Multi-model forecasting of respiratory disease activity in Germany during the 2024-2025 season

An eight-model real-time forecasting system for Germany’s ARI (GP consultations) and SARI (hospitalizations) produced well-calibrated nowcasts and forecasts during 2024/25, though peak timing was often too early.

• Why it matters: Respiratory disease activity in autumn and winter is a public-health monitoring priority, and Germany uses multiple surveillance streams that require reliable real-time prediction despite retrospective data revisions.
• What they did: They built a collaborative multi-target, multi-model system for 2024/25, combining nowcasting with model ensembles for ARI and SARI and using single models for confirmed seasonal influenza and RSV.
• The result: Nowcasts performed convincingly despite early-January Christmas break bias, forecasts were generally well-calibrated with many models beating simple benchmarks, but peak magnitude/timing was difficult and ensemble performance was not consistently superior to individual models.

AI summary of the abstract · may contain errors · not medical advice · read the paper.