From Internal Validation to External Validation: An Artificial Intelligence-Based Study on Predicting Optimal Timing...

Area of Science:

• Critical Care Medicine
• Artificial Intelligence in Healthcare
• Predictive Analytics

Background:

• Mechanical ventilation weaning is crucial for Intensive Care Unit (ICU) patients.
• Prolonged or premature weaning can lead to adverse outcomes and increased resource utilization.
• Optimizing the weaning process is essential for patient recovery and efficient healthcare delivery.

Purpose of the Study:

• To develop and validate two-stage AI predictive models for optimizing mechanical ventilation weaning.
• To determine the effectiveness of AI-assisted tools in reducing ventilation duration and associated complications.
• To assess the potential of AI in minimizing healthcare costs related to mechanical ventilation.

Main Methods:

• Utilized six years of ICU data from Chi Mei Medical Center to develop AI predictive models.
• Implemented a two-stage approach to predict optimal timing for
• Trying Weaning
• and
• Actual Weaning
• .
• Validated the models externally at Kaohsiung Medical University Hospital.

Main Results:

• The AI models demonstrated high performance, with AUCs of 0.981 for
• Trying Weaning
• and 0.915 for
• Actual Weaning
• in the original center.
• External validation showed strong performance with AUCs of 0.915 for
• Trying Weaning
• and 0.866 for
• Actual Weaning
• .
• AI-assisted tools showed significant potential in reducing ventilation time, complications, and costs.

Conclusions:

• AI-assisted tools can effectively optimize mechanical ventilation weaning timing in ICU patients.
• The developed AI models show robust performance and generalizability through external validation.
• Implementing AI in weaning protocols can lead to improved patient outcomes and resource management.