality, prolonged IMV duration, and longer ICU and hospital lengths of stay—were successfully reproduced in the external eICU validation cohort. MP was independently associated with 28-day mortality in septic patients with IMV; three distinct MP trajectory patterns were identified—low-stable, moderate-flat, and high-rising-then-declining—were identified, each associated with meaningful differences in baseline characteristics and a graded increase in adverse outcomes.