A new look at weather-related health impacts through functional regression

Pierre Masselot,Pierre Gosselin,Diane Bélanger,André St-Hilaire,Fateh Chebana,Taha B M J Ouarda

doi:10.1038/s41598-018-33626-1

Abstract

A major challenge of climate change adaptation is to assess the effect of changing weather on human health. In spite of an increasing literature on the weather-related health subject, many aspect of the relationship are not known, limiting the predictive power of epidemiologic models. The present paper proposes new models to improve the performances of the currently used ones. The proposed models are based on functional data analysis (FDA), a statistical framework dealing with continuous curves instead of scalar time series. The models are applied to the temperature-related cardiovascular mortality issue in Montreal. By making use of the whole information available, the proposed models improve the prediction of cardiovascular mortality according to temperature. In addition, results shed new lights on the relationship by quantifying physiological adaptation effects. These results, not found with classical model, illustrate the potential of FDA approaches.

Highlights

Climate change adaptation is an important challenge for years to come
Functional regression is a subset of functional data analysis models (FDA)[17] which represent a statistical framework to deal with continuous curves instead of series of discrete data points
Comparison results show that the predictive performance of functional regression is not vastly superior to those of generalized additive models (GAM) and distributed lag nonlinear models (DLNM)

Summary

Introduction

Climate change adaptation is an important challenge for years to come. With the expected increasing number of heat wave events as well as other meteorological changes, human health could greatly be impacted by climate change[1]. They are not able to estimate other important aspects of weather-related health, such as the impact of seasonal anomalies (e.g. early heat waves or late cold spells) or the potential impact of intra-day variations Not accounting for these effects significantly hinders the predictive performances of classical models and the assessment of the impacts of climate change. Using each hourly measurement as a different explanatory variable in a multiple regression model would lead to unstable estimates because of the high collinearity between them, as well as the high number of coefficients to estimate which increases the uncertainty[18] These issues are addressed by the functional linear model for scalar response (SFLM, schematized in Fig. 2a) which is a linear model taking one or several functional variables as exposure to predict a scalar response. This first part of the study is thereafter referred to as ‘Application 1’

Objectives

Methods

Results

Conclusion