2683

Background/Aim: Several studies have investigated associations between non-optimal temperature and mortality, with now established analytical methods. However, important aspects of these associations, such as the substantial geographical and temporal variability, still need to be elucidated. For instance, limited evidence exists on how risks changes within a season due to acclimatization and adaptive behaviours. In this contribution, we apply novel methods based on functional data analysis to study the evolution of temperature-mortality risks during the year.
Methods: FDA models consider data as continuous curves instead of scalar variables, thus offering a flexible and elegant framework to depict multi-dimensional changes in complex exposure-response relationships along time. Here we demonstrate the use of FDA models to estimate the association between heat and cold with mortality risk using data from 70 cities in UK in the years 1990-2016. Specifically, we apply nonlinear additive functional regression models, a flexible FDA extension that allows for a nonlinear associations as well as a historical effect, i.e. lags. In addition, the association is allowed to evolve across the time domain of the curve, in this case the season.
Results: Preliminary results show that susceptibility to non-optimal temperature diminishes during the season, both heat in the summer and, equivalently, cold in the winter. In parallel, changes also affect the lag structure, with sharper variations in risk early in the season compared to later periods, both for summer and winter.
Conclusion: This case study demonstrates that FDA models can be applied to study complex aspects of health associations with environmental risk factors, with increased flexibility when compared to more traditional regression methods. These methods can represent a powerful and complementary tool in the set of methodologies developed for environmental epidemiology, with a number of potential application in various areas.