Short-term Bayesian influenza forecasting in a hospital environment using a linear Kalman filter

Accurately forecasting the weekly number of influenza (flu) lab tests and positive cases is important for hospitals as they work to plan staffing, manage supplies, and provide timely patient care. In this paper, we describe a practical use of a Bayesian Kalman filter to forecast weekly flu tests and positives in a hospital setting. Using data from a large hospital system in Ohio, we show that this approach, which blends real-time hospital information with historical patterns, offers a dependable way to anticipate flu-related patient volume as far as four weeks ahead.