Photo Credit: freepik

The following is a summary of “Development of a machine learning model for prediction of the duration of unassisted spontaneous breathing in patients during prolonged weaning from mechanical ventilation,” published in the August 2024 issue of Critical Care by Fritsch et al.

Repeated spontaneous breathing trials (SBTs) are crucial to prolonged mechanical ventilation weaning, balancing respiratory muscle load, and preventing over/underloading.

Researchers conducted a retrospective study to develop a machine-learning model predicting unassisted spontaneous breathing duration.

They used structured clinical data from a specialized weaning unit to develop a classifier model for qualitatively predicting an increase in duration, a regressor model for quantitatively predicting the precise duration of SBTs on the next day, and a model for predicting the duration difference between the current and following day. The analysis included 61 features known to influence weaning, integrated into a histogram-based gradient boosting model, which was trained and evaluated using separate data sets.   

The results showed that 18.948 patient days from 1,018 patients were included. The classifier model yielded a Receiver operating curve – Area under the curve (ROC-AUC) of 0.713. The regressor models displayed a mean absolute error of 2:50 hours to predict absolute durations and 2:47 hours for day-to-day difference.

They concluded that the developed machine learning model, while promising results for predicting spontaneous breathing capacity, lacked the prognostic quality necessary for immediate clinical application.

Source: sciencedirect.com/science/article/pii/S088394412400282X