Searching for the best modeling specification for assessing the effects of temperature and humidity on health: a time series analysis in three European cities
- 467 Downloads
Epidemiological time series studies suggest daily temperature and humidity are associated with adverse health effects including increased mortality and hospital admissions. However, there is no consensus over which metric or lag best describes the relationships. We investigated which temperature and humidity model specification most adequately predicted mortality in three large European cities. Daily counts of all-cause mortality, minimum, maximum and mean temperature and relative humidity and apparent temperature (a composite measure of ambient and dew point temperature) were assembled for Athens, London, and Rome for 6 years between 1999 and 2005. City-specific Poisson regression models were fitted separately for warm (April–September) and cold (October–March) periods adjusting for seasonality, air pollution, and public holidays. We investigated goodness of model fit for each metric for delayed effects up to 13 days using three model fit criteria: sum of the partial autocorrelation function, AIC, and GCV. No uniformly best index for all cities and seasonal periods was observed. The effects of temperature were uniformly shown to be more prolonged during cold periods and the majority of models suggested separate temperature and humidity variables performed better than apparent temperature in predicting mortality. Our study suggests that the nature of the effects of temperature and humidity on mortality vary between cities for unknown reasons which require further investigation but may relate to city-specific population, socioeconomic, and environmental characteristics. This may have consequences on epidemiological studies and local temperature-related warning systems.
KeywordsAmbient temperature Apparent temperature Relative humidity Distributed lags Time-series Mortality
Degrees of freedom
International classification of diseases
Distributed lag nonlinear model
Generalized additive model
Akaike information criterion
Partial autocorrelation function
Conflict of interest
No competing financial interests.
Our work does not involved human subjects or any experiment.
- Allen MP (1997) The problem of multicollinearity. Springer, New York, pp 176–180, Chapter 37, Understanding Regression AnalysisGoogle Scholar
- Anderson GB, Bell ML, Peng RD (2013) Methods to calculate the heat index as an exposure metric in environmental health research. Environ Health Perspect 121:1111–1119Google Scholar
- Bolker Β (2009) Dealing with quasi- models in R. The Comprehensive R Archive Network. http://cran.r-project.org/web/packages/bbmle/vignettes/quasi.pdf. Accessed 28 Nov 2012
- Hajat S, Sheridan SC, Allen MJ, Pascal M, Laaidi K, Yagouti A, Bickis U, Tobias A, Bourque D, Armstrong BG et al (2010) Heat-health warning systems: a comparison of the predictive capacity of different approaches to identifying dangerously hot days. Am J Public Health 100(6):1137–1144CrossRefGoogle Scholar
- Metzger KB, Ito K, Matte TD (2010) Summer heat and mortality in New York City: how hot is too hot? Environ Health Perspect 118(1):80–86Google Scholar
- Muggeo VMR (2008) Segmented: an R package to fit regression models with broken-line relationships. R News 8(1):20–25, http://cran.r-project.org/doc/Rnews/. Accessed 14 August 2013Google Scholar
- R-development Core Team (2012), ISBN 3-900051-07-0,URL http://www.R-project.org.
- Samoli E, Stafoggia M, Rodopoulou S, Ostro B, Declercq C, Alessandrini E et al (2013) Associations between fine and coarse particles and mortality in Mediterranean cities: results from the MED-PARTICLES Project. Environ Health Perspect 121(8):932–938Google Scholar
- WHO (2013) Review of evidence on health aspects of air pollution—REVIHAAP Project Technical Report. WHO Publications, WHO Regional Office for Europe, Copenhagen, DenmarkGoogle Scholar
- Wood SN (2001) mgcv: GAMs and generalized ridge regression for R. R News 1(2):20–25, http://cran.r-project.org/doc/Rnews/. Accessed 15 February 2013Google Scholar