The goal of this project was to assess the impact of weather on public health, particularly unusually hot and cold periods.

We used data from England and Wales, where statistics report weekly hospitalizations and deaths for a population of ~60 million, with breakdowns by age, sex, and region. This dataset is granular, and large enough to be informative without the fragmentation seen in very large countries with heterogeneous reporting.

We linked health outcomes to historical weather data and tested for statistically significant effects of heat/cold waves. Despite high noise, our models captured underlying trends and outperformed a zero-change baseline on unseen weeks. Estimated effects were directionally and broadly consistent across regions.

Stakeholder value is mixed because the magnitude of the weather's effect is significantly less than week-to- week fluctuations. The models therefore have use for long term studies or public health messaging, but not for e.g. surge planning.