z-logo
Premium
Analysis of health outcome time series data in epidemiological studies
Author(s) -
Touloumi Giota,
Atkinson Richard,
Tertre Alain Le,
Samoli Evangelia,
Schwartz Joel,
Schindler Christian,
Vonk Judith M.,
Rossi Giuseppe,
Saez Marc,
Rabszenko Daniel,
Katsouyanni Klea
Publication year - 2004
Publication title -
environmetrics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.68
H-Index - 58
eISSN - 1099-095X
pISSN - 1180-4009
DOI - 10.1002/env.623
Subject(s) - confounding , covariate , bivariate analysis , poisson regression , air pollution , statistics , econometrics , poisson distribution , health effect , epidemiology , multilevel model , environmental science , environmental health , regression analysis , geography , demography , mathematics , medicine , chemistry , population , sociology , organic chemistry
Abstract Several recent studies have reported significant health effects of air pollution even at low levels of air pollutants. These studies have been criticized for the statistical methods and for inconsistency in results between cities. An important development in air pollution epidemiology has come from multicenter studies. Within the APHEA‐2 project we have developed a statistical methodology to evaluate short‐term health effects of air pollution using data from 30 cities across Europe. For the analysis, a hierarchical modelling approach was adopted and implemented in two stages: (a) data from each city were analyzed separately to allow for local differences, using generalized additive Poisson regression models; (b) city‐specific effects estimates were regressed on city‐specific covariates to obtain an overall estimate and to explore heterogeneity across cities. In order to illustrate our methodology we present results for PM 10 effects. It was found that a 10 μg/m 3 increase in PM 10 or NO 2 concentrations is associated with a 0.67% (95% CI: 0.50 to 0.90) and 0.33% (0.20 to 0.40) increase in total mortality, respectively. After mutual adjustment, the PM 10 effect was reduced by 40% and that of NO 2 by 20%, but both pooled estimates remained significant. Long‐term mean NO 2 concentrations act as an effect modifier for PM 10 effects, even after adjustment for NO 2 confounding effects. In the second stage we explored two different models for combining the adjusted for NO 2 , PM 10 effects across cities: bivariate, which accounts for within‐city correlation of PM 10 and NO 2 ; and univariate, which ignores this correlation. Both models gave broadly the same results. Copyright © 2004 John Wiley & Sons, Ltd.

This content is not available in your region!

Continue researching here.

Having issues? You can contact us here