Practical development and operationalization of a 12-hour hospital census prediction algorithm.

Area of Science:

• Health Informatics
• Machine Learning in Healthcare
• Hospital Operations Management

Background:

• Hospital crowding, exacerbated by COVID-19, highlights the need for efficient resource utilization.
• Accurate hospital census prediction is crucial for optimizing resource allocation and patient flow.
• Existing methods may not adequately address short-term (12-hour) prediction needs.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting hospital census 12 hours in advance.
• To assess the accuracy and feasibility of simple empirical models for short-term census forecasting.
• To create a practical tool for hospital decision-makers to utilize census predictions.

Main Methods:

• Development of empirical machine learning models using historical hospital data.
• Application of linear models with ElasticNet regularization for predictive analysis.
• Creation of a dashboard application for real-time visualization of predicted census data.

Main Results:

• Linear models with ElasticNet regularization achieved a relative 95% error of +/- 3.4%.
• The developed models provide accurate and useful predictions for 12-hour hospital census.
• The project was completed within approximately 7 months with limited resources.

Conclusions:

• Simple, accurate machine learning models can effectively predict short-term hospital census.
• The developed prediction tool can enhance hospital resource management and decision-making.
• This approach offers a feasible solution for improving hospital operational efficiency.