Fine-Grid Spatial Interaction Matrices for Surveillance Models, with Application to Influenza in Germany

Accurate spatial forecasting of infectious disease outbreaks is critical for public health planning, yet models often rely on administrative boundaries that may not reflect true transmission patterns. We propose the Fine-Grid Spatial Interaction Matrix (FGSIM), which models transmission risk between districts based on distances between individuals to improve both forecast accuracy and spatial resolution of risk mapping. FGSIM creates district-level transmission matrices by defining individual-level transmission risk and aggregating to the district level. We transformed five distance metrics into contact intensity measures to create weight matrices within the endemic-epidemic modelling framework. We evaluated FGSIM using German influenza data (2001-2020) and compared its performance to established methods across four study regions for one to eight weeks ahead forecasts. FGSIM outperformed simplified variants and models without spatial dependence in most cases when evaluated in-sample. For one-week-ahead forecasts, a centroid-based model performed best in three of four regions. For longer-term forecasts, a circle-population based model consistently outperformed others. Risk maps at 100m resolution demonstrated FGSIM’s ability to identify high-risk areas not aligned with administrative boundaries. FGSIM provides a flexible, computationally feasible approach to incorporating individual-level risk into district-level infectious disease models, showing competitive performance in forecasting and enabling fine-scale risk mapping for targeted public health interventions.