Forecasting Hospital Emergency Department Patient Volume Using Internet Search Data

Abstract

We present an efficient and scalable system to predict emergency department (ED) patient volume in hospitals using publicly available Google Trends search data. Search volume data are retrieved for a selected set of context-relevant query keywords with refinements, on which a series of correlation analyses are performed, and a multiple regression predictive model is constructed. We also develop a software suite to enable convenient access to data visualization and prediction capabilities by medical and administrative staff. A preliminary demonstration of the method and software is presented with data from a large public hospital as a form of validation. This paper enables informed resource and manpower allocation in hospitals and thus improved ability to respond to patient influx surges, and importantly, can serve as a key mitigation measure against worsening ED congestion problems that plague hospitals.

Keywords

• Data analytics
• data-driven
• predictive model
• multiple regression
• Google Trends
• medical