Prediction Model of Extubation Outcomes in Critically Ill Patients: A Multicenter Prospective Cohort Study.

Area of Science:

• Critical Care Medicine
• Pulmonology
• Medical Informatics

Background:

• Prolonged mechanical ventilation poses risks, making timely extubation essential.
• Predicting extubation success is vital to prevent complications and optimize patient recovery.
• Current prediction methods may lack accuracy or rely on complex parameters.

Purpose of the Study:

• To develop and validate a versatile model for predicting extubation outcomes in critical care.
• To identify key physiological predictors for successful extubation after a spontaneous breathing trial (SBT).
• To create an accessible tool for clinicians to assess extubation readiness.

Main Methods:

• Prospective, multicenter study involving 499 patients undergoing extubation after a 30-minute SBT.
• Development of a multivariable logistic regression model using eight parameters: age, heart failure, respiratory disease, RSBI, PaO2/FIO2, GCS, fluid balance, and endotracheal suctioning episodes.
• Model validation using bootstrap method and development of an online calculator.

Main Results:

• The novel prediction model demonstrated good performance with AUCs of 0.69 for successful extubation and 0.70 for uneventful extubation.
• These AUCs were significantly higher than those of a conventional model using only the rapid shallow breathing index (RSBI).
• Successful extubation was achieved by 90.8% and uneventful extubation by 65.7% of patients.

Conclusions:

• A novel, accessible prediction model incorporating readily obtainable physiological parameters can accurately forecast extubation outcomes post-SBT.
• This model offers an improvement over conventional methods, aiding clinical decision-making in mechanical ventilation liberation.
• The developed online application facilitates easy integration into clinical practice for predicting extubation success.