Emergency medicine patient wait time multivariable prediction models: a multicentre derivation and validation study

Area of Science:

• Emergency Medicine
• Health Informatics
• Machine Learning

Background:

• Patients and families desire real-time visibility into emergency department (ED) wait times to enhance their experience.
• Improving patient flow and predictability in emergency medicine is a key objective for healthcare systems.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting ED patient wait times.
• To assess the applicability of these predictive models across diverse ED settings.
• To evaluate model performance during the COVID-19 pandemic.

Main Methods:

• Utilized retrospective administrative data from 12 Australian EDs (2017-2019) encompassing over 1.9 million patient episodes.
• Developed and validated statistical and ML models, including random forest and linear regression, for site-specific and global wait time prediction.
• Tested model performance on data from the COVID-19 lockdown period (January-June 2020).

Main Results:

• Median ED wait times varied significantly across sites (24-54 minutes).
• Site-specific ML models demonstrated varying prediction accuracy (median absolute errors from ±22.6 to ±44.0 minutes).
• Important predictors included triage category, recent patient wait times, and arrival time; models were not universally transferable, but performed well during COVID-19.

Conclusions:

• Electronic health record and patient flow data can effectively approximate ED wait times.
• Site-specific models are essential for accurate wait time prediction, as general models lack precision when applied broadly.