Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This study introduces Risk Graph Neural Networks (RGNNs) to enhance the estimation of heat-related mortality risk by incorporating demographic and geographic context, which traditional Distributed Lag Non-linear Models (DLNMs) overlook. By optimizing DLNM coefficient vectors with granular census features, RGNNs achieve improved predictive calibration while maintaining interpretable risk curve outputs. Evaluated during two extreme heat years in England and Wales, RGNNs demonstrated significantly lower point-errors and better uncertainty coverage compared to baseline models, particularly during the 2022 heatwave when traditional models faltered.
RGNNs not only enhance predictive accuracy for heat mortality risk but also maintain interpretability, outperforming traditional models during extreme heat events.
Estimating heat-related mortality risk is a core task in environmental epidemiology, typically addressed with Distributed Lag Non-linear Models (DLNMs); interpretable exposure-response surfaces fitted to temperature-mortality time series. DLNMs are effective but ignore demographic and geographic context, despite well-established relevance to heat vulnerability. We propose Risk Graph Neural Networks (RGNNs), a hierarchical GNN encoder that uses granular census features to optimise DLNM coefficient vectors, preserving interpretable risk curve outputs while substantially improving predictive calibration. Evaluated across 10 regions of England and Wales on two unprecedented heat years, RGNN variants maintain both lower point-errors and near-nominal uncertainty coverage during the 2022 heatwave where baselines collapse.