Epidemiological studies indicate that exposure to extreme warm and cold temperatures is associated with an increase in mortality by cardiovascular and respiratory diseases. As part of the PHEWE project (Assessment and Prevention of acute health effects of weather conditions in Europe) a preliminary evaluation of the short term health effects of temperature and humidity on mortality have been analysed in 17 European cities (Athens, Barcelona, Bucharest, Budapest, Cracow, Helsinki, Dublin, Ljubljana, London, Milan, Paris, Prague, Rome, Stockholm, Turin, Valencia, Zurich).
All cities provided daily counts of deaths for cardiovascular, cerebrovascular, respiratory and all causes and 3-hourly meteorological data (namely temperature and humidity) retrieved from the nearest airport weather station. Daily indicators (mean, maximum, minimum temperature, dew point, relative humidity) and apparent temperature (a combination of the dry bulb and dew point temperature) were computed for each city. Data on several confounders including other meteorological variables and air pollution variables, and on effect modifiers (variables associated with each city’s population, climate and environment) were also provided. A large variability in climatic characteristics and in mortality was observed in the cities involved. For example, mean temperatures ranged from 5.6°C (Helsinki) to 18.6°C (Valencia) and daily counts of deaths ranged from 6.5 (Ljubljana) to 163 (London).
Separate analyses were performed for hot and cold seasons. An exploratory analysis was based on the transfer function approach combined with the use of the genetic algorithm to determine the proper lags and the dimension of the auto-regressive component. A common GEE model was then specified and fitted to the data from each city. The assumption of no correlation between the observations of separate summers (winters) was considered. As suggested by the transfer function approach, an AR(1) covariance structure was specified for summer (winter). The effect of temperature was included in the model as a regression spline. The model was adjusted for day of the week, holiday, air pollution variables and influenza epidemics.
A lagged effect of up to 3 weeks was observed of the apparent temperature on health events. Variations of shape of the curve for mortality-temperature were observed across cities in relation to latitude and other climatic conditions. The investigators will present current results and in-progress reports.