Factors Associated With Late Extubation in Postoperative Patients in the Surgical Intensive Care Unit

Delayed Extubation, Mechanical Ventilation Weaning, Postoperative Intensive Care

Delayed extubation, Surgical intensive care unit, Mechanical ventilation, Weaning, Machine learning, Postoperative care, Rapid shallow breathing index, serum calcium, cumulative fluid balance, weaning predictors

Timely removal of a breathing tube after surgery is an important part of recovery for patients admitted to the intensive care unit (ICU). In some cases, patients may require longer use of mechanical ventilation, which can be associated with longer hospital stays and a higher risk of complications. The purpose of this study is to examine clinical factors that may be associated with delayed removal of breathing tubes in postoperative ICU patients. This study will use information that has already been routinely collected during standard medical care, such as patient characteristics, breathing-related measurements, fluid balance, and laboratory test results. This is a retrospective observational study and does not involve any additional treatments, procedures, or changes to routine clinical care. By analyzing these existing data, the study aims to improve understanding of factors related to extubation timing and support future efforts to enhance postoperative ICU care.

Delayed extubation after surgery is a common clinical challenge in postoperative intensive care unit (ICU) patients and is associated with increased risks of pulmonary complications, prolonged mechanical ventilation, and extended hospitalization. Determining optimal extubation timing remains complex because respiratory mechanics, metabolic status, neurologic recovery, and perioperative stress responses must align for successful ventilator liberation. This study is a retrospective observational study designed to identify clinical factors associated with delayed extubation in postoperative ICU patients and to develop predictive models using routinely collected clinical data. The institutional cohort includes adult postoperative patients admitted to the surgical ICU at Kaohsiung Armed Forces General Hospital who required invasive mechanical ventilation. An external cohort derived from the MIMIC-IV database is used to support model development and cross-domain learning. No additional interventions, procedures, or changes to standard clinical care are involved. Delayed extubation is defined as removal of the endotracheal tube more than 24 hours after initiation of mechanical ventilation, while early extubation is defined as extubation within 24 hours. Demographic characteristics, disease severity scores, respiratory parameters, perioperative laboratory data, and fluid balance variables routinely documented in medical records are included for analysis. Statistical analyses include univariable and multivariable regression models to identify factors independently associated with delayed extubation and hospital length of stay. In addition, machine learning and deep learning approaches are applied to explore complex interactions among clinical variables and to improve prediction performance. Transfer learning techniques are used to enhance model generalizability by integrating information from external ICU datasets while adapting to local institutional data. The primary objective of this study is to improve understanding of physiologic, metabolic, and respiratory determinants of delayed extubation in postoperative ICU patients. The findings may support more individualized ventilator weaning strategies, reduce prolonged mechanical ventilation, and improve postoperative ICU care. This study involves secondary analysis of existing clinical data and poses no additional risk to patients.

No related papers found yet.

Taiwan