Demographically-Informed Heat-Mortality Risk Curves via Risk Graph Neural Networks
For environmental epidemiologists, this method enhances predictive calibration of heat-mortality models by leveraging granular census data, addressing a known limitation of existing DLNMs.
The paper introduces Risk Graph Neural Networks (RGNNs) that incorporate demographic and geographic features to improve heat-mortality risk curve estimation, achieving lower point errors and near-nominal uncertainty coverage during the 2022 heatwave where baselines failed.
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.