Identifying Subjects at Risk for Diaphragm Atrophy During Mechanical Ventilation Using Routinely Available Clinical...

Area of Science:

• Critical Care Medicine
• Respiratory Physiology
• Medical Informatics

Background:

• Diaphragmatic respiratory effort is crucial for patient outcomes during mechanical ventilation.
• Directly measuring diaphragmatic contractility is complex and requires specialized tools.
• Identifying non-invasive predictors of diaphragm atrophy is clinically significant.

Purpose of the Study:

• To assess if routine clinical variables can predict diaphragmatic contractility.
• To determine if clinical variables can stratify the risk of diaphragm atrophy.
• To evaluate the utility of machine learning in this prediction.

Main Methods:

• Secondary analysis of a prospective cohort study involving mechanically ventilated subjects.
• Longitudinal collection of clinical variables including breathing frequency, ventilator settings, and blood gases.
• Application of machine learning and statistical models (logistic regression, random-effects trees) to predict diaphragm atrophy risk.

Main Results:

• No single or combined routine clinical variable accurately predicted diaphragm contractility or atrophy (AUC 0.63-0.75).
• Diaphragm atrophy occurred in 38% of subjects (73/191).
• Diaphragm thickening fraction <15% within 48 hours was associated with increased atrophy risk (OR 2.28).

Conclusions:

• Routinely collected clinical variables and ventilator settings are insufficient for reliably assessing diaphragmatic contractility or atrophy risk.
• Early diaphragm thickening fraction (<15% within 48 hours) serves as a valuable tool for stratifying atrophy risk.
• This finding aids in early identification and potential intervention for diaphragm atrophy.