Artificial Intelligence-Assisted Emergency Department Vertical Patient Flow Optimization.

Area of Science:

• Emergency Medicine
• Artificial Intelligence
• Operations Research

Background:

• Emergency departments (EDs) face challenges in optimizing patient flow.
• Advances in AI and ML offer opportunities for targeted operational improvements.

Purpose of the Study:

• To evaluate the impact of an AI/ML-driven optimized vertical processing pathway (VPP) on ED patient throughput.
• To assess changes in ED length of stay (LOS) and quality metrics.

Main Methods:

• A non-linear ML model was trained on triage data to predict VPP suitability.
• The model informed a VPP protocol integrated into a queueing theory framework.
• A 13-week prospective trial compared before-and-after ED performance data.

Main Results:

• The optimized VPP protocol reduced average ED LOS by 10.75 minutes (4.15%).
• Adjusted analyses estimated LOS reductions between 7.5 and 11.9 minutes.
• No adverse effects on quality metrics like 72-hour ED revisits or hospitalizations were observed.

Conclusions:

• AI/ML-driven VPP personalization significantly enhances ED throughput and maintains care quality.
• This adaptive approach surpasses standard fast-track systems by responding to ED saturation and patient acuity.
• The methodology is adaptable for personalized patient flow optimization in various emergency care settings.