Population Adaptation Phenomena Modelled through Functional Regression

Abstract

The majority of weather-related health studies use past data in regression models to model the relationship between a weather exposure and a health issue. A large variety of models have been used over the years for this purpose, progressively increasing their complexity and relevance. However, gaps still exist in the understanding of weather/health relationships which, when filled, would help predicting more accurately the impact of climate change on human health. In this purpose, the present paper introduces functional regression models for environmental epidemiology studies. These models consider continuous curves instead of series of scalar values as input and/or output. Functional models are thus flexible and able to take advantage of the whole information contained in the data to estimate the weather/health relationship. Two types of functional regression models are applied in the present work to model temperature-related cardiovascular mortality in the city of Montréal, Canada. The first type of model seeks to estimate the impact of hourly temperature variations on cardiovascular mortality. The second one seeks to model how the studied relationship evolves during the year. Both applications allows new insights on the temperature/cardiovascular mortality relationship in Montréal. They especially suggest physiological adaptation effects in the response of the population to temperature, either during summer or winter. From this simple application, it is hoped that functional regression models will be used in more complex studies.