Modeling cold-related excess deaths via stationary vine copulas

Extreme cold temperature events have long been associated with excess mortality via many different causes of death. Climate change is expected to intensify the frequency and severity of these extreme temperature events. To quantify and model cold-related excess deaths and, in turn, to better understand the potential impact of climate change on future mortality levels, we propose a new approach based on the state-of-the-art stationary vine copulas. We adopt the S-vine model for the first time in the context of climate-driven mortality risk, and introduce a special case of the model to aid model comparison and enhance interpretability of the results. This model is referred to as a (stationary) centrally connected C-vine (CCC-vine). Three types of dependence are captured by the proposed models, which are temporal dependence, contemporaneous cross-sectional dependence, and non-contemporaneous cross-sectional dependence. We fit the CCC-vine model to the US regional cause-specific death data over the period 1999–2018 and conclude that the model outperforms various benchmark models including the Gaussian copula model and the VAR model. Based on the fitted models, we generate several temperature scenarios and assess cause-specific excess deaths and overall excess deaths due to extreme cold temperatures. We also analyze and compare the geographical differences in cold-related excess deaths across six continental US regions. The results from our study can help public health interventions during extreme cold events to reduce temperature-driven excess deaths.