Skip Navigation LinksHome > July 2004 - Volume 15 - Issue 4 > EXCESS DEATHS DURING THE 2003 HEAT WAVE IN FRANCE: WHICH CON...
Epidemiology:
The Sixteenth Conference of the International Society for Environmental Epidemiology (ISEE): Abstracts

EXCESS DEATHS DURING THE 2003 HEAT WAVE IN FRANCE: WHICH CONTRIBUTION OF AIR POLLUTION?

Cassadou, Sylvie*; Le Tertre, Alain*; Medina1, Sylvia*; Eilstein, Daniel*; Declercq, Christophe†; Pascal, Laurence*; Filleul, Laurent*; Lefranc, Agnes‡; Fabre, Pascal*; Prouvost, Hélène†; D'Helf, Myriam*; Jusot, Jean François*; Chardon, Benoit‡; Ledrans, Martine*

Free Access
Article Outline
Collapse Box

Author Information

*InVS, †ORS Nord Pas de Calais; ‡ORS Ile de France

ISEE-298

Back to Top | Article Outline

Introduction:

Hot temperatures that occurred during the summer 2003 in Europe were associated to an excess of thousands deaths. It was estimated that about 15 000 excess deaths occurred during the August 1–20 heat wave in France. Simultaneously, air pollution levels, particularly ozone, were dramatically elevated all over the country. Indeed, patterns of air pollution and meteorological conditions are very closely related. Thus, some of the deaths attributed to the heat wave may have been caused by photochemical and particulate air pollution.

A study was designed to examine the temperature - air pollution - mortality relationships, especially in extreme pollution and weather conditions. This approach presents innovative features: two explanatory variables taken into account and focus on exceptional exposure conditions.

Back to Top | Article Outline

Methods:

Time series design is used to analyze the short-term effects of temperature and air pollution on mortality. Daily values of temperature and ozone concentrations (8 hours daily mean - suburban stations) are collected in the nine biggest French cities for a multi-year period including 2003. Mortality daily counts include all causes of deaths (ICD10 A000-Y98) for all the cities. Statisticians from each city perform Poisson's time-series regression with their air pollution, temperature and mortality own data. Autoregressive Generalized linear models allowing for overdispersion are fitted to data. Short-term confounding factors as influenza outbreaks (national registry), public holidays, school holidays, day of the week and humidity are included in the model. Besides these short-term confounders, we control long term trends and seasons. Smoothing functions (penalized splines) are used to account for non-linear relationships between mortality and confounders. In contrast with previous analyses, temperature is included as an explanatory and not as a confounding variable. Possible interactions between same day of O3 and temperature are taken into account with an interaction term. As part of the diagnostic testing of the models, we test the residuals for autocorrelation, apply Akaïke's information criteria and plot the residuals. Quality control by the coordination center ensures homogeneity of the modeling processes. In order to obtain a quantitative summary of the findings across the nine cities, we use a hierarchical approach giving a pooled regression coefficient corresponding to a weighted mean of the city-specific regression coefficients. The relative contribution of air pollution and weather conditions to the excess deaths observed during the 2003 heat wave in those nine cities will be presented at the ISEE conference.

© 2004 Lippincott Williams & Wilkins, Inc.

Twitter  Facebook

Login

Article Tools

Share