Nowcasting combines statistical modelling with real-time data to estimate current levels of activity when complete information is not yet available. Originally used in fields like economics and epidemiology, it also has potential application in healthcare operations.
This script was used during the COVID-19 pandemic. In it, nowcasting methods were used to predict near-term hospital admissions and bed occupancy despite delays in data reporting. By accounting for those delays, nowcasting provided more accurate, up-to-date estimates of system pressure than raw data alone. This approach supports operational planning and resource allocation, helping healthcare teams respond proactively to emerging trends rather than reacting to outdated figures.
As a near-term forecasting approach, the ‘nowcasts’ predict COVID-19 bed occupancy rates for the following 7 days using trends in recent data. One of the advantages of this method is that each prediction does not require data from any factors other than that being measured. The forecasts are produced using the time series model with the lowest cross-validation error. This may either be a fitted exponential smoothing state space model or a fitted autoregressive integrated moving average (ARIMA) model. - Nowcasting for improved management of COVID-19 acute bed capacity, Richard M Wood, British Journal of Healthcare Management 2021 27:2, 1-3 doi.org/10.12968/bjhc.2020.0179
While it is a comparatively simple approach with its limitations, it does have its uses. A full read of the paper is recommended to understand this method and whether it is suitable for your use case, or whether alternative forecasting options would be better suited to the problem at hand.
Project status
Status: Bronze
Rationale: Useful script for short-term operational forecasting. As it is a single script, and does not have documentation, tests or extra features, this has been rated as “Bronze”.