Table of Contents
- Background
- Purpose
- How does RespiCast work?
- What is a Probabilistic Forecast?
- What is an Ensemble Forecast?
- Disclaimers
- References
Background
The European Centre for Disease Prevention and Control (ECDC) first launched the COVID-19 Forecasting Hub in March 2021. The hub attracted a large number model submissions and became a mandatory reference for ensemble forecasts in the EU/EEA region and globally. Based on this, the hub was expanded to include other respiratory disease surveillance indicators like influenza like illness (ILI) and acute respiratory infections (ARI). The new expanded hub, RespiCast, was launched in November 2023 and it continues to provide robust short-term forecasts for major respiratory diseases. ECDC acknowledges the technical contribution of the ISI Foundation and the London School of Medicine and Tropical Hygiene. We also thank colleagues from the US COVID-19 Forecast Hub and the HubVerse team for their support to this project.
Purpose
RespiCast aims to provide health sector decision-makers and the general public with reliable information about the near-term burden of respiratory illnesses in the EU/EEA region. Our secondary objectives are improving the predictive performance of disease forecasts through ongoing evaluation and development, and fostering a community of infectious disease modellers committed to an open-science philosophy.
How does RespiCast work?
RespiCast compiles probabilistic short-term forecasts (1 to 4 weeks) for respiratory disease surveillance indicators across Europe, including ILI, ARI, and COVID-19 metrics. Forecasts are submitted by independent teams employing a variety of modelling approaches. These individual forecasts are then combined to create an ensemble forecast—an established method for improving forecasting accuracy[1]. For more information on technical details and contributors, visit the RespiCast-Covid19 and RespiCast-SyndromicIndicators pages on GitHub.
What is a Probabilistic Forecast?
In RespiCast, forecasts are presented as predictive intervals rather than single-point estimates. The probability assigned to the interval (e.g., 95%) reflects the forecast's confidence, with lower confidence levels resulting in narrower intervals. Probabilistic forecasts better capture the uncertainty of epidemics, which are influenced by factors like human behaviour and seasonality. Point estimates alone do not effectively communicate this uncertainty.
What is an Ensemble Forecast?
An ensemble forecast combines predictions from multiple models to produce more accurate projections for complex systems, such as weather patterns or disease dynamics[2] [3]. This approach addresses structural uncertainty—the variability that arises from different ways of representing a system through mathematical equations or computational models. By incorporating multiple models, ensemble forecasting mitigates the limitations of relying on the assumptions of a single model. Various methods are used to create ensemble forecasts, including taking the median or mean of all forecasted trajectories, or applying weights based on past performance metrics. Research into optimising these methods remains an active field[4] [5].
Disclaimers
Please note that forecasts are built using data from disease surveillance systems and will hence reflect the reported data and not necessarily the underlying disease activity. Due to differences in surveillance systems and data sources between locations, we recommend focusing on trends rather than on absolute numbers to assess the current epidemiological situation, or to compare different countries. Forecasts are inherently uncertain. Reliability of the forecast depends on several factors, including number of submitted models and reporting delays. Past performance is no indicator of future performance
References
[1]: https://royalsocietypublishing.org/doi/full/10.1098/rsif.2022.0659
[3]: https://ajph.aphapublications.org/doi/full/10.2105/AJPH.2022.306831
[4]: https://royalsocietypublishing.org/doi/full/10.1098/rsif.2022.0659