Statistical learning methodologies and admission prediction in an emergency department

Area of Science:

• Healthcare Management
• Data Science in Medicine
• Emergency Medicine

Background:

• Emergency department (ED) quality relies on efficient patient treatment.
• Key performance indicators (KPIs) are crucial for measuring ED performance.
• Expert-selected KPIs informed this study's data analysis and model development.

Purpose of the Study:

• To perform exploratory data analysis on ED performance indicators.
• To develop a predictive model for patient admission.
• To utilize expert-selected KPIs for data-driven insights.

Main Methods:

• Retrospective analysis of 172,695 ED records.
• Comparison of three machine learning algorithms for admission prediction.
• Model development based on initial patient information.

Main Results:

• Analysis revealed consistent mean length of stay across weekdays.
• A positive linear relationship was observed between length of stay and patient age.
• The developed admission predictive model achieved an Area Under the Curve (AUC) of 0.79.

Conclusions:

• Selected KPIs can assess ED resource allocation efficiency, particularly for overcrowding.
• The predictive model can support hospital and ED administrators in decision-making.
• The model helps fill information gaps to improve key performance indicators.