Driving pressure, simplified elastic mechanical power, and time to liberation from mechanical ventilation: a competing-risks and time-anchored observational study

Driving pressure ( \(Δ\) P) and mechanical power are associated with liberation from mechanical ventilation, but the temporal interpretation of measurements obtained late in the ventilation course is uncertain. We analysed adults ventilated \(≥\) 24 h in the Medical Information Mart for Intensive Care IV (MIMIC-IV). \(Δ\) P and simplified elastic mechanical power (MP el ) were calculated from contemporaneously charted variables in the final 6-h controlled-mode window before the end of the first invasive-ventilation episode. The primary competing-risk analysis was landmarked at the end of this exposure window; successful liberation was extubation without reintubation, non-invasive ventilation, or death within 48 h, with death as a competing event. Binary outcome associations were examined in the eICU Collaborative Research Database (eICU-CRD). A separate time-zero-aligned, entropy-balanced analysis compared observed \(Δ\) P \(≤\) 14 versus >14 cmH 2 O at the first operational low-support time point. Of 7,126 MIMIC-IV patients with measurable \(Δ\) P, complete-case landmark models included 6,316 ( \(Δ\) P) and 6,296 (MP el ); higher values were associated with slower successful liberation (cause-specific hazard ratios per standard deviation [SD]: 0.90 and 0.76; Fine–Gray subdistribution hazard ratios 0.85 and 0.68). In eICU-CRD (n=1,260), higher \(Δ\) P was associated with 48-h failure (adjusted odds ratio per SD 1.79), although 419 of 420 failures were deaths. In the weighted comparison, the 28-day difference in area under the cumulative-incidence curve for successful liberation was \(-1.12\) d (95% confidence interval [CI] \(-1.70\) to \(-0.58\) ) for \(Δ\) P >14 versus \(≤\) 14. These findings support a prognostic association but do not establish the effect of actively lowering \(Δ\) P.