Determining pre-procedure fasting alert time using procedural and scheduling data.

Area of Science:

• Anesthesiology
• Medical Informatics
• Healthcare Operations

Background:

• Standard pre-anesthesia fasting protocols often result in durations exceeding minimum recommendations.
• Uncertainty in procedure start times leads to prolonged fasting for patients.

Purpose of the Study:

• To compare standard fasting durations with novel approaches using machine learning (ML) and simulation.
• To optimize the calculation of pre-procedure fasting timepoints.

Main Methods:

• Retrospective observational study using scheduling data from a Canadian academic hospital (Jan 2020 - Apr 2022).
• Four ML and simulation-based approaches were employed to predict procedure start times.
• Tree-based ML models were utilized to determine optimal fasting instruction times.

Main Results:

• Standard practice median fasting duration was 10.08 hours for both food and clear fluids.
• The best ML approach reduced median fasting to 7.7 hours (food/non-clear) and 3.7 hours (clear liquids).
• 97.3% of patients met minimum fasting requirements with the optimized approach.

Conclusions:

• Machine learning models can effectively predict procedure start times to optimize pre-anesthesia fasting.
• Optimized fasting protocols reduce patient fasting duration while maintaining safety standards.
• Further research is needed to implement automated fasting alert systems.