Simulation of AI-driven CT Imaging Queue Prioritization in the Emergency Department

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Healthcare
• Health Systems Engineering

Background:

• Emergency department (ED) CT scan wait times impact patient outcomes.
• Optimizing CT queue management is crucial for timely diagnosis.
• Current first-in-first-out (FIFO) systems may not prioritize critical cases effectively.

Purpose of the Study:

• To evaluate the impact of an AI-driven CT queue prioritization system on ED CT wait times.
• To compare AI-based dispatch against traditional FIFO policies using discrete-event simulation (DES).
• To assess the system's efficiency in identifying and prioritizing actionable CT studies.

Main Methods:

• Retrospective analysis of 313,966 ED visits.
• Development and training of a LightGBM model to predict CT study actionability.
• Discrete-event simulations (DES) comparing FIFO and AI prioritization policies on 20,795 studies.
• Analysis of wait times for actionable and nonactionable studies, including 90th-percentile and median values.

Main Results:

• AI prioritization reduced median wait times for actionable CT studies by 10.75 minutes and 90th-percentile wait times by 43.36 minutes.
• The proportion of actionable findings obtained within one hour increased from 48.33% to 57.30%.
• AI prioritization achieved 80-87% of the maximum benefit possible with perfect predictions, with minimal increases in wait times for non-actionable studies.

Conclusions:

• AI-driven CT queue prioritization effectively reduces wait times for critical findings.
• The system improves diagnostic timeliness without requiring additional hardware or significant disruption to lower-acuity patients.
• This approach offers a scalable solution for optimizing ED imaging workflows using existing data.