Predicting hospital admission for Emergency Department patients using a Bayesian network

Area of Science:

• Health Informatics
• Emergency Medicine
• Machine Learning Applications

Background:

• Hospital admission delays in the Emergency Department (ED) exacerbate ED diversion issues and reduce overall capacity.
• Effective early prediction of hospital admission status is crucial for optimizing patient flow and resource management within healthcare systems.

Purpose of the Study:

• To evaluate the accuracy of a Bayesian network model for the early prediction of hospital admission status in Emergency Department encounters.
• To identify key variables for predicting admission status that are readily available in most ED settings.

Main Methods:

• A Bayesian network model was developed and validated using data from 16,900 Emergency Department encounters.
• The model incorporated nine commonly available variables within the ED setting.
• Model performance was assessed using the area under the receiver operating characteristic curve (AUC).

Main Results:

• The Bayesian network achieved a high accuracy, with an AUC of 0.894 (95% CI: 0.887-0.902) on the validation set.
• The model demonstrated strong predictive capability for hospital admission status.

Conclusions:

• The developed Bayesian network is a highly accurate tool for the early prediction of hospital admission.
• This system can serve as a valuable alert for clinicians to initiate admission processes earlier in the patient's ED visit.
• Early initiation of admission processes can potentially alleviate ED overcrowding and diversion.