Optimizing mechanical ventilation: Personalizing mechanical power to reduce ICU mortality ‐ a retrospective cohort study

Alkhalifah AS, Rumindo K, Brincat E, Blanchard F, Helleberg J, Clarke D, Popoff B, Duranteau O, Mohamed ZU, Senosy A. Optimizing mechanical ventilation: Personalizing mechanical power to reduce ICU mortality - a retrospective cohort study. PLoS One. 2025 Feb 13;20(2):e0318018. doi: 10.1371/journal.pone.0318018. PMID: 39946423; PMCID: PMC11825045.

Summary of "Optimizing Mechanical Ventilation: Personalizing Mechanical Power to Reduce ICU Mortality – A Retrospective Cohort Study"

Abstract

This retrospective cohort study explored the relationship between mechanical power (MP) and ICU mortality, evaluating whether MP individualized to ideal body weight (IBW) could reduce ventilator-induced lung injury (VILI) and improve patient outcomes. Using data from the AmsterdamUMCdb, the authors analyzed 2,338 patients who underwent pressure-controlled mechanical ventilation for at least 48 hours. They identified hypoxemia-specific thresholds for MP, applied machine learning models for mortality prediction, and developed a dynamic optimization algorithm to personalize ventilator settings.

Key Points:

1. Study Objective and Scope: The study assessed how individualized mechanical power, adjusted to patient physiology and hypoxemia severity, correlates with ICU mortality, using a large retrospective dataset of over 2,300 ICU patients.

2. Mechanical Power Calculation: MP was calculated using a surrogate formula incorporating tidal volume, respiratory rate, PEEP, and inspiratory pressure, then normalized to IBW to account for patient variability.

3. Mortality and Mechanical Power Correlation: Non-survivors exhibited higher 48-hour time-weighted average mechanical power (TWA-MP) compared to survivors; every 1 J/min increase in IBW-adjusted MP raised the odds of mortality by 2%.

4. Hypoxemia-Stratified Thresholds: Safe upper limits for IBW-adjusted MP were 0.22 J/min/kg for nonhypoxemic, 0.27 for mildly hypoxemic, and 0.34 for moderately hypoxemic patients. No reliable threshold was found for severely hypoxemic individuals.

5. Machine Learning for Prediction: Among six tested models, XGBoost emerged as the most effective for ICU mortality prediction (AUROC 0.88), informing the basis for real-time ventilator optimization.

6. Dynamic Individualization Algorithm: A novel algorithm was proposed to personalize mechanical ventilation by iteratively adjusting tidal volume, respiratory rate, and driving pressure based on real-time acidosis markers and MP limits, aiming to reduce mortality.

7. Clinical Case Impact: Applying the individualized ventilation strategy to non-survivors led to a predicted mortality reduction of 9.4%, highlighting potential translational benefit in real-world settings.

8. Covariate Significance: While MP was a significant predictor, factors like age, lactate levels, and SOFA score had stronger influence on mortality, reinforcing the multifactorial nature of ICU outcomes.

9. Limitations and Generalizability: The study’s retrospective design, single-center data source, and focus on pressure-controlled ventilation limit generalizability. The findings call for prospective validation and external dataset testing.

10. Future Directions: Integration of AI-guided, real-time ventilation strategies into clinical workflows could enable personalized lung-protective ventilation, though challenges include system infrastructure, clinician training, and real-time data acquisition.

Conclusion

This study reinforces the prognostic value of individualized mechanical power in ventilated patients, particularly when stratified by hypoxemia severity. By integrating machine learning with ventilator settings optimization, the authors propose a practical path forward for precision mechanical ventilation. While the results are promising, prospective studies and multicenter trials are necessary to validate these findings and translate them into clinical protocols.

ACCESS FULL ARTICLE HERE

Watch the following video on "Journal Club: Reduce mechanical power to minimize VILI" by Hamilton Medical

Discussion Questions

1. How can individualized MP thresholds be incorporated into daily ICU rounds or automated decision-support systems to improve clinical workflow?

2. What are the barriers to implementing machine learning models like XGBoost in real-time ICU ventilator management, and how can they be overcome?

3. Could incorporating additional physiologic parameters (e.g., compliance, imaging data) refine MP estimates and enhance prediction accuracy for VILI and mortality?

Javier Amador-Castañeda, BHS, RRT, FCCM, PNAP

Interprofessional Critical Care Network (ICCN)

Take Advantage of This Resource

I encourage you to explore this growing library of articles and leverage it to stay informed on the latest in critical care. Visit the collection today at: https://perfusfind.com/ic/

This is another step in making high-quality, evidence-based information easily accessible to the critical care community. As always, thank you for your continued support!

As always, don’t forget to like, share, and subscribe. See you on the other side!

This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.